Career Employer

Your FREE CompTIA Data+ Practice Test 2026 – 200+ Q&A

Prepare with realistic, CompTIA Data+ exam-style questions — take a full DA0-002 practice test or drill one domain.

How ready are you?

To find us again, just search “Career Employer CompTIA Data+”

By
CompTIA Data+ exam by the numbersStandard mode

Where the CompTIA Data+ exam's points come from

The official outline weights, the exam rules and our 204 free practice questions mapped domain by domain. Use it to plan your study hours.

  • Number of questions:Maximum of 90
  • Question types:Multiple-choice and performance-based (PBQs)
  • Administered by:Pearson VUE (in-person test center or OnVUE online proctored)
  • Recertification:Valid 3 years; renew with 30 CEUs
Facts from the official exam outline and rules
90 min
Time limit
675
Passing score
18-24
Recommended experience
~$255
Exam cost

Where the CompTIA Data+ points are

The official weight for each of the five domains. Hover or tap a domain to see how many questions that means on your exam.

See these numbers as a table
Weights from the CompTIA Data+ exam outline. Questions on your exam = weight × 90 questions, rounded. Practice questions: Career Employer's CompTIA Data+ bank.
DomainOfficial weightQuestions on your examOur practice questions
Data Concepts and Environments20%1842
Data Acquisition and Preparation22%2045
Data Analysis24%2246
Visualization and Reporting20%1838
Data Governance14%1333

What to do: Data Analysis is the biggest domain at 24%, which means 22 of your questions. Data Governance is the smallest at 14%.

Career Employer study data · CompTIA Data+ exam

What CompTIA Data+ studiers got right (and wrong) so far

4,232 first-try answers on our free CompTIA Data+ practice questions. Each part below switched on by itself once it had enough people behind it.[6]

Data through Nov 17, 2026 · updates daily
84.3%
accuracy on the second try, up from 74.6%

Where CompTIA Data+ points are lost

Out of 100 questions on a real exam, how many a typical studier would miss in each section. Accuracy alone hides this: a big section with average accuracy costs more points than a small, hard one.

Not shown yet (below our sample-size minimum): Data Governance.

See these numbers as a table
First-try answers from CompTIA Data+ studiers on Career Employer, through Nov 17, 2026. Weights from Pearson VUE's CompTIA Data+ content outline.
SectionExam weightFirst-try accuracyPoints lost per 100First-try answersStudiers
Data Analysis24.4%59.3%9.91,06053
Visualization and Reporting20%70%6.082255
Data Concepts and Environments20%82.6%3.590757
Data Acquisition and Preparation22.2%85%3.391855

Split your study time

Your hours, divided by the domain weights.

Data Concepts and Environments
12 hrs
Data Acquisition and Preparation
13 hrs
Data Analysis
14 hrs
Visualization and Reporting
12 hrs
Data Governance
8.4 hrs

More CompTIA Data+ student data is on its way

We’ve logged 4,232 first-try answers on our CompTIA Data+ questions so far. This section switches to real CompTIA Data+ student results by itself once each part has enough people behind it:

What switches onNeeds
Headline numbers150 studiers
Accuracy by domain300 answers + 50 people per domain
First full exam score100 full exams
Finish rate150 people starting

Then it shows you vs everyone.

How hard are the questions?

192 CompTIA Data+ questions with at least 15 first-try answers, grouped by the share of people who got them right first time.

See these numbers as a table
Every version of each question since Aug 29, 2026, through Nov 17, 2026.
Got it right first tryQuestions
0–10%0
10–20%1
20–30%7
30–40%10
40–50%13
50–60%18
60–70%15
70–80%24
80–90%42
90–100%62

What it means: 31 of 192 questions (16%) trip up more than half of studiers on the first try. That's on purpose: our free CompTIA Data+ bank is built to be hard, so a passing score here means something.[6]

CompTIA Data+ study activity, week by week

How much CompTIA Data+ studying happens on Career Employer each week. Hover or tap a week for details.

See these numbers as a table
Site analytics, week of Jul 6 through Nov 15, 2026.
Week ofQuestions answeredPractice tests startedFlashcard sessions
Nov 9724322
Nov 2462172
Oct 26778320
Oct 19411120
Oct 12314170
Oct 513030
Sep 28574170
Sep 2118021
Sep 14484120
Sep 7233100
Aug 3129683
Aug 2410770
Aug 17132119
Aug 1050012–
Aug 31418–
Jul 273885–
Jul 201242–
Jul 1300–
Jul 600–

Source: anonymous first-try answers on Career Employer CompTIA Data+ practice tests, Aug 29, 2026 – Nov 17, 2026. Staff traffic, repeats and peeked answers excluded. Our practice questions are written to the official outline, not the official exam; self-selected sample; a studier is one browser.

Weights and rules from the published CompTIA Data+ exam outline and certification pages. Practice-question counts from Career Employer's CompTIA Data+ bank. Fees and requirements can change; check the official exam site before you register.

CompTIA Data+ — current exam code DA0-002 (V2) — is an early-career data analytics certification that validates your ability to turn raw data into actionable business insight.[1] These free CompTIA Data+ practice questions mirror the current 2026 content outline so you practice the way the real exam is built.[3] Pair these with our free study guide, flashcards, and cheat sheet.

CompTIA Data+ is one of the 14 CompTIA certifications — explore all our CompTIA practice tests to compare and prep across the whole family.

What Is on the CompTIA Data+ Exam?

The CompTIA Data+ DA0-002 (V2) exam covers five domains: Data Concepts & Environments (20%), Data Acquisition & Preparation (22%), Data Analysis (24%), Visualization & Reporting (20%), and Data Governance (14%).[1]

Data Analysis carries the most weight, and the V2 blueprint rebalanced the older DA0-001 split — renaming Data Mining to Data Acquisition & Preparation and broadening Visualization to Visualization & Reporting.[2] Our full practice test is weighted to match.

CompTIA Data+ practice test — practice questions by domain with answer explanations

Practice Questions by Domain

Use Start Test for a full weighted DA0-002 simulation, or open the hub and pick a single domain to drill your weak area. After each full exam, your results show a per-domain breakdown so you know exactly where to focus — most candidates need the most reps on Data Analysis and the statistics-heavy topics.

What Are the Requirements to Take CompTIA Data+?

CompTIA Data+ has no mandatory prerequisites — anyone can register. CompTIA recommends 18-24 months of experience in a data analyst or similar role, with exposure to databases and analytical tools, a basic understanding of statistics, and hands-on data visualization experience.[3] Earning CompTIA Tech+ or A+ first can help build foundational IT knowledge, but neither is required.

How Do You Register for the CompTIA Data+ Exam?

You register for CompTIA Data+ by purchasing an exam voucher from the CompTIA Store or an authorized reseller, then scheduling your exam through Pearson VUE.[1] You can test in person at a Pearson VUE test center or remotely via OnVUE online proctoring from home.

Online testing requires a webcam, microphone, stable internet, a clean private workspace, and a valid government-issued ID for check-in. Have your voucher number ready when you book.

What Is the Passing Score for CompTIA Data+?

The passing score for CompTIA Data+ is 675 on a scale of 100 to 900.[1] Scores use a weighted, scaled model rather than a simple percentage correct, so not every question carries equal weight.

Performance-based questions (PBQs) typically appear first and can be worth more than standard multiple-choice items. You receive a pass/fail result immediately at the end of your session, along with a score report.

How Hard Is CompTIA Data+? (Pass Rate)

CompTIA does not publish official pass rates for Data+, but industry estimates put the first-attempt pass rate near 70-85%. Most of the challenge comes from the breadth of topics across five domains and from the performance-based questions, rather than from extreme technical depth — candidates often find statistics, data-mining concepts, and visualization best-practice questions the trickiest.

~70-85%
Estimated pass rate
no official figure published
24%
Data Analysis domain
largest section

The takeaway: drill until you’re consistently scoring above target on full-length practice — especially Data Analysis and statistics — before you book your exam date.

On Career Employer, CompTIA Data+ students miss Data Analysis most (59% right on the first try)[6] — see the CompTIA Data+ student data above.

What to Expect on Exam Day

Arrive at your Pearson VUE test center at least 15 minutes early to check in — bring a valid, unexpired government-issued photo ID whose name matches your registration.[1] You’ll store phones and personal items in a locker; no notes are allowed.

A short tutorial precedes the exam, then you have 90 minutes to answer up to 90 multiple-choice and performance-based questions, with PBQs typically appearing first. If you test via OnVUE online proctoring, expect a clean private room, a webcam check, and an ID scan.

You receive a pass/fail result immediately, and CompTIA posts your score report to your account. Having simulated the full timing with practice tests makes that clock feel routine.

What CompTIA Data+ Test-Takers Report

What people who recently sat the CompTIA Data+ said about the real exam in public forums, quoted word for word:

Real exam tougher than practice

“It was a challenging exam, and although I was scoring well in practice tests, the real exam was tougher and tested deeper understanding.”

— u/Maleficent-Collar207r/CompTIA, Aug 2026Passed ✓

80 questions, only one PBQ

“I did my exam today and managed to get a score of 775. I got 80 questions and there was only 1 PBQ.”

— u/wiwiwiad13r/CompTIA, Apr 2026Passed ✓

Missed twice by a narrow margin

“First attempt: 625/900 … Second attempt: 651/900 … What’s frustrating is that I’m literally just a 02 questions away from passing. It feels worse than failing by a big margin because I don’t even know where I’m going wrong.”

— u/Disastrous_Sea1457r/CompTIA, Mar 2026Did not pass this attempt

How to Use This CompTIA Data+ Practice Test

  • Recreate exam conditions. Take the full test timed, with no notes.[4]
  • Diagnose, then drill. Use a full DA0-002 simulation to find weak domains, then drill them.
  • Prioritize Data Analysis + statistics. They’re the biggest score-movers.
  • Learn the why. Read every explanation — understanding beats memorizing.
  • Answer everything. There’s no guessing penalty, so never leave a question blank.

Why Get CompTIA Data+ Certified?

CompTIA Data+ is a vendor-neutral credential that proves you can mine and prepare data, run descriptive and exploratory analysis, build clear visualizations, and apply governance and quality controls — exactly the skills employers want in early-career data analysts.[3] These free CompTIA Data+ practice tests are the most efficient way to get there.

Conclusion

Passing the DA0-002 exam comes down to knowing data concepts, analysis, visualization, and governance cold. Use this free CompTIA Data+ practice test to find your weak domains and drill them to mastery — and round out your prep with our free study guide, flashcards, and cheat sheet. On Career Employer, CompTIA Data+ students lose the most points on Data Analysis (59% correct on the first try), so start your drilling there.[6]

CompTIA Data+ Practice Test FAQ

DA0-002 (V2) is the current version. It launched October 14, 2025. The older DA0-001 (V1) retired in English on April 14, 2026 (with Japanese and Thai versions available until July 16, 2026). If you are testing now, you will take DA0-002, which has updated content on cloud data environments, AI, governance, and modern visualization.

What changed on this page

Every real change, with the date.

  1. Interactive charts added. Built from the official outline and our question bank.
  2. 204 questions rewritten.
  3. 1 question corrected.
  4. 99 new practice questions. 2 questions retired. 1 answer key corrected. The bank now has 204 questions.
Career Employer CompTIA Data+ practice-test data, through Nov 17, 2026 · 57+ students
Every published Career Employer CompTIA Data+ practice-test number, with its sample size, source and date
MetricValuenStudentsSourceData through
First-try accuracy: Data Analysis (24.4% of the exam; costs 9.9 of every 100 exam points)59.3%1,060 answers53current question setNov 17, 2026
First-try accuracy: Visualization and Reporting (20% of the exam; costs 6 of every 100 exam points)70%822 answers55current question setNov 17, 2026
First-try accuracy: Data Concepts and Environments (20% of the exam; costs 3.5 of every 100 exam points)82.6%907 answers57current question setNov 17, 2026
First-try accuracy: Data Acquisition and Preparation (22.2% of the exam; costs 3.3 of every 100 exam points)85%918 answers55current question setNov 17, 2026

Sections below our sample-size minimum (not published yet): Data Governance.

First attempt at each question only; repeats, answers after revealing the explanation, bots and staff excluded. Aug 29, 2026 – Nov 17, 2026. Our practice questions written to the official outline, not the official exam; self-selected sample; a student is one browser. Free to reuse under CC BY 4.0 — cite “Career Employer practice-test data, careeremployer.com/data”.

CompTIA Data+ question bank

All 204 questions, by domain

A reference copy of every question in this practice test. Each answer stays hidden until you choose to show it. To practice with scoring, timing and your readiness score, use Start Test at the top of the page.

Data Concepts and Environments (42)

  1. In data analytics, what does the term "Data Lake" primarily refer to?

    • A.A repository that holds clean copies of source data in its common format
    • B.A repository that holds massive volumes of raw data in its native format
    • C.A repository that holds old files of inactive data in its compact format
    • D.A repository that holds small slices of sales data in its summary format
    Show answerHide answer

    Correct answer: A repository that holds massive volumes of raw data in its native format

    "A repository that holds massive volumes of raw data in its native format" is correct: a data lake stores structured, semi-structured, and unstructured data at very large scale without transforming it first, and a schema is applied only when the data is read (schema-on-read). "A repository that holds clean copies of source data in its common format" is incorrect: that describes a data warehouse, which cleanses and integrates source data into a predefined, conformed schema before loading it (schema-on-write). "A repository that holds old files of inactive data in its compact format" is incorrect: that describes a data archive, which moves rarely used data to low-cost, often compressed storage for retention; a data lake keeps raw data available for active analysis. "A repository that holds small slices of sales data in its summary format" is incorrect: that describes a data mart, a subject-oriented subset of warehouse data (such as sales) that is usually summarized for one department or business line.

  2. What is the primary purpose of a Data Warehouse?

    • A.To process operational transactions and data updates
    • B.To collect unstructured documents and raw recordings
    • C.To maintain redundant backups and recovery snapshots
    • D.To support analytical reports and structured queries
    Show answerHide answer

    Correct answer: To support analytical reports and structured queries

    "To support analytical reports and structured queries" is correct: a data warehouse is a central repository of integrated, historical data that is modeled in advance (schema-on-write) so analysts can run reports and structured queries efficiently. "To process operational transactions and data updates" is incorrect: that is the job of an OLTP (online transaction processing) database, which is optimized for many small inserts and updates rather than for analysis. "To collect unstructured documents and raw recordings" is incorrect: keeping raw, unstructured content in its native format is the role of a data lake; a warehouse stores cleansed, structured data. "To maintain redundant backups and recovery snapshots" is incorrect: backup and recovery copies protect data against loss; a warehouse is built for analysis, not as a backup system.

  3. In the context of data analytics, what is "Big Data" characterized by?

    • A.The complex and variable nature of data
    • B.The clean and structured nature of data
    • C.The static and aggregate nature of data
    • D.The personal and private nature of data
    Show answerHide answer

    Correct answer: The complex and variable nature of data

    "The complex and variable nature of data" is correct: Big Data describes data sets too large or complex for traditional tools, usually summarized by volume, velocity, and variety, with variability often added; its complexity and variability are defining traits. "The clean and structured nature of data" is incorrect: clean, structured data fits neatly into the rows and columns of a traditional relational database; Big Data is defined by large volume and a messy mix of structured, semi-structured, and unstructured formats. "The static and aggregate nature of data" is incorrect: Big Data usually arrives continuously at high velocity and at a detailed level; static, pre-aggregated figures describe summary tables in a reporting system. "The personal and private nature of data" is incorrect: personal and private data raise privacy and compliance concerns, but sensitivity is a classification of data, not what characterizes Big Data; much Big Data, such as sensor and log data, is not personal at all.

  4. In a data environment, what does "Metadata" refer to?

    • A.Data that describes other data
    • B.Data that is aggregated from other data
    • C.Data that repeats other data
    • D.Data that is derived from other data
    Show answerHide answer

    Correct answer: Data that describes other data

    "Data that describes other data" is correct: metadata is data about data, such as a file's author, creation date, format, and size, or a column's name, type, and source, which explains what the underlying data is and how to use it. "Data that is aggregated from other data" is incorrect: totals and other aggregates are summary data calculated from the values themselves; they report the content of the data rather than describing its characteristics. "Data that repeats other data" is incorrect: duplicate or backup copies are redundant data kept for recovery or availability; they repeat the data instead of describing it. "Data that is derived from other data" is incorrect: derived data, such as a profit column calculated from revenue and cost, is produced by formulas from existing values; it is new data, not a description of data.

  5. Which type of database is optimized for retrieving and organizing large amounts of unstructured data?

    • A.Relational Database
    • B.Analytical Database
    • C.Spatial Database
    • D.NoSQL Database
    Show answerHide answer

    Correct answer: NoSQL Database

    "NoSQL Database" is correct: NoSQL databases (document, key-value, column-family, and graph stores) use flexible or schema-less models and scale out across many servers, which suits storing, retrieving, and organizing large volumes of unstructured and semi-structured data. "Relational Database" is incorrect: relational databases store structured data in tables with a predefined schema and are best for consistent, transactional data, not large volumes of unstructured content. "Analytical Database" is incorrect: analytical databases, often columnar, are tuned for fast aggregate queries over structured, modeled data such as a warehouse, not for organizing unstructured content. "Spatial Database" is incorrect: spatial databases specialize in geographic and geometric data such as points, lines, and polygons; they are not designed for large amounts of general unstructured data.

  6. What is an OLAP (Online Analytical Processing) system primarily used for in a data environment?

    • A.Simple inserts and frequent updates
    • B.Scheduled extracts and staged loads
    • C.Complex queries and ad-hoc analysis
    • D.Periodic backups and quick restores
    Show answerHide answer

    Correct answer: Complex queries and ad-hoc analysis

    "Complex queries and ad-hoc analysis" is correct: OLAP systems organize data multidimensionally, as cubes of dimensions and measures, so analysts can run complex queries, slice and dice, drill down, and explore data ad hoc, usually on data warehouse data. "Simple inserts and frequent updates" is incorrect: large numbers of simple inserts and updates are the workload of OLTP (online transaction processing) systems, which record day-to-day business transactions. "Scheduled extracts and staged loads" is incorrect: scheduled extracts and staged loads are ETL tasks that populate the warehouse an OLAP system later queries. "Periodic backups and quick restores" is incorrect: backups and restores protect data against loss; they are database administration tasks, not analytical processing.

  7. Which technology is primarily used for distributed processing of large data sets across clusters of computers?

    • A.Talend
    • B.Pandas
    • C.Hadoop
    • D.Looker
    Show answerHide answer

    Correct answer: Hadoop

    "Hadoop" is correct: Apache Hadoop stores data across many machines with HDFS and processes it in parallel (originally with MapReduce, with YARN managing cluster resources), so very large data sets can be processed across clusters of commodity computers. "Talend" is incorrect: Talend is an ETL and data integration tool for extracting, transforming, and loading data; it can hand jobs to a cluster engine, but distributed processing is not its primary purpose. "Pandas" is incorrect: pandas is a Python library for working with data frames in the memory of a single machine; it does not distribute processing across a cluster. "Looker" is incorrect: Looker is a business intelligence and visualization platform that queries existing databases to build reports and dashboards; it does not perform distributed processing itself.

  8. What is the main characteristic of "semi-structured" data?

    • A.It follows a predefined schema but accepts nulls or blanks to denote gaps
    • B.It takes a free form but needs algorithms or analysts to extract insights
    • C.It lacks a formal structure but uses tags or markers to separate elements
    • D.It keeps a structured design but repeats columns or values to avoid joins
    Show answerHide answer

    Correct answer: It lacks a formal structure but uses tags or markers to separate elements

    "It lacks a formal structure but uses tags or markers to separate elements" is correct: semi-structured data such as JSON, XML, and email does not fit a fixed relational schema, but it carries its own tags, keys, or markers that separate data elements and express their hierarchy. "It follows a predefined schema but accepts nulls or blanks to denote gaps" is incorrect: data that follows a predefined schema is structured data, even when some values are null; semi-structured data has no fixed schema at all. "It takes a free form but needs algorithms or analysts to extract insights" is incorrect: that describes unstructured data, such as free text, images, and audio, which has no organizing tags and must be interpreted before it can be analyzed. "It keeps a structured design but repeats columns or values to avoid joins" is incorrect: that describes denormalized relational data, which is still structured; repeating values to avoid joins is a performance choice, not a sign of semi-structure.

  9. What does the term "Data Mart" refer to in a data environment?

    • A.A store of integrated data covering many products or regions
    • B.A subset of warehouse data serving one subject or department
    • C.A pool of raw data holding documents or recordings unchanged
    • D.A copy of operational data showing current sales or balances
    Show answerHide answer

    Correct answer: A subset of warehouse data serving one subject or department

    "A subset of warehouse data serving one subject or department" is correct: a data mart is a smaller, subject-oriented subset of a data warehouse built for one business line or department, such as sales or finance, so its users get only the data they need. "A store of integrated data covering many products or regions" is incorrect: that describes the enterprise data warehouse itself, which integrates data across the whole organization; a mart is a focused slice of it. "A pool of raw data holding documents or recordings unchanged" is incorrect: that describes a data lake, which keeps raw data of any type in its native format. "A copy of operational data showing current sales or balances" is incorrect: that describes an operational data store (ODS), which integrates current operational data for day-to-day reporting rather than serving as a subject-specific analytical subset.

  10. In data management, what is "Schema-on-Read" primarily associated with?

    • A.Data Marts
    • B.Data Cubes
    • C.Data Lakes
    • D.Data Vault
    Show answerHide answer

    Correct answer: Data Lakes

    "Data Lakes" is correct: schema-on-read means raw data is stored as it arrives and a structure is applied only when the data is read or queried, which is how data lakes handle large volumes of varied data. "Data Marts" is incorrect: a data mart is a subject-oriented subset of a warehouse whose tables are modeled before loading, so it follows schema-on-write. "Data Cubes" is incorrect: an OLAP data cube is built from predefined dimensions and measures, so its structure is fixed before any data is loaded. "Data Vault" is incorrect: Data Vault is a warehouse modeling method (hubs, links, and satellites) that loads data into a defined model, which is schema-on-write.

  11. A retail analyst is choosing a storage system for nightly sales reporting that aggregates several years of cleaned, integrated transaction history from many source systems. Which storage system is purpose-built for this kind of enterprise-wide analytical reporting?

    • A.A data mart
    • B.A data warehouse
    • C.A data lake
    • D.A data dashboard
    Show answerHide answer

    Correct answer: A data warehouse

    A data warehouse is correct: it is built to hold cleaned, integrated, historical data from many source systems so the whole organization can run analytical queries and reports across years of history. A data mart is incorrect: a mart is a smaller subject-specific slice, usually for one department, rather than the enterprise-wide store described. A data lake is incorrect: a lake keeps raw data in its native formats and applies structure only when it is read, so it is not the cleaned, integrated store this reporting needs. A data dashboard is incorrect: a dashboard displays metrics drawn from a data source; it is a reporting front end, not a storage system that integrates years of history.

  12. A marketing team wants its own narrow slice of the company data warehouse containing only campaign and lead data so its analysts can query quickly without sifting through unrelated subjects. What is this subject-focused subset called?

    • A.A data warehouse
    • B.A data lake
    • C.A data lakehouse
    • D.A data mart
    Show answerHide answer

    Correct answer: A data mart

    A data mart is correct: a data mart is a subject-oriented subset of a data warehouse built for one business area, such as marketing, so that team can query its own data quickly without the rest of the warehouse in the way. A data warehouse is incorrect: the warehouse is the enterprise-wide store the team is drawing from, not the narrow subject-focused slice it wants. A data lake is incorrect: a data lake holds raw data of many formats for the whole organization rather than a curated slice for one team. A data lakehouse is incorrect: a lakehouse combines low-cost lake storage with warehouse-style tables and management for the whole organization; it is a platform, not a subject-focused subset for one department.

  13. An analyst is asked to explain how a data warehouse differs from a data mart to a new team member. Which statement most accurately captures the relationship?

    • A.A data warehouse is an isolated departmental extract, while a data mart is a central organization-wide repository spanning several source systems
    • B.A data warehouse is a landing zone holding raw, unprocessed source files, while a data mart is a cleaned query-ready reporting layer
    • C.A data warehouse is an enterprise-wide integrated store, while a data mart is a smaller subject-specific subset commonly serving one department
    • D.A data warehouse is a high-level summary of quarterly sales totals, while a data mart is a detailed line-item transaction history
    Show answerHide answer

    Correct answer: A data warehouse is an enterprise-wide integrated store, while a data mart is a smaller subject-specific subset commonly serving one department

    A data warehouse is an enterprise-wide integrated store, while a data mart is a smaller subject-specific subset commonly serving one department is correct: the warehouse integrates data from across the organization, and a mart narrows that data to one subject area for a particular team. A data warehouse is an isolated departmental extract, while a data mart is a central organization-wide repository spanning several source systems is incorrect: it reverses the two, because the warehouse is the broad organization-wide store and the mart is the narrow slice. A data warehouse is a landing zone holding raw, unprocessed source files, while a data mart is a cleaned query-ready reporting layer is incorrect: a landing zone for raw files describes a data lake or staging area, whereas a warehouse holds cleaned, integrated data. A data warehouse is a high-level summary of quarterly sales totals, while a data mart is a detailed line-item transaction history is incorrect: a warehouse keeps detailed, granular history rather than just summaries, and marts are often more summarized than the warehouse they are drawn from.

  14. A company stores massive volumes of raw clickstream logs, sensor readings, images, and JSON documents in their native formats, applying structure only when data is read for a specific analysis. Which environment is described?

    • A.A data mart
    • B.A data cube
    • C.A data silo
    • D.A data lake
    Show answerHide answer

    Correct answer: A data lake

    A data lake is correct: a lake stores large volumes of raw structured, semi-structured and unstructured data in native formats and applies a schema only when the data is read for a specific analysis (schema-on-read). A data mart is incorrect: a mart is a curated, structured subset of a warehouse for one subject area. A data cube is incorrect: a cube is a pre-aggregated multidimensional structure used by OLAP tools and built from structured data, not a store for raw files of many formats. A data silo is incorrect: a silo is data isolated within one department or system; the term describes poor integration, not an architecture for storing raw multi-format data.

  15. An analyst must explain the core difference between a data warehouse and a data lake. Which distinction is correct?

    • A.A data warehouse loads raw log files under schema-on-read, while a data lake validates inbound multi-source records under schema-on-write
    • B.A data warehouse stores curated structured records under schema-on-write, while a data lake keeps raw multi-format files under schema-on-read
    • C.A data warehouse captures live transactions under strict normalization, while a data lake archives final reports under formal version control
    • D.A data warehouse publishes shared tables openly under minimal governance, while a data lake restricts access under strict compliance controls
    Show answerHide answer

    Correct answer: A data warehouse stores curated structured records under schema-on-write, while a data lake keeps raw multi-format files under schema-on-read

    A data warehouse stores curated structured records under schema-on-write, while a data lake keeps raw multi-format files under schema-on-read is correct: a warehouse cleans and models data to a schema defined before loading, while a lake accepts raw data of any format and applies structure only when the data is read. A data warehouse loads raw log files under schema-on-read, while a data lake validates inbound multi-source records under schema-on-write is incorrect: it swaps the two approaches, since warehouses validate and model data on the way in and lakes take raw files and interpret them later. A data warehouse captures live transactions under strict normalization, while a data lake archives final reports under formal version control is incorrect: capturing live transactions in normalized tables is the job of an OLTP database, and a lake is a store of raw source data rather than an archive of finished reports. A data warehouse publishes shared tables openly under minimal governance, while a data lake restricts access under strict compliance controls is incorrect: warehouses are among the most tightly governed data stores, and lakes are usually the less curated environment, so the governance claims are reversed.

  16. An analyst is told a system handles thousands of small, fast inserts and updates such as recording each individual customer purchase as it happens. Which type of system is this?

    • A.OLAP (online analytical processing)
    • B.BI (business intelligence) reporting
    • C.BPM (business process management)
    • D.OLTP (online transaction processing)
    Show answerHide answer

    Correct answer: OLTP (online transaction processing)

    OLTP (online transaction processing) is correct: OLTP systems handle very large numbers of short, fast inserts and updates, such as recording each purchase as it happens, and are usually highly normalized to protect data integrity. OLAP (online analytical processing) is incorrect: OLAP supports complex, read-heavy analysis of aggregated historical data, not rapid individual writes. BI (business intelligence) reporting is incorrect: BI reporting reads data that has already been captured and summarizes it in reports and dashboards; it does not record each purchase as it occurs. BPM (business process management) is incorrect: BPM is a discipline for modeling and improving business workflows, not a type of system for capturing individual transactions.

  17. A business intelligence team runs large, complex, ad hoc queries that scan millions of historical rows to analyze sales trends but rarely modify any data. Which processing type best describes this workload?

    • A.OLTP (online transaction processing)
    • B.OLAP (online analytical processing)
    • C.TPS (transaction processing systems)
    • D.EDP (electronic data processing)
    Show answerHide answer

    Correct answer: OLAP (online analytical processing)

    OLAP (online analytical processing) is correct: OLAP is optimized for large, complex, read-heavy queries over historical data, such as scanning millions of rows to find sales trends, while the data itself is rarely changed. OLTP (online transaction processing) is incorrect: OLTP handles many small inserts and updates of current records, the opposite workload. TPS (transaction processing systems) is incorrect: a transaction processing system records day-to-day business transactions such as orders and payments, which is again a write-heavy operational workload rather than ad hoc analysis. EDP (electronic data processing) is incorrect: EDP is the older general term for automated, repetitive processing of business records, such as applying payments to accounts, not for exploratory analytical queries over years of history.

  18. Which pairing correctly contrasts OLTP and OLAP systems?

    • A.OLTP executes lengthy analytical scans and is typically denormalized; OLAP processes numerous quick insertions and is typically normalized
    • B.OLTP favors compressed columnar formats and is typically refreshed nightly; OLAP prefers row-oriented layouts and is typically updated live
    • C.OLTP serves internal business analysts and is typically refreshed nightly; OLAP serves online retail shoppers and is typically updated live
    • D.OLTP processes frequent short transactions and is typically normalized; OLAP runs complex analytical queries and is typically denormalized
    Show answerHide answer

    Correct answer: OLTP processes frequent short transactions and is typically normalized; OLAP runs complex analytical queries and is typically denormalized

    OLTP processes frequent short transactions and is typically normalized; OLAP runs complex analytical queries and is typically denormalized is correct: OLTP systems process large numbers of short insert and update transactions and are normalized to protect integrity and avoid redundancy, while OLAP systems run complex read-heavy analytical queries and are denormalized, for example into star schemas, so queries need fewer joins. OLTP executes lengthy analytical scans and is typically denormalized; OLAP processes numerous quick insertions and is typically normalized is incorrect: it reverses the two, since long analytical scans and denormalization belong to OLAP, while quick inserts and normalization belong to OLTP. OLTP favors compressed columnar formats and is typically refreshed nightly; OLAP prefers row-oriented layouts and is typically updated live is incorrect: OLTP databases usually store data row by row and change continuously, while analytical systems favor columnar storage and are commonly refreshed by scheduled batch loads. OLTP serves internal business analysts and is typically refreshed nightly; OLAP serves online retail shoppers and is typically updated live is incorrect: OLTP serves the customers and operational staff who create transactions as they happen, and OLAP serves the analysts who study them, so both the audiences and the refresh patterns are reversed.

  19. A data engineer designs a warehouse with a central fact table holding sales measures, connected directly to denormalized dimension tables for product, date, and region. Which schema design is this?

    • A.A snowflake schema
    • B.A galaxy schema
    • C.A star schema
    • D.A federated schema
    Show answerHide answer

    Correct answer: A star schema

    A star schema is correct: a star schema places one central fact table, here holding sales measures, directly beside denormalized dimension tables for product, date and region, so the layout looks like a star. A snowflake schema is incorrect: a snowflake schema normalizes the dimensions into further sub-tables, such as a product table linked to a separate category table, which adds joins. A galaxy schema is incorrect: a galaxy, or fact constellation, schema has several fact tables sharing dimension tables, not one central fact table. A federated schema is incorrect: a federated schema is a combined view over several separate databases that are queried together, not a warehouse layout built around a fact table.

  20. A warehouse design normalizes its dimension tables into multiple related sub-tables, so a product dimension links out to separate category and supplier tables. Which schema is being used?

    • A.A federated schema
    • B.A snowflake schema
    • C.A star schema
    • D.A galaxy schema
    Show answerHide answer

    Correct answer: A snowflake schema

    A snowflake schema is correct: a snowflake schema normalizes dimension tables into related sub-tables, so a product dimension links out to separate category and supplier tables, creating a branching shape with more joins. A federated schema is incorrect: a federated schema combines several independent databases into one queryable view; it does not describe splitting one dimension into sub-tables. A star schema is incorrect: a star schema keeps each dimension in one denormalized table joined directly to the fact table, so a product dimension would hold its category and supplier details itself. A galaxy schema is incorrect: a galaxy schema links several fact tables to shared dimensions; it does not describe splitting one dimension into sub-tables.

  21. A data warehouse must keep both the prior and the current value of a customer's address by adding a new row each time the address changes while preserving the old row. Which slowly changing dimension approach is this?

    • A.Type 1 (overwrite the stored value, keep current)
    • B.Type 3 (add a field, keep previous)
    • C.Type 0 (ignore the incoming value, keep original)
    • D.Type 2 (add a new row, keep history)
    Show answerHide answer

    Correct answer: Type 2 (add a new row, keep history)

    Type 2 (add a new row, keep history) is correct: a Type 2 slowly changing dimension adds a new row each time the attribute changes, usually with effective dates or a current-row flag, and keeps the old row, so both the prior and the current address are preserved. Type 1 (overwrite the stored value, keep current) is incorrect: Type 1 overwrites the existing value, so only the current address is kept and the prior address is lost. Type 3 (add a field, keep previous) is incorrect: Type 3 adds a column to the existing row to hold the previous value, rather than adding a new row for each change. Type 0 (ignore the incoming value, keep original) is incorrect: Type 0 retains the original value and ignores later changes, so the new address would never be stored.

  22. A dimension table is configured so that whenever an attribute changes, the new value overwrites the old one and no history is retained. Which slowly changing dimension type is in use?

    • A.Type 2 (add a new row, keep history)
    • B.Type 3 (store prior values, keep current)
    • C.Type 1 (overwrite values, keep current)
    • D.Type 4 (add mini-dimension, keep history)
    Show answerHide answer

    Correct answer: Type 1 (overwrite values, keep current)

    Type 1 (overwrite values, keep current) is correct: a Type 1 slowly changing dimension overwrites the old attribute value with the new one, so only the current state remains and no history is retained. Type 2 (add a new row, keep history) is incorrect: Type 2 preserves history by adding a new row for each change. Type 3 (store prior values, keep current) is incorrect: Type 3 keeps the previous value in an extra column, so some history is retained. Type 4 (add mini-dimension, keep history) is incorrect: Type 4 splits rapidly changing attributes into a mini-dimension referenced from each fact row, so history is still retained.

  23. An organization stores customer records in tables with predefined columns, enforced data types, and relationships joined by keys. Which database category is described?

    • A.A relational database
    • B.A document database
    • C.A wide-column database
    • D.A time-series database
    Show answerHide answer

    Correct answer: A relational database

    A relational database is correct: it stores records in tables with predefined columns and enforced data types, and it links tables through primary and foreign keys. A document database is incorrect: a document database stores flexible JSON-like documents whose fields can vary from record to record, without a fixed table schema. A wide-column database is incorrect: a wide-column store such as Cassandra groups data into column families whose columns can differ from row to row and does not enforce relationships through foreign keys. A time-series database is incorrect: a time-series database is optimized for timestamped measurements such as sensor readings, not for customer records linked across related tables.

  24. A startup needs to store rapidly changing, schema-flexible JSON documents that vary from record to record without a fixed table layout. Which database type best fits?

    • A.A non-relational (NoSQL) database
    • B.A relational (SQL-based) database
    • C.A distributed (NewSQL) database
    • D.A dimensional (OLAP) database
    Show answerHide answer

    Correct answer: A non-relational (NoSQL) database

    A non-relational (NoSQL) database is correct: NoSQL document stores accept records whose structure varies, such as JSON documents with different fields, without a fixed table layout, and they adapt easily as that structure changes. A relational (SQL-based) database is incorrect: a relational database requires a predefined schema of tables and typed columns, so highly variable documents fit it poorly. A distributed (NewSQL) database is incorrect: NewSQL systems scale out across many servers but keep the relational model and its fixed schemas. A dimensional (OLAP) database is incorrect: a dimensional database organizes cleaned data into facts and dimensions for analysis, not schema-flexible operational documents.

  25. An analyst classifies the field that stores values like 'pending', 'shipped', or 'delivered' for an order. Which data type best describes these labeled categories with no inherent numeric meaning?

    • A.Categorical (dimension) data
    • B.Discrete (whole-number) data
    • C.Currency (monetary) data
    • D.Date (calendar-based) data
    Show answerHide answer

    Correct answer: Categorical (dimension) data

    Categorical (dimension) data is correct: values such as 'pending', 'shipped' and 'delivered' are labels that place each order into a group and carry no numeric magnitude, which is what categorical, or dimension, data means. Discrete (whole-number) data is incorrect: discrete data is countable numbers, such as the number of items in an order, and status labels are not counts. Currency (monetary) data is incorrect: currency data stores money amounts with fixed decimal precision. Date (calendar-based) data is incorrect: date data stores calendar values such as an order date, not status labels.

  26. A field records the exact temperature of a process, which can take any value such as 72.4 or 72.41 degrees within a range. How should an analyst classify this variable?

    • A.A continuous variable
    • B.A discrete variable
    • C.A categorical variable
    • D.A dichotomous variable
    Show answerHide answer

    Correct answer: A continuous variable

    A continuous variable is correct: a continuous variable can take any value within a range, including fractions such as 72.4 or 72.41 degrees, which describes a measured temperature. A discrete variable is incorrect: a discrete variable takes separate countable values, such as a number of items, with nothing in between. A categorical variable is incorrect: a categorical variable holds labels or groups rather than measured quantities. A dichotomous variable is incorrect: a dichotomous variable has only two possible categories, such as yes or no, while a temperature can take infinitely many values.

  27. A column stores the number of items in each shopping cart, which can only be whole counts like 0, 1, or 5. Which classification applies to this variable?

    • A.A discrete variable
    • B.A continuous variable
    • C.A categorical variable
    • D.A dichotomous variable
    Show answerHide answer

    Correct answer: A discrete variable

    A discrete variable is correct: the number of items in a cart is a count that can only take whole values such as 0, 1 or 5, with no values in between. A continuous variable is incorrect: a continuous variable can take any value within a range, including fractions, and a cart cannot hold 2.5 items. A categorical variable is incorrect: a categorical variable holds labels or groups, while an item count is a numeric quantity. A dichotomous variable is incorrect: a dichotomous variable has only two possible categories, such as yes or no, while a cart count can take many values.

  28. A data analyst must store product codes such as 'A12B' and 'X9Z' that mix letters and numbers and are never used in calculations. Which data type is most appropriate?

    • A.Alphabetic
    • B.Numeric
    • C.Floating-point
    • D.Alphanumeric
    Show answerHide answer

    Correct answer: Alphanumeric

    Alphanumeric is correct: alphanumeric data mixes letters and digits and is treated as text rather than as a number, which fits product codes such as 'A12B' and 'X9Z' that are never used in calculations. Alphabetic is incorrect: alphabetic data contains letters only, and these codes also contain digits. Numeric is incorrect: a numeric type holds values used in arithmetic, and these codes contain letters and are never calculated. Floating-point is incorrect: a floating-point type stores numbers that may include a fractional part, so it cannot hold the letters in a code such as 'A12B'.

  29. An analyst needs to store monetary amounts so that values keep a fixed two-decimal precision and a currency context for financial reporting. Which data type is designed for this?

    • A.Discrete
    • B.Varchar
    • C.Currency
    • D.Integer
    Show answerHide answer

    Correct answer: Currency

    Currency is correct: a currency data type stores monetary amounts with a fixed number of decimal places, usually two, and a currency context, which avoids the rounding errors of general floating-point numbers. Discrete is incorrect: discrete describes countable values such as item counts; it is not a storage type built for money. Integer is incorrect: an integer type holds whole numbers only, so it cannot keep the two decimal places of an amount such as 19.99 without extra scaling, and it carries no currency context. Varchar is incorrect: a varchar column stores characters as text, so amounts would have to be converted before any reliable arithmetic or aggregation.

  30. An analyst is told a dataset contains structured data. Which example best fits the definition of structured data?

    • A.A folder of scanned invoices and receipts with handwritten notes
    • B.A news feed of tags and attributes with optional nested elements
    • C.A relational table of rows and columns with enforced field types
    • D.A media library of recordings and videos with spoken commentary
    Show answerHide answer

    Correct answer: A relational table of rows and columns with enforced field types

    A relational table of rows and columns with enforced field types is correct: structured data follows a predefined model, such as rows and columns with defined data types in a relational table, which makes it easy to query. A folder of scanned invoices and receipts with handwritten notes is incorrect: scanned images and handwriting are unstructured data with no queryable schema. A news feed of tags and attributes with optional nested elements is incorrect: tag-based feeds such as XML are semi-structured, because their tags describe the data but the structure can vary from one record to the next. A media library of recordings and videos with spoken commentary is incorrect: audio and video files are unstructured data.

  31. Which statement correctly contrasts structured and unstructured data?

    • A.Structured data follows a predefined tabular format; unstructured data (such as imagery, audio, and free text) lacks a fixed schema
    • B.Structured data takes a free-form file layout; unstructured data (such as ledgers, invoices, and tariffs) uses a rigid typed grid
    • C.Structured data is a numeric record; unstructured data (such as customer names, comments, and street addresses) is a written record
    • D.Structured data is a machine-generated record; unstructured data (such as notes, letters, and chat posts) is a human-written record
    Show answerHide answer

    Correct answer: Structured data follows a predefined tabular format; unstructured data (such as imagery, audio, and free text) lacks a fixed schema

    Structured data follows a predefined tabular format; unstructured data (such as imagery, audio, and free text) lacks a fixed schema is correct: structured data conforms to a defined model of rows, columns and data types, while unstructured content such as images, audio, video and free-form text has no fixed schema. Structured data takes a free-form file layout; unstructured data (such as ledgers, invoices, and tariffs) uses a rigid typed grid is incorrect: it reverses the two, because tabular data with typed columns is the hallmark of structured data. Structured data is a numeric record; unstructured data (such as customer names, comments, and street addresses) is a written record is incorrect: structured data often holds text, such as names and addresses in table columns, so the difference is the predefined model, not numbers versus words. Structured data is a machine-generated record; unstructured data (such as notes, letters, and chat posts) is a human-written record is incorrect: both kinds can come from machines or from people, for example sensor tables and handwritten forms, so the source does not define the category.

  32. An analyst describes JSON and XML as 'semi-structured' rather than fully structured. Why do these formats fall into the semi-structured category?

    • A.They archive data using plain prose or images but do not include a single organizing marker
    • B.They describe data using tags or key-value pairs but do not require a strict tabular schema
    • C.They record data using blank or missing entries but do not support a complete column layout
    • D.They encode data using binary or compressed blocks but do not permit a readable text format
    Show answerHide answer

    Correct answer: They describe data using tags or key-value pairs but do not require a strict tabular schema

    They describe data using tags or key-value pairs but do not require a strict tabular schema is correct: JSON and XML carry markers such as tags or key-value pairs that describe the data, yet they do not force every record into fixed columns, which places them between structured tables and unstructured content. They archive data using plain prose or images but do not include a single organizing marker is incorrect: content with no organizing markers at all is unstructured, while JSON and XML are full of tags and keys. They record data using blank or missing entries but do not support a complete column layout is incorrect: missing values can occur in any kind of data and are not what makes a format semi-structured. They encode data using binary or compressed blocks but do not permit a readable text format is incorrect: JSON and XML are plain, human-readable text formats rather than binary encodings.

  33. Which set of examples are all considered semi-structured data?

    • A.JSON documents, XML files, and email messages with headers
    • B.SQL tables, CSV extracts, and spreadsheet rows with totals
    • C.WAV files, JPEG photos, and video clips with soundtracks
    • D.CSV extracts, PDF scans, and server logs with timestamps
    Show answerHide answer

    Correct answer: JSON documents, XML files, and email messages with headers

    JSON documents, XML files, and email messages with headers is correct: each of these carries organizing markers, such as keys, tags or header fields, without a rigid table schema, which is what makes data semi-structured. SQL tables, CSV extracts, and spreadsheet rows with totals is incorrect: these follow fixed rows and columns, so they are structured data. WAV files, JPEG photos, and video clips with soundtracks is incorrect: audio, images and video are unstructured data. CSV extracts, PDF scans, and server logs with timestamps is incorrect: the set is mixed, because CSV extracts are structured and scanned PDFs are unstructured, even though log files are semi-structured, so not every example is semi-structured.

  34. A developer receives data in a lightweight text format that organizes information as nested key-value pairs inside braces, commonly used for web APIs. Which file format is this?

    • A.XML (Extensible Markup Language)
    • B.EDI (Electronic Data Interchange)
    • C.RSS (Really Simple Syndication)
    • D.JSON (JavaScript Object Notation)
    Show answerHide answer

    Correct answer: JSON (JavaScript Object Notation)

    JSON (JavaScript Object Notation) is correct: JSON is a lightweight, human-readable text format that organizes data as key-value pairs and nested objects inside braces, and it is the most common format for web API responses. XML (Extensible Markup Language) is incorrect: XML also nests data, but it uses opening and closing tags in angle brackets rather than key-value pairs in braces. EDI (Electronic Data Interchange) is incorrect: EDI is a set of standards for exchanging business documents, such as purchase orders, as delimited segments, not nested key-value objects. RSS (Really Simple Syndication) is incorrect: RSS is an XML-based web feed format that marks up items with tags, not a format of key-value pairs inside braces.

  35. An analyst opens a file that uses opening and closing tags like <customer> and </customer> to define and nest data elements. Which file format is this?

    • A.JSON (JavaScript Object Notation)
    • B.XML (Extensible Markup Language)
    • C.EDI (Electronic Data Interchange)
    • D.PDF (Portable Document Format)
    Show answerHide answer

    Correct answer: XML (Extensible Markup Language)

    XML (Extensible Markup Language) is correct: XML uses paired opening and closing tags, such as <customer> and </customer>, that the author defines to describe and nest data elements, which makes it a self-describing semi-structured format. JSON (JavaScript Object Notation) is incorrect: JSON also nests data, but it uses braces and key-value pairs rather than angle-bracket tags. EDI (Electronic Data Interchange) is incorrect: EDI exchanges business documents as delimited segments and elements, not as tagged markup. PDF (Portable Document Format) is incorrect: a PDF is a fixed-layout document format for presenting pages, not a text file of custom data tags.

  36. An analyst exports data to a text file where each field on a line is separated by a comma. What type of file format is this an example of?

    • A.A fixed-width file
    • B.A spreadsheet file
    • C.A delimited file
    • D.A columnar file
    Show answerHide answer

    Correct answer: A delimited file

    A delimited file is correct: a delimited file separates the fields on each line with a chosen character such as a comma, tab or pipe, and a comma-separated values (CSV) file is the most common example. A fixed-width file is incorrect: a fixed-width file places each field at set character positions rather than separating fields with a character. A spreadsheet file is incorrect: a workbook such as an .xlsx file is a binary or zipped package with sheets and formatting, not plain text lines split by commas, even though spreadsheet programs can open CSV files. A columnar file is incorrect: columnar formats such as Parquet store values column by column in a binary layout instead of as comma-separated lines of text.

  37. What is the key difference between a flat file and a relational database for storing data?

    • A.A flat file enforces integrity through foreign keys and constraints, while a relational database places data in one unlinked file
    • B.A flat file holds data in one standalone table, while a relational database links many tables through keys and enforces structure
    • C.A flat file scales to enormous workloads using built-in indexing, while a relational database fits data in small text spreadsheets
    • D.A flat file stores data in encrypted proprietary blocks, while a relational database writes plain text records separated by commas
    Show answerHide answer

    Correct answer: A flat file holds data in one standalone table, while a relational database links many tables through keys and enforces structure

    A flat file holds data in one standalone table, while a relational database links many tables through keys and enforces structure is correct: a flat file holds data in a single table or file with no enforced relationships, while a relational database spreads data across related tables joined by primary and foreign keys and enforces data types and integrity. A flat file enforces integrity through foreign keys and constraints, while a relational database places data in one unlinked file is incorrect: it reverses the two, because enforcing relationships between tables is what a relational database does and a flat file cannot. A flat file scales to enormous workloads using built-in indexing, while a relational database fits data in small text spreadsheets is incorrect: flat files have no built-in indexing or concurrency controls, and relational databases run large enterprise systems. A flat file stores data in encrypted proprietary blocks, while a relational database writes plain text records separated by commas is incorrect: flat files are usually plain text, such as CSV, while a relational database manages its own internal storage rather than comma-separated lines.

  38. A web analyst inspects a file built from tags such as <p>, <div>, and <a> that define how content is structured and displayed in a browser. Which format is this?

    • A.XML (Extensible Markup Language)
    • B.HTML (Hypertext Markup Language)
    • C.KML (Keyhole Markup Language)
    • D.GML (Geography Markup Language)
    Show answerHide answer

    Correct answer: HTML (Hypertext Markup Language)

    HTML (Hypertext Markup Language) is correct: HTML uses a predefined set of tags, such as <p>, <div> and <a>, to structure and display content in a web browser. XML (Extensible Markup Language) is incorrect: XML also uses tags, but its tags are defined by the author to describe data, and it has no built-in display tags such as <p> or <div>. KML (Keyhole Markup Language) is incorrect: KML is an XML-based format for map features such as placemarks and routes, not for web page content. GML (Geography Markup Language) is incorrect: GML is an XML-based format for exchanging geographic features and geometry, not for laying out a web page.

  39. In a database, what does a schema primarily define?

    • A.The stored contents and values of the database, including records, rows, null entries, and summary totals
    • B.The user accounts and permissions of the database, including logins, roles, password hashes, and privileges
    • C.The logical layout and structure of the database, including tables, fields, data types, and relationships
    • D.The backup schedule and history of the database, including jobs, copies, restore points, and failure alerts
    Show answerHide answer

    Correct answer: The logical layout and structure of the database, including tables, fields, data types, and relationships

    The logical layout and structure of the database, including tables, fields, data types, and relationships is correct: a schema is the blueprint of a database, defining its tables, the fields in each table, their data types and keys, and how the tables relate. The stored contents and values of the database, including records, rows, null entries, and summary totals is incorrect: the data actually held at a given moment is the database's contents, or instance, which fills the structure the schema defines. The user accounts and permissions of the database, including logins, roles, password hashes, and privileges is incorrect: accounts and privileges are security settings managed through access control, not the schema. The backup schedule and history of the database, including jobs, copies, restore points, and failure alerts is incorrect: backups are an operational recovery process and have nothing to do with how the data is structured.

  40. An analyst stores a library of customer-uploaded photos, voice recordings, and promotional videos for a campaign analysis. How are these media files best classified?

    • A.Semi-structured data (log, email, and feed formats)
    • B.Structured data (row, column, and field values)
    • C.Unstructured data (image, audio, and video formats)
    • D.Categorical data (label, group, and class values)
    Show answerHide answer

    Correct answer: Unstructured data (image, audio, and video formats)

    Unstructured data (image, audio, and video formats) is correct: photos, voice recordings and videos do not follow a predefined row-and-column model and cannot be queried directly without extra processing, so they are unstructured data. Semi-structured data (log, email, and feed formats) is incorrect: semi-structured data carries organizing markers such as tags, keys or header fields, which raw media files do not have. Structured data (row, column, and field values) is incorrect: structured data fits fixed rows and columns with defined data types, and media files do not. Categorical data (label, group, and class values) is incorrect: categorical data is a set of labels that place records into groups, not media content.

  41. A field stores values such as '2026-06-11' to represent calendar information used for filtering and time-based grouping. Which data type should the analyst assign?

    • A.Date
    • B.Text
    • C.Numeric
    • D.Integer
    Show answerHide answer

    Correct answer: Date

    Date is correct: a date data type stores calendar values such as '2026-06-11' so they sort in time order and support filtering, grouping by month or year, and date arithmetic. Text is incorrect: stored as text, the value is only a string of characters, so sorting and date calculations become unreliable. Numeric is incorrect: a general numeric type would treat the value as a quantity such as 20260611, which breaks calendar logic such as adding days or grouping by month. Integer is incorrect: an integer is a whole-number type for counts and quantities, and a date stored as an integer loses its calendar meaning in the same way.

  42. An organization wants a single, large, centralized repository that ingests raw structured, semi-structured, and unstructured data at scale so future analytical use cases can be defined later. Which environment best meets this need?

    • A.A data warehouse
    • B.A data lake
    • C.An OLTP database
    • D.An OLAP database
    Show answerHide answer

    Correct answer: A data lake

    A data lake fits this need: it is one large, centralized repository that stores raw structured, semi-structured, and unstructured data at scale and applies a schema only when the data is read, so use cases can be decided later. A data warehouse is also centralized, but it loads cleaned, structured data into a schema designed in advance for known reporting needs. An OLTP database runs day-to-day transactions in a fixed, normalized schema, and an OLAP database holds data already modeled into dimensions and aggregates for analysis that was planned in advance.

Data Acquisition and Preparation (45)

  1. Which concept in data management focuses on the use of data across different domains and formats for improved decision-making?

    • A.Data Warehousing
    • B.Data Aggregation
    • C.Data Stewardship
    • D.Data Integration
    Show answerHide answer

    Correct answer: Data Integration

    "Data Integration" is correct: data integration combines data from different sources, domains, and formats into a unified, consistent view so it can be used together for analysis and better decisions. "Data Warehousing" is incorrect: warehousing stores integrated historical data in a central repository for reporting; it depends on integration, but it is a storage and query architecture rather than the concept of unifying data across domains and formats. "Data Aggregation" is incorrect: aggregation summarizes detailed records into totals, averages, or counts; it condenses data instead of bringing different domains and formats together into one view. "Data Stewardship" is incorrect: stewardship makes people accountable for the quality, definitions, and proper use of data within a domain; it is a governance role, not the combining of data across domains.

  2. Which term describes the process of dividing a data set into smaller chunks to facilitate analysis?

    • A.Data Partitioning
    • B.Data Sampling
    • C.Data Normalization
    • D.Data Deduplication
    Show answerHide answer

    Correct answer: Data Partitioning

    "Data Partitioning" is correct: partitioning divides a large table or data set into smaller, more manageable segments (for example by date range, list, or hash) so each piece can be stored, processed, and analyzed more efficiently. "Data Sampling" is incorrect: sampling selects a representative subset of records to analyze in place of the whole; it does not divide the entire data set into chunks. "Data Normalization" is incorrect: normalization decomposes tables to remove redundancy and update anomalies; its aim is data integrity, not breaking a data set into chunks for analysis. "Data Deduplication" is incorrect: deduplication finds and removes duplicate records or blocks; it shrinks the data rather than dividing it into pieces.

  3. What is the main objective of data normalization in a database?

    • A.To duplicate columns and accelerate data access
    • B.To maintain indexes and expedite data retrieval
    • C.To reduce redundancy and improve data integrity
    • D.To partition tables and distribute data storage
    Show answerHide answer

    Correct answer: To reduce redundancy and improve data integrity

    "To reduce redundancy and improve data integrity" is correct: normalization organizes tables through the normal forms so each fact is stored once, which removes duplicated data and prevents insert, update, and delete anomalies. "To duplicate columns and accelerate data access" is incorrect: deliberately duplicating columns to speed up reads is denormalization, the reverse of what normalization sets out to do. "To maintain indexes and expedite data retrieval" is incorrect: indexing speeds up lookups; normalization often adds joins and is not designed to make retrieval faster. "To partition tables and distribute data storage" is incorrect: partitioning splits a large table into segments to spread storage and improve manageability; normalization splits tables to remove redundancy, not to distribute storage.

  4. In data environments, what is the primary purpose of ETL (Extract, Transform, Load.) processes?

    • A.Integrating records from various sources into a data warehouse
    • B.Converting aggregates from data marts into a dynamic dashboard
    • C.Mirroring transaction data from servers into a standby replica
    • D.Masking sensitive data from production into a sandbox database
    Show answerHide answer

    Correct answer: Integrating records from various sources into a data warehouse

    "Integrating records from various sources into a data warehouse" is correct: ETL extracts data from source systems, transforms it into a consistent format, and loads it into a target, most often a data warehouse, so data from many sources can be analyzed together. "Converting aggregates from data marts into a dynamic dashboard" is incorrect: turning summarized data into dashboards is the work of business intelligence and visualization tools, which use the data that ETL has already loaded into the warehouse and its marts. "Mirroring transaction data from servers into a standby replica" is incorrect: mirroring, or replication, keeps a synchronized copy of a database for failover and high availability; it copies data as it is instead of transforming and integrating it. "Masking sensitive data from production into a sandbox database" is incorrect: data masking replaces sensitive values with realistic substitutes so a copy of production data can be used safely in a test or sandbox database; it protects privacy rather than integrating sources for analysis.

  5. Which concept is essential for understanding the relationships between different data elements in a database?

    • A.Data Mining
    • B.Data Merging
    • C.Data Masking
    • D.Data Mapping
    Show answerHide answer

    Correct answer: Data Mapping

    "Data Mapping" is correct: data mapping defines how individual data elements in one data set or model correspond to elements in another, such as which source field feeds which target column, which documents the relationships between data elements and underpins integration and migration. "Data Mining" is incorrect: data mining discovers patterns, correlations, and trends in large data sets; it analyzes the values in the data rather than defining how data elements relate to one another. "Data Merging" is incorrect: merging physically combines two data sets into one, for example by appending or joining records; it relies on a mapping of the fields but is the operation, not the concept that defines the relationships. "Data Masking" is incorrect: masking hides sensitive values by replacing them with realistic substitutes; it protects data rather than describing how data elements relate.

  6. In a data environment, what is "data deduplication" primarily used for?

    • A.Increasing the retrieval speed of data
    • B.Reducing the storage footprint of data
    • C.Redacting the sensitive fields of data
    • D.Maintaining the standby copies of data
    Show answerHide answer

    Correct answer: Reducing the storage footprint of data

    "Reducing the storage footprint of data" is correct: deduplication finds repeated copies of data, keeps one unique instance, and replaces the other copies with references to it, so much less storage space is needed. "Increasing the retrieval speed of data" is incorrect: faster retrieval comes from indexing, caching, or query tuning; deduplication targets storage savings, and rebuilding deduplicated data can even add read overhead. "Redacting the sensitive fields of data" is incorrect: redaction and masking hide confidential values to protect privacy; they do not remove duplicate copies. "Maintaining the standby copies of data" is incorrect: keeping standby copies is replication for availability and recovery, which deliberately adds redundant copies, the opposite of deduplication.

  7. In the context of data environments, what does "Data Hygiene" refer to?

    • A.The practice of keeping data organized, accurate, and current
    • B.The practice of keeping data encrypted, masked, and protected
    • C.The practice of keeping data centralized, unified, and shared
    • D.The practice of keeping data indexed, partitioned, and cached
    Show answerHide answer

    Correct answer: The practice of keeping data organized, accurate, and current

    "The practice of keeping data organized, accurate, and current" is correct: data hygiene is the ongoing work of keeping data clean, for example by validating, correcting, deduplicating, standardizing, and updating records so they stay accurate and up to date. "The practice of keeping data encrypted, masked, and protected" is incorrect: encryption and masking are data security and privacy controls; they protect data from exposure but do not make it accurate or current. "The practice of keeping data centralized, unified, and shared" is incorrect: centralizing and unifying data is integration or consolidation; data can be fully integrated and still be inaccurate or out of date. "The practice of keeping data indexed, partitioned, and cached" is incorrect: indexing, partitioning, and caching are performance techniques that speed access; they do not clean the data.

  8. In data mining, which method is used to handle missing data values?

    • A.Data smoothing
    • B.Data imputation
    • C.Data deduplication
    • D.Data normalization
    Show answerHide answer

    Correct answer: Data imputation

    Data imputation is the method for handling missing values: each gap is filled with a substitute such as the mean, median or mode of the column, or a value predicted by a model, so the record can still be used in the analysis. Data smoothing reduces random noise in values that are present, for example with binning or moving averages, and does not fill gaps. Data deduplication finds and removes repeated records. Data normalization rescales values to a common range, such as 0 to 1, and leaves missing entries missing.

  9. A data analyst needs to pull live currency exchange rates from a financial services provider directly into a reporting pipeline without manually downloading files. The provider exposes a documented set of endpoints that return structured data on request. Which data acquisition method is the analyst using?

    • A.Electronic data interchange (EDI) transactions
    • B.Secure file transfer protocol (SFTP) downloads
    • C.Application programming interface (API) access
    • D.Optical character recognition (OCR) processing
    Show answerHide answer

    Correct answer: Application programming interface (API) access

    Application programming interface (API) access is the method: the provider publishes documented endpoints, and the pipeline sends requests to them and receives structured data in return, with no files to download by hand. Secure file transfer protocol (SFTP) downloads move whole files from a server, which is the file handling the analyst wants to avoid. Electronic data interchange (EDI) transactions exchange standardized business documents, such as orders and invoices, between trading partners, and optical character recognition (OCR) processing converts images of printed or scanned documents into text; neither one requests live values from a provider's endpoints.

  10. A retailer wants to collect product prices from a competitor's public website that offers no data feed or programmatic access. An analyst writes a routine that downloads the pages and extracts price fields from the page markup. What is this acquisition technique called?

    • A.Data mining
    • B.API integration
    • C.Public database
    • D.Web scraping
    Show answerHide answer

    Correct answer: Web scraping

    Web scraping is the technique: a routine downloads the competitor's pages and extracts the price fields from the HTML markup, which is how data is collected from a site that offers no feed or programmatic access. API integration would need the site to expose endpoints, which this one does not. A public database is a published dataset, such as a government open-data portal, that is queried or downloaded as data rather than pulled out of page markup, and data mining searches data that has already been collected for patterns rather than gathering it.

  11. In the context of the CompTIA Data+ objectives, which statement best describes data acquisition?

    • A.Copying and storing data from production databases and servers into an offsite vault for recovery
    • B.Moving and compressing data from active systems and files into an archive for long-term retention
    • C.Summarizing and charting data from tables and queries into an interactive dashboard for managers
    • D.Collecting and importing data from internal and external sources into an environment for analysis
    Show answerHide answer

    Correct answer: Collecting and importing data from internal and external sources into an environment for analysis

    Collecting and importing data from internal and external sources into an environment for analysis is data acquisition, the front-end step of gathering raw data through methods such as APIs, web scraping, surveys, and public databases. Copying and storing data from production databases and servers into an offsite vault for recovery is a backup, which protects data rather than gathering it for analysis. Moving and compressing data from active systems and files into an archive for long-term retention is archiving, which takes data out of active use. Summarizing and charting data from tables and queries into an interactive dashboard for managers is visualization and reporting, which comes after the data has been acquired and prepared.

  12. A nightly process loads only the records that were inserted or changed in the source system since the previous run, rather than reloading the entire table. Which integration method does this describe?

    • A.Full load
    • B.Initial load
    • C.Delta load
    • D.Trickle load
    Show answerHide answer

    Correct answer: Delta load

    A delta load is the method: each nightly run picks up only the rows that were inserted or changed since the previous run instead of reloading the whole table. A full load copies the entire table on every run, and an initial load is the first complete population of a new target. A trickle load also moves changed records, but it feeds them in continuously throughout the day rather than in one scheduled nightly batch.

  13. A team extracts raw data from several sources, loads it unchanged into a cloud data warehouse, and then runs transformations using the warehouse's compute engine. Which integration approach is being used?

    • A.ETL
    • B.CDC
    • C.ESB
    • D.ELT
    Show answerHide answer

    Correct answer: ELT

    ELT (extract, load, transform) is the approach: the raw data is loaded into the cloud warehouse unchanged and transformed afterward with the warehouse's own compute engine. ETL (extract, transform, load) would transform the data in a staging layer before loading it, which is the opposite order. CDC (change data capture) identifies the rows that were inserted, updated, or deleted in a source so that only the changes are copied, and an ESB (enterprise service bus) is middleware that routes messages between applications; neither describes loading raw data first and transforming it inside the target.

  14. What is the primary difference between ETL and ELT?

    • A.ETL batches data and then loads it, while ELT streams raw data and then loads it in milliseconds
    • B.ETL transforms data and then loads it, while ELT loads raw data and then transforms it in place
    • C.ETL copies data and then loads it, while ELT maps raw data and then queries it in source tables
    • D.ETL detects changed data and then loads it, while ELT pulls raw data and then reloads it in full
    Show answerHide answer

    Correct answer: ETL transforms data and then loads it, while ELT loads raw data and then transforms it in place

    The defining difference is the order of the steps: ETL transforms data and then loads it, while ELT loads raw data and then transforms it in place, using the target system's own processing power. The statement that ETL batches data and then loads it, while ELT streams raw data and then loads it in milliseconds confuses the two with batch and streaming delivery, which either pattern can use. The statement that ETL copies data and then loads it, while ELT maps raw data and then queries it in source tables describes data virtualization, not ELT, which does load the data into the target. The statement that ETL detects changed data and then loads it, while ELT pulls raw data and then reloads it in full confuses them with delta and full loads, which are scheduling choices available to both.

  15. An analyst is documenting how raw data should be staged before it reaches an on-premises data warehouse that has limited storage and strict schema requirements. Transformations must be completed before any data lands in the warehouse. Which process fits this requirement?

    • A.ELT
    • B.CDC
    • C.EAI
    • D.ETL
    Show answerHide answer

    Correct answer: ETL

    ETL (extract, transform, load) fits: the data is transformed and conformed in a staging layer, and only the finished result is loaded into the warehouse, which suits a target with limited storage and a strict schema. ELT (extract, load, transform) would land raw data in the warehouse first and transform it there, which the storage limit and the schema rules do not allow. CDC (change data capture) identifies which source rows have changed, and EAI (enterprise application integration) connects business applications so they can share data and processes; neither one completes the transformations before data reaches the warehouse.

  16. A market research group gathers opinions by distributing a structured questionnaire to 500 customers and recording their responses. Which data collection method is this?

    • A.Focus groups
    • B.Surveys
    • C.Web scraping
    • D.Observation
    Show answerHide answer

    Correct answer: Surveys

    Surveys are the method: a structured questionnaire is sent to a set of respondents and their answers are recorded, which is survey-based data collection. Focus groups gather opinions too, but through a moderated discussion with a small group rather than a fixed questionnaire sent to 500 people. Observation records behavior as it happens without asking anyone questions, and web scraping extracts existing data from web pages rather than collecting new responses.

  17. A quality analyst stands on a factory floor and records how long each assembly step takes without interacting with the workers or systems. Which data collection method best describes this?

    • A.Surveys
    • B.Observation
    • C.Experiments
    • D.Simulations
    Show answerHide answer

    Correct answer: Observation

    Observation is the method: the analyst gathers data by watching and recording each assembly step as it happens, without questioning the workers or pulling records from a system. Surveys depend on asking people questions, which the analyst deliberately avoids. Experiments change a condition on purpose to measure its effect, and simulations generate data from a model of the process, while here the real line is simply timed as it runs.

  18. A population dataset contains 2,000,000 records, which is too large to analyze in full within the project deadline. The analyst selects a representative subset of 5,000 records using a random selection method to draw conclusions about the whole. Which acquisition concept is being applied?

    • A.Filtering
    • B.Profiling
    • C.Sampling
    • D.Binning
    Show answerHide answer

    Correct answer: Sampling

    Sampling is the concept: a representative subset of 5,000 records is drawn at random so that conclusions can be generalized to all 2,000,000 records without processing every one. Filtering also keeps a subset, but it selects the rows that meet a stated condition rather than drawing them at random to stand for the whole population. Profiling examines a dataset's structure and quality, and binning groups continuous values into ranges; neither one selects a subset of records for drawing conclusions.

  19. During data profiling, an analyst discovers that the same customer appears in the table three times with identical values across all fields. Which data quality issue does this represent?

    • A.Redundant data
    • B.Corrupted data
    • C.Duplicate data
    • D.Invalid data
    Show answerHide answer

    Correct answer: Duplicate data

    Duplicate data is the issue: the same customer record appears three times with identical values in every field, which inflates counts and totals until the extra copies are removed. Redundant data is the same piece of information stored in more than one place, such as a customer's address kept in two different tables or columns, rather than a whole record repeated as extra rows. Corrupted data has been damaged so that values are garbled or unreadable, and invalid data holds values that break a rule, such as an impossible date; the repeated rows here are intact and valid.

  20. An analyst is profiling a survey export and finds that a column expected to contain only the values 'Yes' or 'No' also contains entries such as 'Maybe', 'Y', and 'unknown'. Which data quality problem is this?

    • A.Specification mismatch
    • B.Duplicate observations
    • C.Statistical outliers
    • D.Missing value
    Show answerHide answer

    Correct answer: Specification mismatch

    Specification mismatch is the problem: the column's specification allows only 'Yes' or 'No', and entries such as 'Maybe', 'Y', and 'unknown' do not conform to that defined set of values. Duplicate observations are the same record repeated, which is duplicate data. Statistical outliers are values far outside the typical range of a numeric field, and missing values are empty entries; neither describes values that are present but fall outside a column's allowed list.

  21. While profiling a dataset, an analyst notices a column meant to store dates contains several entries like 'N/A' and free-text notes, which breaks date calculations. Identifying this problem is an example of which profiling check?

    • A.Data type validation
    • B.Uniqueness validation
    • C.Cross-field validation
    • D.Value range validation
    Show answerHide answer

    Correct answer: Data type validation

    Data type validation is the check: it confirms that every value in a field matches the field's expected data type, so text such as 'N/A' or a free-form note in a date column is caught before it breaks date calculations. Uniqueness validation looks for repeated values in a field that should hold each value once, such as an ID. Value range validation tests whether values of the right type fall between allowed limits, and cross-field validation tests whether related fields agree, such as an end date that falls after its start date; neither one targets values of the wrong type.

  22. What is data profiling in the data preparation process?

    • A.Inspecting a dataset to learn its structure, content, quality, and statistics before it is used
    • B.Scrubbing a dataset to resolve its errors, blanks, duplicates, and formatting before it is used
    • C.Restructuring a dataset to convert its layout, data types, units, and formats before it is used
    • D.Masking a dataset to obscure its names, emails, addresses, and phone numbers before it is used
    Show answerHide answer

    Correct answer: Inspecting a dataset to learn its structure, content, quality, and statistics before it is used

    Inspecting a dataset to learn its structure, content, quality, and statistics before it is used is data profiling: it reports value distributions, completeness, formats, and anomalies so that the cleansing work can be planned. Scrubbing a dataset to resolve its errors, blanks, duplicates, and formatting before it is used is data cleansing, the step that profiling informs. Restructuring a dataset to convert its layout, data types, units, and formats before it is used is data transformation. Masking a dataset to obscure its names, emails, addresses, and phone numbers before it is used is a privacy control rather than an assessment of the data.

  23. What is data cleansing?

    • A.Reviewing and summarizing or charting data types, numeric ranges, null counts, or distributions in a dataset
    • B.Detecting and fixing or removing inaccurate, incomplete, duplicated, or badly formatted records in a dataset
    • C.Converting and reshaping or recoding text, timestamps, currencies, or category code fields in a dataset
    • D.Scrambling and substituting or encrypting surnames, emails, birthdates, or bank account numbers in a dataset
    Show answerHide answer

    Correct answer: Detecting and fixing or removing inaccurate, incomplete, duplicated, or badly formatted records in a dataset

    Detecting and fixing or removing inaccurate, incomplete, duplicated, or badly formatted records in a dataset is data cleansing, which improves quality by repairing or eliminating the problems that profiling uncovers. Reviewing and summarizing or charting data types, numeric ranges, null counts, or distributions in a dataset is data profiling, which describes the data without changing it. Converting and reshaping or recoding text, timestamps, currencies, or category code fields in a dataset is data transformation, which changes format and structure rather than correcting errors. Scrambling and substituting or encrypting surnames, emails, birthdates, or bank account numbers in a dataset is data masking, a privacy control.

  24. A new analyst asks how data wrangling differs from data cleaning. Which explanation is most accurate?

    • A.Data cleaning is the overall process of validating and turning raw data into a trustworthy dataset, and data wrangling is one part of it that reshapes tables and columns
    • B.Data wrangling is the overall process of collecting and importing raw data into a central store, and data cleaning is one part of it that loads and indexes table records
    • C.Data wrangling is the overall process of transforming and mapping raw data into a usable form, and data cleaning is one part of it that fixes errors and inconsistencies
    • D.Data cleaning is the overall process of mining and summarizing raw data into a written report, and data wrangling is one part of it that charts and presents key results
    Show answerHide answer

    Correct answer: Data wrangling is the overall process of transforming and mapping raw data into a usable form, and data cleaning is one part of it that fixes errors and inconsistencies

    Data wrangling is the overall process of transforming and mapping raw data into a usable form, and data cleaning is one part of it that fixes errors and inconsistencies; cleaning is a stage inside wrangling, not the other way around. The statement that data cleaning is the overall process of validating and turning raw data into a trustworthy dataset, and data wrangling is one part of it that reshapes tables and columns reverses that hierarchy. The statement that data wrangling is the overall process of collecting and importing raw data into a central store, and data cleaning is one part of it that loads and indexes table records describes acquisition and loading instead. The statement that data cleaning is the overall process of mining and summarizing raw data into a written report, and data wrangling is one part of it that charts and presents key results both reverses the hierarchy and describes analysis and visualization instead.

  25. What is data wrangling?

    • A.The overall process of transforming, restructuring, and enriching raw data into a clean format for analysis
    • B.The scheduled process of extracting, converting, and loading source data into a data warehouse for reporting
    • C.The exploratory work of searching, clustering, and modeling a large data set with algorithms for prediction
    • D.The visual work of charting, summarizing, and presenting finished data with a live dashboard for executives
    Show answerHide answer

    Correct answer: The overall process of transforming, restructuring, and enriching raw data into a clean format for analysis

    The overall process of transforming, restructuring, and enriching raw data into a clean format for analysis is data wrangling, also called data munging, which spans cleansing, structuring, and enriching data for a particular analysis. The scheduled process of extracting, converting, and loading source data into a data warehouse for reporting describes an ETL pipeline, an engineered process that runs the same steps on a schedule. The exploratory work of searching, clustering, and modeling a large data set with algorithms for prediction is data mining, which works on data that has already been prepared. The visual work of charting, summarizing, and presenting finished data with a live dashboard for executives is visualization and reporting.

  26. How does data wrangling differ from ETL?

    • A.Data wrangling is typically automated, engineer-run maintenance of scheduled batch jobs, while ETL is a manual, improvised cleanup that copies and fixes spreadsheets between analysts
    • B.Data wrangling is typically reserved for unstructured text, images, and audio files, while ETL is reserved for tabular datasets that move between relational databases
    • C.Data wrangling is typically exploratory, analyst-led preparation of a specific dataset, while ETL is a structured, repeatable pipeline that moves and transforms data between systems
    • D.Data wrangling is typically the modeling stage for training, testing, and tuning predictive models, while ETL is the reporting stage that shares charts between business teams
    Show answerHide answer

    Correct answer: Data wrangling is typically exploratory, analyst-led preparation of a specific dataset, while ETL is a structured, repeatable pipeline that moves and transforms data between systems

    Data wrangling is typically exploratory, analyst-led preparation of a specific dataset, while ETL is a structured, repeatable pipeline that moves and transforms data between systems: wrangling is iterative and tied to a particular analysis, whereas ETL is engineered and scheduled to feed warehouses. The claim that data wrangling is typically automated, engineer-run maintenance of scheduled batch jobs, while ETL is a manual, improvised cleanup that copies and fixes spreadsheets between analysts swaps the two roles. The claim that data wrangling is typically reserved for unstructured text, images, and audio files, while ETL is reserved for tabular datasets that move between relational databases invents a split by data type; both can work with structured and unstructured sources. The claim that data wrangling is typically the modeling stage for training, testing, and tuning predictive models, while ETL is the reporting stage that shares charts between business teams confuses preparation with modeling and reporting.

  27. A dataset has 1,000 customer rows, but 80 rows have a blank value in the 'annual_income' field. The analyst must decide how to handle these gaps before modeling. Which preparation issue is being addressed?

    • A.Missing values
    • B.Outlying values
    • C.Invalid entries
    • D.Stale entries
    Show answerHide answer

    Correct answer: Missing values

    The issue is missing values: 80 of the 1,000 rows have no entry in the 'annual_income' field, and those gaps must be filled or the rows handled some other way before modeling. Outlying values are present but extreme, such as an income far above every other customer's. Invalid entries are present but break a rule, such as a negative income, and stale entries are values that were once correct but are now out of date; a blank field is none of these, because nothing was recorded at all.

  28. An analyst chooses to remove every row that has any missing value before running a statistical test. Which missing-data handling method is this, and what is its main risk?

    • A.Median imputation, which can understate variance and distort the correlations
    • B.Pairwise deletion, which can produce uneven sample sizes and blur the results
    • C.Listwise deletion, which can bias the results and discard many usable records
    • D.Regression imputation, which can overstate the fit and understate uncertainty
    Show answerHide answer

    Correct answer: Listwise deletion, which can bias the results and discard many usable records

    Listwise deletion, which can bias the results and discard many usable records, is what the analyst is doing: every row with any gap is dropped, which can shrink the sample sharply and skew it when the values are not missing at random. Pairwise deletion, which can produce uneven sample sizes and blur the results, drops a case only from the calculations that need the missing variable, so it does not remove whole rows up front. Median imputation, which can understate variance and distort the correlations, and regression imputation, which can overstate the fit and understate uncertainty, both fill the gaps instead of deleting anything.

  29. What is imputation in data preparation?

    • A.Summarizing numeric values with functions such as the sum, count, average, or a running total
    • B.Rescaling numeric values with formulas such as the z-score, min-max, decimal scaling, or a log transform
    • C.Replacing sensitive values with substitutes such as the hash, a token, or a dummy string
    • D.Replacing missing values with estimates such as the mean, median, mode, or a predicted value
    Show answerHide answer

    Correct answer: Replacing missing values with estimates such as the mean, median, mode, or a predicted value

    Replacing missing values with estimates such as the mean, median, mode, or a predicted value is imputation, which fills gaps so that the records can stay in the analysis. Summarizing numeric values with functions such as the sum, count, average, or a running total is aggregation. Rescaling numeric values with formulas such as the z-score, min-max, decimal scaling, or a log transform is normalization, which changes the range of values that already exist. Replacing sensitive values with substitutes such as the hash, a token, or a dummy string is data masking, which protects real values rather than filling in absent ones.

  30. An analyst compares listwise deletion with imputation for a dataset where about 30 percent of rows have a missing value in one column. Which consideration favors imputation in this scenario?

    • A.Imputation recovers the genuine values missing from 30 percent of rows, while the analysis would stay unbiased
    • B.Imputation adds rows for the missing 30 percent, so the statistical power of the analysis would rise sharply
    • C.Imputation is required, as the intended statistical test would be blocked by empty cells in 30 percent of rows
    • D.Imputation keeps the sample size by filling gaps, while deleting 30 percent of rows would sharply cut the data
    Show answerHide answer

    Correct answer: Imputation keeps the sample size by filling gaps, while deleting 30 percent of rows would sharply cut the data

    Imputation keeps the sample size by filling gaps, while deleting 30 percent of rows would sharply cut the data; avoiding that loss of data is the consideration that favors imputation here. The claim that imputation recovers the genuine values missing from 30 percent of rows, while the analysis would stay unbiased, is false, because imputed values are only estimates and can introduce bias of their own. The claim that imputation adds rows for the missing 30 percent, so the statistical power of the analysis would rise sharply, is false, because no rows are missing: imputation fills empty cells in rows that already exist and keeps the sample at its original size rather than enlarging it. The claim that imputation is required, as the intended statistical test would be blocked by empty cells in 30 percent of rows, is false, because most tools simply drop incomplete rows and run anyway, which is exactly the loss of data that imputation avoids.

  31. In a sales dataset where most transactions fall between 50 and 500 dollars, one record shows 4,200,000 dollars and appears to be a data-entry error. What is this extreme value called?

    • A.Missing value
    • B.Outlier
    • C.Duplicate
    • D.Redundant data
    Show answerHide answer

    Correct answer: Outlier

    An outlier is a value that lies far outside the typical range of the other observations, here 4,200,000 dollars among sales of 50 to 500 dollars, whether it turns out to be an error or a genuine extreme. A missing value is an empty entry, a duplicate repeats a record that already exists, and redundant data is the same information stored in more than one place; this value is present, unique, and a properly formatted number, but extreme.

  32. An analyst finds several extreme values in a dataset and is unsure whether they are errors or genuine. Which approach reflects sound handling of outliers?

    • A.Delete each extreme value, then note it in the audit log
    • B.Replace each extreme value with the mean, median, or mode, then treat it as genuine data
    • C.Investigate each extreme value, then fix, keep, or exclude it as the cause dictates
    • D.Retain each extreme value, then add a note about it in the report
    Show answerHide answer

    Correct answer: Investigate each extreme value, then fix, keep, or exclude it as the cause dictates

    Investigate each extreme value, then fix, keep, or exclude it as the cause dictates: an outlier can be a data-entry mistake or a genuine extreme, so whether it is corrected, kept, or excluded has to follow what caused it. To delete each extreme value, then note it in the audit log, documents the change but still throws away genuine extremes along with the mistakes. To replace each extreme value with the mean, median, or mode, then treat it as genuine data, invents values and hides real variation. To retain each extreme value, then add a note about it in the report, leaves any mistakes in the analysis without ever checking them.

  33. What is data manipulation in the preparation stage?

    • A.Changing the structure, content, or layout of data through operations such as filtering, sorting, aggregating, and reshaping
    • B.Assigning the owners, stewards, or custodians of data through policies such as access, retention, quality, and sharing rules
    • C.Protecting the privacy, integrity, or availability of data through controls such as encryption, masking, backups, and audits
    • D.Communicating the trends, patterns, or findings of data through visuals such as charts, graphs, heat maps, and dashboards
    Show answerHide answer

    Correct answer: Changing the structure, content, or layout of data through operations such as filtering, sorting, aggregating, and reshaping

    Changing the structure, content, or layout of data through operations such as filtering, sorting, aggregating, and reshaping is data manipulation, which reorganizes data so that it is usable for analysis. Assigning the owners, stewards, or custodians of data through policies such as access, retention, quality, and sharing rules is data governance. Protecting the privacy, integrity, or availability of data through controls such as encryption, masking, backups, and audits is data security. Communicating the trends, patterns, or findings of data through visuals such as charts, graphs, heat maps, and dashboards is visualization and reporting.

  34. An analyst keeps only the rows where 'region' equals 'West' so the report focuses on western sales, removing all other rows from view. Which manipulation technique is this?

    • A.Sampling
    • B.Filtering
    • C.Profiling
    • D.Appending
    Show answerHide answer

    Correct answer: Filtering

    Filtering is the technique: it keeps only the rows that meet a stated condition, here 'region' equals 'West', and hides the rest. Sampling also produces a subset, but it draws rows at random to represent the whole rather than selecting them by a condition. Profiling examines a dataset's structure and quality without removing anything, and appending adds rows from another dataset to the bottom of the table.

  35. What is data aggregation?

    • A.Combining many records into summary values such as sums, counts, or averages for each category
    • B.Breaking single text strings into components such as street, city, or postal code for each address
    • C.Inserting statistical estimates such as means, medians, or modes into blank fields for each column
    • D.Splitting full datasets into subsets such as training, validation, or test sets for each model
    Show answerHide answer

    Correct answer: Combining many records into summary values such as sums, counts, or averages for each category

    Combining many records into summary values such as sums, counts, or averages for each category is data aggregation, which rolls detailed rows up into grouped totals, such as total sales by region. Breaking single text strings into components such as street, city, or postal code for each address is parsing. Inserting statistical estimates such as means, medians, or modes into blank fields for each column is imputation. Splitting full datasets into subsets such as training, validation, or test sets for each model is data partitioning, which divides records rather than summarizing them.

  36. A column contains full names stored as 'Last, First' in one string. The analyst splits each value into separate 'last_name' and 'first_name' fields by interpreting the comma delimiter. Which manipulation technique is this?

    • A.Blending
    • B.Recoding
    • C.Pivoting
    • D.Parsing
    Show answerHide answer

    Correct answer: Parsing

    Parsing is the technique: it breaks one value into meaningful parts by reading a delimiter or pattern, here the comma that separates the last name from the first name. Blending combines data from different sources on a shared field, usually inside a visualization tool, so it joins data together rather than splitting it. Recoding changes the values of a variable into new codes or labels, and pivoting rotates rows into columns to summarize them; neither one separates a single string into parts.

  37. What does transposing data accomplish?

    • A.Sorting a dataset so that its rows run from highest to lowest
    • B.Rotating a dataset so that its rows and columns exchange places
    • C.Filtering a dataset so that its rows from other groups drop out
    • D.Fixing a dataset so that its duplicated rows and typos are gone
    Show answerHide answer

    Correct answer: Rotating a dataset so that its rows and columns exchange places

    Rotating a dataset so that its rows and columns exchange places is what transposing does: each row becomes a column and each column becomes a row, which reshapes the layout for a different view of the same values. Sorting a dataset so that its rows run from highest to lowest changes only the order of the rows. Filtering a dataset so that its rows from other groups drop out removes rows that fail a condition, and fixing a dataset so that its duplicated rows and typos are gone is data cleansing; neither one changes the orientation of the table.

  38. A survey stored gender as 1, 2, and 3, but the analyst converts these codes into 'Female', 'Male', and 'Nonbinary' to make a report readable. Which manipulation technique is this?

    • A.Parsing
    • B.Recoding
    • C.Sampling
    • D.Pivoting
    Show answerHide answer

    Correct answer: Recoding

    Recoding is the technique: it maps the values of a variable to different categories or labels, here turning the codes 1, 2, and 3 into readable text. Parsing splits one value into several parts, sampling selects a subset of records, and pivoting rotates rows into columns to summarize them; none of them relabels the existing values of a variable.

  39. An analyst stacks December sales records onto the bottom of a table that already holds January through November sales, so all rows share the same columns. Which manipulation technique is this?

    • A.Appending
    • B.Merging
    • C.Splitting
    • D.Filtering
    Show answerHide answer

    Correct answer: Appending

    Appending is the technique: it adds the rows of one dataset to the bottom of another that has the same columns, so December's records are stacked under January through November. Merging combines datasets side by side by matching key fields, which adds columns rather than stacking rows. Splitting divides one table or field into several, and filtering keeps only the rows that meet a condition; neither one adds records to a table.

  40. What is a dataset merge?

    • A.Stacking the rows of two datasets vertically, usually when both share the same columns
    • B.Collapsing the rows of a dataset into summary totals, usually split by a category field
    • C.Removing the rows of a dataset that have gaps, usually when a test needs complete cases
    • D.Combining the rows of related datasets into one table, usually by matching common keys
    Show answerHide answer

    Correct answer: Combining the rows of related datasets into one table, usually by matching common keys

    Combining the rows of related datasets into one table, usually by matching common keys, is a dataset merge: records from each source that share a key value in one or more fields are joined into a single, wider dataset. Stacking the rows of two datasets vertically, usually when both share the same columns, is appending, which adds records below one another without matching keys. Collapsing the rows of a dataset into summary totals, usually split by a category field, is aggregation. Removing the rows of a dataset that have gaps, usually when a test needs complete cases, is listwise deletion.

  41. Two tables are merged so that the result keeps only the rows whose key exists in BOTH tables. Which join type is this?

    • A.Inner join
    • B.Full outer join
    • C.Left join
    • D.Right join
    Show answerHide answer

    Correct answer: Inner join

    An inner join is the type: it keeps only the rows whose key appears in both tables and drops unmatched rows from either side. A full outer join keeps every row from both tables and fills the gaps with nulls. A left join keeps every row from the left table plus the matches from the right, and a right join does the reverse, so neither one limits the result to keys found in both tables.

  42. An analyst wants a result that keeps every customer from the Customers table and attaches order details where they exist, leaving order fields blank for customers with no orders. Which join produces this?

    • A.Left outer join with Customers as the first table
    • B.Right outer join with Orders as the second table
    • C.Right anti join with Customers as the first table
    • D.Left anti join with Orders as the second table
    Show answerHide answer

    Correct answer: Left outer join with Customers as the first table

    A left outer join with Customers as the first table (FROM Customers LEFT OUTER JOIN Orders) keeps every customer row and fills the order columns with nulls where no order matches, which is exactly the result described. A right outer join with Orders as the second table keeps every order instead, so customers who have never ordered drop out. A left anti join with Orders as the second table returns only the customers who have no orders, with no order details attached, and a right anti join with Customers as the first table returns only orders that have no matching customer.

  43. What does a full outer join return when combining two tables?

    • A.Only rows from both tables where join keys match, with the remaining data filtered out
    • B.All rows from both tables, matched where keys align, with unmatched fields left null
    • C.All possible pairings of rows from both tables, with no join condition applied to them
    • D.Rows from both tables, stacked one upon another, with no duplicates kept in the result
    Show answerHide answer

    Correct answer: All rows from both tables, matched where keys align, with unmatched fields left null

    A full outer join returns all rows from both tables, matched where keys align, with unmatched fields left null, so no record from either side is lost. Only rows from both tables where join keys match, with the remaining data filtered out, describes an inner join; all possible pairings of rows from both tables, with no join condition applied to them, describes a cross join; and rows from both tables, stacked one upon another, with no duplicates kept in the result, describes a UNION, which appends rows instead of joining them on keys.

  44. What is database normalization?

    • A.Converting raw dates into one consistent display layout, typically by applying one parsing rule to each source column
    • B.Organizing tables and columns to cut redundancy and protect integrity, typically by splitting data into related tables
    • C.Rescaling numeric input fields to a common range for fair comparison, typically by mapping values between zero and one
    • D.Deleting duplicate records to keep a single copy per customer, typically by matching rows on shared identifying fields
    Show answerHide answer

    Correct answer: Organizing tables and columns to cut redundancy and protect integrity, typically by splitting data into related tables

    Database normalization means organizing tables and columns to cut redundancy and protect integrity, typically by splitting data into related tables so that each fact is stored once. Converting raw dates into one consistent display layout, typically by applying one parsing rule to each source column, is data standardization, a cleansing step that is sometimes loosely called normalizing formats. Rescaling numeric input fields to a common range for fair comparison, typically by mapping values between zero and one, is min-max normalization of features for analysis, and deleting duplicate records to keep a single copy per customer, typically by matching rows on shared identifying fields, is deduplication; neither changes the schema.

  45. A reporting database is intentionally restructured to combine related tables into fewer, wider tables so that analysts can read it with fewer joins, accepting some redundancy for faster queries. Which design choice describes the difference being applied?

    • A.Normalization, which splits wide tables into related ones, unlike denormalization, which combines them back together
    • B.Indexing, which adds lookup structures for faster searches, unlike a full scan, which reads through every stored row
    • C.Partitioning, which splits one large table into date ranges, unlike sharding, which places rows on several servers
    • D.Denormalization, which trades duplicated data for quicker lookups, unlike normalization, which minimizes duplication
    Show answerHide answer

    Correct answer: Denormalization, which trades duplicated data for quicker lookups, unlike normalization, which minimizes duplication

    The team applied denormalization, which trades duplicated data for quicker lookups, unlike normalization, which minimizes duplication by storing each fact once. Normalization, which splits wide tables into related ones, unlike denormalization, which combines them back together, names the opposite of what was done. Indexing, which adds lookup structures for faster searches, unlike a full scan, which reads through every stored row, speeds queries without merging any tables, and partitioning, which splits one large table into date ranges, unlike sharding, which places rows on several servers, divides data rather than combining it.

Data Analysis (46)

  1. In data mining, what is the primary purpose of the Apriori algorithm?

    • A.To build decision trees that classify new records
    • B.To find item sets that frequently appear together
    • C.To create clusters that group similar data points
    • D.To flag rare outliers that signal unusual charges
    Show answerHide answer

    Correct answer: To find item sets that frequently appear together

    "To find item sets that frequently appear together" is correct: Apriori scans transaction data for frequent itemsets, pruning candidates on the principle that every subset of a frequent itemset must also be frequent, and the itemsets it finds are then used to build association rules for market basket analysis. "To build decision trees that classify new records" is incorrect: building decision trees for classification is the job of algorithms such as ID3, C4.5, and CART, not Apriori. "To create clusters that group similar data points" is incorrect: grouping similar points is clustering, done with algorithms such as k-means or DBSCAN; Apriori looks for items that occur together, not for similar records. "To flag rare outliers that signal unusual charges" is incorrect: flagging rare, unusual records is anomaly detection; Apriori deliberately searches for frequent patterns rather than rare ones.

  2. What does the term 'overfitting' refer to in the context of data mining?

    • A.A model matched loosely to the underlying data
    • B.A model granted accidental access to the test data
    • C.A model fitted too closely to the training data
    • D.A model applied to the drifting data in production
    Show answerHide answer

    Correct answer: A model fitted too closely to the training data

    "A model fitted too closely to the training data" is correct: an overfit model learns the noise and quirks of its training data along with the real pattern, so it scores very well on that data but generalizes poorly to new, unseen data. "A model matched loosely to the underlying data" is incorrect: a model that fits the data too loosely to capture the underlying pattern is underfitting, which produces high error even on the data it was trained on. "A model granted accidental access to the test data" is incorrect: letting test information reach the model during training is data leakage; it inflates evaluation scores, but it is a flaw in how the data was split, not the definition of overfitting. "A model applied to the drifting data in production" is incorrect: a model losing accuracy because the data it receives in production has changed since training is data drift (or concept drift), which happens after deployment and is different from fitting the training data too closely.

  3. Which data mining technique is primarily used for identifying clusters of similar data points within a dataset?

    • A.Factor analysis
    • B.Cluster analysis
    • C.Regression analysis
    • D.Diagnostic analysis
    Show answerHide answer

    Correct answer: Cluster analysis

    "Cluster analysis" is correct: cluster analysis (clustering) is an unsupervised technique, using methods such as k-means or hierarchical clustering, that groups data points so points in the same cluster are more similar to each other than to points in other clusters. "Factor analysis" is incorrect: factor analysis groups correlated variables (columns) into underlying factors; it does not group the data points (rows) themselves. "Regression analysis" is incorrect: regression analysis models how a dependent variable changes with one or more predictors in order to predict values; it does not group data points by similarity. "Diagnostic analysis" is incorrect: diagnostic analysis drills into data to explain why an outcome occurred, for example by comparing segments or tracing causes; it explains results rather than discovering clusters of similar points.

  4. In data mining, what is the primary function of a decision tree?

    • A.To classify data into predefined categories
    • B.To partition unlabeled data into discovered subgroups
    • C.To transform data into principal components
    • D.To distill data into descriptive statistics
    Show answerHide answer

    Correct answer: To classify data into predefined categories

    "To classify data into predefined categories" is correct: a decision tree is a supervised method that splits records on attribute values, branch by branch, until each leaf assigns a record to one of the known classes, which makes it a standard classification tool. "To partition unlabeled data into discovered subgroups" is incorrect: discovering natural subgroups in unlabeled data is clustering, an unsupervised task; a decision tree learns from labeled examples. "To transform data into principal components" is incorrect: transforming variables into principal components is principal component analysis, which is used to reduce dimensionality. "To distill data into descriptive statistics" is incorrect: summarizing data with means, counts, and other descriptive statistics is descriptive analysis, not the classification a decision tree performs.

  5. What is a key characteristic of the K-Means clustering algorithm in data mining?

    • A.It requires the classes of records to be supplied in advance
    • B.It requires the radius of neighborhoods to be set in advance
    • C.It automatically determines the number of clusters
    • D.It needs the number of clusters specified beforehand
    Show answerHide answer

    Correct answer: It needs the number of clusters specified beforehand

    "It needs the number of clusters specified beforehand" is correct: k-means needs the analyst to choose k, the number of clusters, before it runs; it then assigns each point to the nearest centroid and recalculates the centroids until the clusters stabilize. "It requires the classes of records to be supplied in advance" is incorrect: needing labeled records is a property of supervised methods such as k-nearest neighbors; k-means is unsupervised and works on unlabeled data. "It requires the radius of neighborhoods to be set in advance" is incorrect: a neighborhood radius (ϵ\epsilon) is a parameter of density-based clustering such as DBSCAN, which finds the number of clusters on its own; k-means has no radius parameter. "It automatically determines the number of clusters" is incorrect: k-means cannot choose k itself; the analyst must set the number of clusters, often after comparing several values with the elbow method or silhouette scores.

  6. Which technique in data mining is used for identifying unusual patterns that might signify fraudulent activity?

    • A.Association rules
    • B.Cluster analysis
    • C.Anomaly detection
    • D.Data aggregation
    Show answerHide answer

    Correct answer: Anomaly detection

    "Anomaly detection" is correct: anomaly (outlier) detection identifies observations or patterns that deviate markedly from normal behavior, which is why it is used to flag possible fraud such as unusual transactions. "Association rules" is incorrect: association rules find items that frequently occur together, such as products bought in the same basket; they describe common patterns, not unusual ones. "Cluster analysis" is incorrect: cluster analysis groups similar records together to describe the groups in the data; flagging rare, unusual activity is the task of anomaly detection. "Data aggregation" is incorrect: aggregation rolls detailed records up into totals and averages, which can hide the individual unusual transactions that signal fraud.

  7. What is the primary goal of feature selection in the data mining process?

    • A.To draw a sample of historical records for training a model
    • B.To pick a subset of relevant variables for building a model
    • C.To generate a set of composite features for fitting a model
    • D.To standardize a range of numeric values for tuning a model
    Show answerHide answer

    Correct answer: To pick a subset of relevant variables for building a model

    "To pick a subset of relevant variables for building a model" is correct: feature selection keeps the input variables (features) that carry the most useful information and drops irrelevant or redundant ones, which simplifies the model, reduces overfitting, and speeds up training. "To draw a sample of historical records for training a model" is incorrect: drawing a subset of rows is sampling; feature selection chooses columns (variables), not records. "To generate a set of composite features for fitting a model" is incorrect: creating new composite features from existing ones is feature extraction or engineering, as in principal component analysis; feature selection keeps a subset of the original variables. "To standardize a range of numeric values for tuning a model" is incorrect: rescaling numeric inputs to a common scale is feature scaling or standardization, a preprocessing step that keeps every variable rather than choosing among them.

  8. Which data mining method is typically used for analyzing data with a time-based component?

    • A.Time series analysis
    • B.Root cause analysis
    • C.Spatial data analysis
    • D.Text analysis
    Show answerHide answer

    Correct answer: Time series analysis

    "Time series analysis" is correct: time series analysis studies observations recorded in time order, such as daily sales, to find trends, seasonality, and cycles and to forecast future values. "Root cause analysis" is incorrect: root cause analysis traces a problem back to its underlying cause; it is a problem-solving method, not a technique for analyzing time-ordered data. "Spatial data analysis" is incorrect: spatial analysis studies data tied to locations, such as coordinates and regions; its focus is where things happen, not when. "Text analysis" is incorrect: text analysis examines the words and themes in documents and other text; it is not built around a time-based component.

  9. In data mining, what is the purpose of using the Support Vector Machine (SVM) algorithm?

    • A.To group data into dense clusters
    • B.To render data into visual charts
    • C.To convert data into binary codes
    • D.To classify data into two classes
    Show answerHide answer

    Correct answer: To classify data into two classes

    "To classify data into two classes" is correct: a support vector machine is a supervised classifier that finds the hyperplane separating two classes with the widest possible margin; problems with more classes are handled by combining several of these two-class models. "To group data into dense clusters" is incorrect: finding dense groups in unlabeled data is density-based clustering, such as DBSCAN; an SVM learns from labeled examples. "To render data into visual charts" is incorrect: charting data is visualization; an SVM is a predictive model, not a charting tool. "To convert data into binary codes" is incorrect: turning categories into 0/1 indicator columns is encoding, such as one-hot encoding, a preprocessing step that may come before an SVM but is not its purpose.

  10. What is the role of 'lift' in association rule mining?

    • A.To measure how reliably a rule predicts consequent items
    • B.To measure how frequently a rule itemset appears overall
    • C.To measure how sharply a rule boosts promotional revenue
    • D.To measure how strongly a rule outperforms random chance
    Show answerHide answer

    Correct answer: To measure how strongly a rule outperforms random chance

    "To measure how strongly a rule outperforms random chance" is correct: lift compares a rule's confidence with how common the consequent is on its own (lift equals confidence divided by the support of the consequent), so a lift above 1 means the items occur together more often than random chance would predict. "To measure how reliably a rule predicts consequent items" is incorrect: that is confidence, the share of transactions containing the antecedent that also contain the consequent; it ignores how common the consequent already is. "To measure how frequently a rule itemset appears overall" is incorrect: that is support, the share of all transactions that contain every item in the rule. "To measure how sharply a rule boosts promotional revenue" is incorrect: lift in association rule mining is a ratio of probabilities, not a measure of the sales or revenue a promotion produces.

  11. Which algorithm in data mining is best suited for handling categorical data in classification problems?

    • A.Naive Bayes Classifier
    • B.Support Vector Machine
    • C.Artificial Neural Network
    • D.Kernel Density Estimation
    Show answerHide answer

    Correct answer: Naive Bayes Classifier

    "Naive Bayes Classifier" is correct: Naive Bayes applies Bayes' theorem with the assumption that predictors are independent and estimates its probabilities directly from category frequencies, so it handles categorical predictors naturally and is a standard choice for text and other categorical classification. "Support Vector Machine" is incorrect: SVMs work with numeric features and distances, so categorical data must first be encoded as numbers; they are not the best fit for categorical inputs. "Artificial Neural Network" is incorrect: neural networks also need categorical inputs converted to numeric encodings and usually require far more data and tuning; they are not specifically suited to categorical data. "Kernel Density Estimation" is incorrect: kernel density estimation estimates the probability distribution of continuous variables; it is not a classifier built for categorical data.

  12. In data mining, what is the main purpose of using Principal Component Analysis (PCA)?

    • A.To normalize the ranges of columns in the data
    • B.To predict the direction of trends in the data
    • C.To determine the count of clusters in the data
    • D.To reduce the number of dimensions in the data
    Show answerHide answer

    Correct answer: To reduce the number of dimensions in the data

    "To reduce the number of dimensions in the data" is correct: PCA transforms correlated variables into a smaller set of uncorrelated principal components that keep most of the variance, reducing the dimensionality of the data. "To normalize the ranges of columns in the data" is incorrect: normalizing or standardizing columns is a preprocessing step often done before PCA, not the purpose of PCA itself. "To predict the direction of trends in the data" is incorrect: forecasting trends is the job of time series and regression methods; PCA does not predict values. "To determine the count of clusters in the data" is incorrect: choosing the number of clusters relies on methods such as the elbow method or silhouette scores; PCA can make clusters easier to see but does not determine how many there are.

  13. Which data mining technique is used for predicting future values based on historical data?

    • A.Correlation Analysis
    • B.Regression Analysis
    • C.Descriptive Analysis
    • D.Cluster Analysis
    Show answerHide answer

    Correct answer: Regression Analysis

    "Regression Analysis" is correct: regression models the relationship between a dependent variable and one or more predictors using historical data, and the fitted equation is then used to predict future values, as in sales or demand forecasting. "Correlation Analysis" is incorrect: correlation measures the strength and direction of the relationship between two variables but does not produce an equation for predicting values. "Descriptive Analysis" is incorrect: descriptive analysis summarizes what has already happened, using measures such as averages and totals; it does not predict future values. "Cluster Analysis" is incorrect: cluster analysis groups similar records together; it is a descriptive, unsupervised technique, not a method for predicting values.

  14. In data mining, what is the primary role of a confusion matrix?

    • A.To show how strongly the variables correlate
    • B.To determine how quickly the model converges
    • C.To visualize how well the algorithm performs
    • D.To measure how cleanly the clusters separate
    Show answerHide answer

    Correct answer: To visualize how well the algorithm performs

    To visualize how well the algorithm performs is the role of a confusion matrix: it lays a classifier's predicted classes against the actual classes in a table, so true positives, false positives, true negatives and false negatives can be read at a glance and metrics such as precision and recall can be derived from them. To show how strongly the variables correlate is the job of a correlation matrix, which compares variables with one another rather than predictions with outcomes. To determine how quickly the model converges calls for a learning or loss curve tracked over training iterations, not a single table of results. To measure how cleanly the clusters separate describes silhouette analysis for unsupervised clustering, where there are no actual class labels to tabulate.

  15. What is the primary benefit of using the Random Forest algorithm in data mining?

    • A.It uses gradient boosting to reduce earlier errors
    • B.It uses fitted coefficients to model linear trends
    • C.It uses ensemble learning to deliver high accuracy
    • D.It uses one decision tree to remain understandable
    Show answerHide answer

    Correct answer: It uses ensemble learning to deliver high accuracy

    It uses ensemble learning to deliver high accuracy is the main benefit of Random Forest: it trains many decision trees on bootstrap samples, each split considering a random subset of features, and combines their votes or averages, which cuts the variance and overfitting of any single tree. It uses gradient boosting to reduce earlier errors describes boosting methods such as gradient-boosted trees, which build trees one after another to correct earlier mistakes, whereas a random forest grows its trees independently. It uses fitted coefficients to model linear trends describes linear regression; a random forest fits no coefficients and extrapolates linear trends poorly. It uses one decision tree to remain understandable describes a single decision tree, whose short list of rules is easy to read, while a forest of hundreds of trees gives up that interpretability in exchange for accuracy.

  16. What is the purpose of 'Gini Index' in the context of decision tree classifiers?

    • A.To determine how the inputs are scaled
    • B.To determine how the nodes are split
    • C.To determine how the votes are counted
    • D.To determine how the models are evaluated
    Show answerHide answer

    Correct answer: To determine how the nodes are split

    To determine how the nodes are split is the purpose of the Gini index (Gini impurity) in a decision tree classifier: at each node the tree scores candidate splits by how mixed the classes would be in the resulting child nodes and chooses the split that reduces impurity the most, giving the purest possible subsets. To determine how the inputs are scaled is a preprocessing choice such as normalization or standardization, which decision trees do not need and which the Gini index plays no part in. To determine how the votes are counted is the aggregation rule of an ensemble such as a random forest, for example a majority vote across trees. To determine how the models are evaluated is a separate step after training: a model's accuracy is measured on its predictions, for example with a confusion matrix, not by the impurity score used to choose splits.

  17. Which algorithm is best suited for identifying unusual patterns in high-dimensional datasets?

    • A.Linear SVM
    • B.LDA Classifier
    • C.One-Class SVM
    • D.KNN Classifier
    Show answerHide answer

    Correct answer: One-Class SVM

    One-Class SVM is best suited: it learns a boundary around the normal observations using only normal (unlabeled) data, and its kernel lets it work well in many dimensions, so new points falling outside that boundary are flagged as unusual. Linear SVM is a two-class classifier that needs labeled examples of both classes, so it cannot learn what is unusual from normal data alone. LDA Classifier (linear discriminant analysis) separates known, labeled classes and assumes each class is normally distributed with a shared covariance, which does not describe rare, unlabeled anomalies. KNN Classifier assigns labels by a majority vote of nearby labeled points, and its distance calculations become unreliable as the number of dimensions grows.

  18. In data mining, what is the primary use of the 'Lift' measure in association rule learning?

    • A.To measure how frequently paired items appear
    • B.To measure how strongly two variables relate
    • C.To measure how reliably rules predict results
    • D.To measure how broadly antecedent items occur
    Show answerHide answer

    Correct answer: To measure how strongly two variables relate

    To measure how strongly two variables relate is the use of lift in association rule learning: it compares how often the antecedent and consequent of a rule occur together with how often they would occur together if they were independent, so a lift above 1 signals a positive association, a lift of 1 signals independence and a lift below 1 signals a negative association. To measure how frequently paired items appear describes support, the share of all transactions that contain the whole itemset. To measure how reliably rules predict results describes confidence, the share of transactions containing the antecedent that also contain the consequent. To measure how broadly antecedent items occur describes coverage, the support of the antecedent on its own, which ignores the consequent.

  19. What role does 'Cross-validation' play in the data mining process?

    • A.Deploying models
    • B.Validating data
    • C.Evaluating models
    • D.Transforming data
    Show answerHide answer

    Correct answer: Evaluating models

    Evaluating models is the role of cross-validation: the data are split into folds, the model is trained on all but one fold and tested on the held-out fold in turn, and the scores are averaged to estimate how well it will generalize to new data and to expose overfitting. Deploying models means putting a finished model into production, which comes after evaluation. Validating data means checking records against rules for type, range and completeness, a data-quality task despite the similar name. Transforming data means reshaping, recoding or aggregating fields during preparation, before any model is trained.

  20. Which data mining technique is used primarily for categorizing or grouping similar data points together?

    • A.Factor Analysis
    • B.Correlation Analysis
    • C.Association Analysis
    • D.Cluster Analysis
    Show answerHide answer

    Correct answer: Cluster Analysis

    Cluster Analysis groups data points so that points in the same cluster are more similar to one another than to points in other clusters, without predefined labels; k-means and hierarchical clustering are common methods. Factor Analysis groups correlated variables into a smaller set of underlying factors, so it condenses columns rather than grouping records. Correlation Analysis measures how strongly pairs of variables move together and does not assign data points to groups. Association Analysis finds items that tend to occur together in transactions and expresses them as if-then rules rather than as groups of similar records.

  21. In data mining, what is a common use of the DBSCAN algorithm?

    • A.Splitting observations into preset spherical groups
    • B.Classifying labeled records into predefined classes
    • C.Grouping spatial points into density-based clusters
    • D.Converting frequent itemsets into association rules
    Show answerHide answer

    Correct answer: Grouping spatial points into density-based clusters

    Grouping spatial points into density-based clusters is a common use of DBSCAN (density-based spatial clustering of applications with noise): it links points that have enough neighbors within a set radius, so it finds clusters of any shape in spatial data, does not need the number of clusters in advance, and leaves isolated points as noise. Splitting observations into preset spherical groups describes k-means, which needs the number of clusters chosen beforehand and favors compact, round clusters. Classifying labeled records into predefined classes is supervised classification, which DBSCAN, an unsupervised method, does not perform. Converting frequent itemsets into association rules is the job of market basket algorithms such as Apriori.

  22. Which metric in data mining is crucial for evaluating the balance between sensitivity and specificity in a classification model?

    • A.F1 Score
    • B.L1 Loss
    • C.Log Loss
    • D.Hit Rate
    Show answerHide answer

    Correct answer: F1 Score

    F1 Score fits best among these: it folds two competing rates into one balanced figure by taking the harmonic mean of precision and recall (recall is the same quantity as sensitivity), so it stays low unless both are reasonably high. Strictly, F1 never uses true negatives, so a measure such as balanced accuracy is the one that averages sensitivity with specificity directly. L1 Loss is the mean absolute error used to fit and judge regression models, not a classification balance. Log Loss scores how well predicted probabilities match the actual classes, heavily penalizing confident mistakes, rather than balancing two rates. Hit Rate is simply sensitivity, the share of actual positives that are caught, so it measures only one side of the balance.

  23. Which statistical test is appropriate for determining if there is a significant difference between the means of two independent groups?

    • A.Paired t-test
    • B.Two-sample t-test
    • C.Chi-square test
    • D.Shapiro-Wilk test
    Show answerHide answer

    Correct answer: Two-sample t-test

    Two-sample t-test is the right choice: it compares the means of two independent groups, judging whether the difference between the sample means is large relative to its standard error, and it suits small samples when the data are roughly normal (Welch's version drops the equal-variance assumption). Paired t-test compares two measurements taken on the same subjects or on matched pairs, so it does not fit two independent groups. Chi-square test examines whether categorical variables are associated by comparing observed and expected counts, not means. Shapiro-Wilk test checks whether a sample comes from a normal distribution, an assumption often checked before running a t-test rather than a test for a difference in means.

  24. In data analysis, what does the term 'heteroscedasticity' refer to?

    • A.Non-constant variance of the error terms in a regression model
    • B.Serial dependence of the error terms in a regression model
    • C.Asymmetric distribution of the error terms in a regression model
    • D.Nonzero average of the error terms in a regression model
    Show answerHide answer

    Correct answer: Non-constant variance of the error terms in a regression model

    Non-constant variance of the error terms in a regression model is what heteroscedasticity means: the spread of the residuals is not constant but grows or shrinks across the range of fitted values or predictors, often visible as a funnel shape in a residual plot. It violates the equal-variance assumption of ordinary least squares, making the coefficient estimates inefficient and their standard errors unreliable. Serial dependence of the error terms in a regression model is autocorrelation, where each residual is related to the ones before it, as is common in time series. Asymmetric distribution of the error terms in a regression model is skewness, a departure from normality, which is a separate assumption. Nonzero average of the error terms in a regression model means the residuals are systematically biased, for example when a needed intercept or variable has been left out.

  25. Which method is used in data analysis to reduce the number of features in a dataset by transforming original variables into a new set of variables?

    • A.Hierarchical Cluster Analysis (HCA)
    • B.Principal Component Analysis (PCA)
    • C.Recursive Feature Elimination (RFE)
    • D.Sequential Feature Selection (SFS)
    Show answerHide answer

    Correct answer: Principal Component Analysis (PCA)

    Principal Component Analysis (PCA) reduces the number of features by transforming the original, possibly correlated variables into a new set of uncorrelated variables called principal components, ordered by how much of the variance each one explains, so the first few can stand in for many. Hierarchical Cluster Analysis (HCA) groups observations into a tree of nested clusters and creates no new variables. Recursive Feature Elimination (RFE) reduces features by repeatedly dropping the weakest original variables, so the variables it keeps are unchanged rather than transformed. Sequential Feature Selection (SFS) likewise adds or removes original features one at a time, choosing a subset instead of building new combined variables.

  26. In the context of data analysis, what does 'multicollinearity' refer to?

    • A.Interaction terms among the predictor variables in a regression model
    • B.Serial dependence among the observed residuals in a regression model
    • C.High correlations among the predictor variables in a regression model
    • D.Unequal variances among the observed residuals in a regression model
    Show answerHide answer

    Correct answer: High correlations among the predictor variables in a regression model

    High correlations among the predictor variables in a regression model is what multicollinearity means: two or more predictors move together so closely that one can be largely predicted from the others, which inflates the standard errors of the coefficients and makes their individual effects hard to separate; a variance inflation factor (VIF) is a common check. Interaction terms among the predictor variables in a regression model are products of predictors added so that the effect of one can depend on the level of another, which can exist even when the predictors are uncorrelated. Serial dependence among the observed residuals in a regression model is autocorrelation. Unequal variances among the observed residuals in a regression model is heteroscedasticity.

  27. Which data analysis technique is used to categorize or group elements of a dataset based on a set of predefined criteria?

    • A.Performance analysis
    • B.Cluster analysis
    • C.Survival analysis
    • D.Correlation analysis
    Show answerHide answer

    Correct answer: Cluster analysis

    Cluster analysis is the technique for grouping the elements of a dataset: it assigns records to clusters so that members of a cluster are similar to one another on the chosen criteria, such as distance across a set of attributes, and dissimilar to members of other clusters. Performance analysis tracks measurements against defined goals or targets rather than sorting records into groups. Survival analysis models the time until an event, such as churn or equipment failure, occurs. Correlation analysis measures how strongly pairs of variables are related, without assigning records to groups.

  28. In data analysis, what is the purpose of a confusion matrix?

    • A.To compare a model's residuals with the predicted values
    • B.To compare a model's predictions with the actual values
    • C.To compare a model's gains with the random baseline
    • D.To compare a model's accuracy with the majority baseline
    Show answerHide answer

    Correct answer: To compare a model's predictions with the actual values

    To compare a model's predictions with the actual values is the purpose of a confusion matrix: in a classification problem it cross-tabulates the predicted classes against the true classes, so the counts of true positives, false positives, true negatives and false negatives show exactly where the model is right and how it errs. To compare a model's residuals with the predicted values describes a residual plot, a regression diagnostic for patterns in the errors. To compare a model's gains with the random baseline describes a lift or cumulative gains chart. To compare a model's accuracy with the majority baseline is a benchmark check against always predicting the most common class, not the job of the matrix itself.

  29. Which of the following is a key indicator of a good fit for a regression model?

    • A.A high model P-value
    • B.A large standard error
    • C.A large feature count
    • D.A high R-squared value
    Show answerHide answer

    Correct answer: A high R-squared value

    A high R-squared value indicates a good fit because R-squared is the proportion of the variance in the dependent variable that the model explains, so values close to 1 mean the fitted line tracks the data closely (adjusted R-squared also guards against inflating it with extra predictors). A high model P-value means the overall F-test cannot rule out that the predictors explain nothing, which signals a weak model. A large standard error means the observed values sit far from the fitted line on average, the opposite of a good fit. A large feature count can inflate R-squared and invite overfitting, so having many predictors is not in itself evidence of a good fit.

  30. In data analysis, which method is commonly used for predicting categorical outcomes?

    • A.Linear regression
    • B.Exponential regression
    • C.Curvilinear regression
    • D.Logistic regression
    Show answerHide answer

    Correct answer: Logistic regression

    Logistic regression is the standard method for predicting a categorical outcome: it models the log-odds of class membership as a linear function of the predictors and converts them through the logistic (sigmoid) function into probabilities between 0 and 1, which are then assigned to classes. Linear regression predicts a continuous numeric outcome and can return values outside any class range. Exponential regression fits a growth or decay curve to a continuous outcome. Curvilinear regression fits a curved relationship, such as a polynomial, to a continuous outcome rather than assigning categories.

  31. What is the primary purpose of conducting a time-series analysis in data analysis?

    • A.To spot trends and patterns across successive periods
    • B.To compare means and variances across treatment groups
    • C.To map values and densities across geographic regions
    • D.To estimate hazards and survival across patient groups
    Show answerHide answer

    Correct answer: To spot trends and patterns across successive periods

    To spot trends and patterns across successive periods is the main purpose of time-series analysis: observations recorded in time order are examined for long-term trend, seasonality, cycles and irregular movements, which also supports forecasting future values. To compare means and variances across treatment groups is the job of tests such as ANOVA, which compare groups rather than follow one measure through time. To map values and densities across geographic regions is geospatial analysis. To estimate hazards and survival across patient groups is survival analysis, which models the time until an event occurs rather than the movement of a series.

  32. In data analysis, what does a high kurtosis in a dataset indicate?

    • A.A distribution with a skewed or asymmetric shape
    • B.A distribution with a flat or uniform shape
    • C.A distribution with double or multiple peaks
    • D.A distribution with extreme or outlier values
    Show answerHide answer

    Correct answer: A distribution with extreme or outlier values

    A distribution with extreme or outlier values is what high kurtosis indicates: kurtosis measures the weight of a distribution's tails, so a high (leptokurtic) value means more observations lie far from the mean than a normal distribution would produce, whatever the shape of the peak. A distribution with a skewed or asymmetric shape is described by skewness, not kurtosis. A distribution with a flat or uniform shape has light tails and therefore low (platykurtic) kurtosis. A distribution with double or multiple peaks is bimodal or multimodal, which tends to lower kurtosis rather than raise it.

  33. Which method is most suitable for identifying underlying dimensions or factors that explain observed correlations between variables?

    • A.Factor analysis
    • B.Cluster analysis
    • C.Correlation analysis
    • D.Time series analysis
    Show answerHide answer

    Correct answer: Factor analysis

    Factor analysis is designed for this: it models the observed variables as combinations of a smaller number of unobserved factors, so variables that load on the same factor share an underlying dimension, which accounts for the correlations among them. Cluster analysis groups observations (rows) by similarity rather than explaining why variables are correlated. Correlation analysis measures how strongly pairs of variables are related but does not identify what lies behind those correlations. Time series analysis studies how a measure moves over time.

  34. In data analysis, which technique is used for dividing data into training and testing sets for model validation?

    • A.Data augmentation
    • B.Cross-validation
    • C.Feature selection
    • D.Regularization
    Show answerHide answer

    Correct answer: Cross-validation

    Cross-validation is the technique: the data are divided into k folds, the model is trained on all but one fold and tested on the fold held out, and this repeats until every fold has served as the test set, giving a more reliable estimate of performance than a single split. Data augmentation creates modified copies of existing training examples to enlarge the training set rather than dividing it. Feature selection chooses which input variables a model uses. Regularization adds a penalty for model complexity to reduce overfitting during training.

  35. Which technique is used to estimate the relationship between a dependent variable and one or more independent variables in data analysis?

    • A.Data clustering
    • B.Covariance matrix
    • C.Factor analysis
    • D.Linear regression
    Show answerHide answer

    Correct answer: Linear regression

    Linear regression estimates how a dependent variable changes with one or more independent variables by fitting the line or plane that minimizes the sum of squared errors, giving a coefficient for each predictor. Data clustering, such as k-means, groups records by similarity and has no dependent variable to predict. Covariance matrix lists how each pair of variables varies together, which shows association but does not model one variable as a function of the others. Factor analysis reduces many correlated variables to a few underlying factors and likewise has no dependent variable.

  36. What does the term 'outlier' refer to in data analysis?

    • A.A value that marks the exact midpoint of a dataset
    • B.A value that marks the third quartile of a dataset
    • C.A value that measures the full spread of a dataset
    • D.A value that breaks the usual pattern of a dataset
    Show answerHide answer

    Correct answer: A value that breaks the usual pattern of a dataset

    A value that breaks the usual pattern of a dataset is an outlier: it lies an unusual distance from the other observations, for example more than 1.5 times the interquartile range beyond the first or third quartile, or several standard deviations from the mean, and it may reflect an error or a genuinely rare case. A value that marks the exact midpoint of a dataset is the median. A value that marks the third quartile of a dataset is the 75th percentile, which helps set outlier fences but is an ordinary summary value itself. A value that measures the full spread of a dataset is the range, the difference between the largest and smallest values.

  37. Which method in data analysis is used to assign weights to different variables based on their importance?

    • A.Weighted average
    • B.Cumulative average
    • C.Arithmetic average
    • D.Moving average
    Show answerHide answer

    Correct answer: Weighted average

    Weighted average is the method: each value is multiplied by a weight that reflects its importance, the products are summed, and the total is divided by the sum of the weights, so more important variables pull the result further. Cumulative average recalculates the plain mean of all values recorded so far, giving each one equal weight. Arithmetic average adds the values and divides by their count, which also treats every value equally. Moving average takes the mean over a sliding window of recent periods to smooth a time series, not to weight variables by importance.

  38. In data analysis, what is the purpose of using Spearman's rank correlation coefficient?

    • A.To measure the strength and direction of linear correlation between two measures
    • B.To compare the ranks and medians of measurements between two independent samples
    • C.To compare the signs and ranks of differences between two dependent samples
    • D.To gauge the direction and strength of association between two ranked variables
    Show answerHide answer

    Correct answer: To gauge the direction and strength of association between two ranked variables

    To gauge the direction and strength of association between two ranked variables is the purpose of Spearman's rank correlation coefficient: the values are converted to ranks and then correlated, giving a coefficient from -1 to +1 that captures any monotonic relationship, so it suits ordinal data or skewed numeric data and is little affected by outliers. To measure the strength and direction of linear correlation between two measures describes Pearson's correlation coefficient, which uses the raw values and assumes a linear relationship. To compare the ranks and medians of measurements between two independent samples describes the Mann-Whitney U test, which compares groups rather than measuring association. To compare the signs and ranks of differences between two dependent samples describes the Wilcoxon signed-rank test for paired or matched measurements.

  39. What is the main objective of using decision trees in data analysis?

    • A.To cluster raw data points into similar groups
    • B.To compress variables into principal components
    • C.To condense transactions into association rules
    • D.To classify records into distinct categories
    Show answerHide answer

    Correct answer: To classify records into distinct categories

    To classify records into distinct categories is the main objective of a decision tree: it learns a sequence of if-then splits on the input features, each branch narrowing the possibilities, until every leaf assigns a class label (regression trees apply the same idea to numeric targets). To cluster raw data points into similar groups is the goal of unsupervised clustering methods such as k-means, which work without predefined classes. To compress variables into principal components is the goal of principal component analysis. To condense transactions into association rules is the goal of association rule mining, such as market basket analysis.

  40. A data analyst is summarizing weekly support-ticket counts for five weeks: 12, 19, 7, 23, and 15. The team lead asks for the range of the dataset. What value should the analyst report, and what does it represent?

    • A.11.5, the spread between the lower and upper quartile values
    • B.16, the gap between the largest and smallest figures listed
    • C.6.2, the typical spread between the mean and observed values
    • D.3, the gap between the earliest and latest figures recorded
    Show answerHide answer

    Correct answer: 16, the gap between the largest and smallest figures listed

    The range is 16, the gap between the largest and smallest figures listed: 23 minus 7. The value 11.5, the spread between the lower and upper quartile values, is the interquartile range (21 minus 9.5, using the medians of the lower and upper halves), and 6.2, the typical spread between the mean and observed values, is the sample standard deviation; both describe spread, but neither is the range. The value 3, the gap between the earliest and latest figures recorded, subtracts the first week's count (12) from the last week's (15) without sorting, which shows change over time rather than spread.

  41. A retention analyst examines account tenures (in months) for five customers: 4, 4, 6, 8, and 18. Recognizing that the single 18-month account skews the average upward, which measure of central tendency best represents the typical customer's tenure, and what is its value?

    • A.The mean tenure of 8 months
    • B.The median of 6 months
    • C.The modal value of 4 months
    • D.The midrange of 11 months
    Show answerHide answer

    Correct answer: The median of 6 months

    The median of 6 months best represents the typical account, because the middle of the sorted tenures (4, 4, 6, 8, 18) is not pulled by the single long account. The mean tenure of 8 months is the true average but is inflated by the 18-month account, which is why the median is preferred for skewed data. The modal value of 4 months reflects only the two shortest accounts, and the midrange of 11 months, halfway between 4 and 18, depends wholly on the two extremes, so it is the measure most distorted by the outlier.

  42. While preparing a statistics summary, an analyst must explain how the mean, median, and mode differ to a non-technical stakeholder. Which description correctly distinguishes all three measures of central tendency?

    • A.The mean is the arithmetic average, the median is the middle of the two extremes, and the mode is a count of occurrences
    • B.The mean is the cumulative total, the median is the middle of the sorted data, and the mode is a count of occurrences
    • C.The mean is the cumulative total, the median is the middle of the two extremes, and the mode is the most frequent value
    • D.The mean is the arithmetic average, the median is the middle of the sorted data, and the mode is the most frequent value
    Show answerHide answer

    Correct answer: The mean is the arithmetic average, the median is the middle of the sorted data, and the mode is the most frequent value

    The mean is the arithmetic average, the median is the middle of the sorted data, and the mode is the most frequent value; these are the three standard measures of central tendency. The version saying the mean is the arithmetic average, the median is the middle of the two extremes, and the mode is a count of occurrences gives the median the definition of the midrange and treats the mode as a frequency rather than the value that occurs most often. The version saying the mean is the cumulative total, the median is the middle of the sorted data, and the mode is a count of occurrences treats the sum as the mean and makes the same mistake about the mode, and the version saying the mean is the cumulative total, the median is the middle of the two extremes, and the mode is the most frequent value gets both the mean and the median wrong.

  43. An analyst reports that exam scores have a standard deviation of 6 points and notes that the variance is 36. A colleague asks how variance and standard deviation relate. Which explanation is correct?

    • A.Variance is the square root of the standard deviation, so the standard deviation is its square and is expressed in squared units
    • B.Variance is the standard deviation divided by the sample size, and it keeps shrinking toward zero as more data are gathered
    • C.Variance is the standard deviation divided by the average, and it is expressed as a percentage to compare spread across datasets
    • D.Variance is the average squared deviation around the mean, and the standard deviation is its square root in the original units
    Show answerHide answer

    Correct answer: Variance is the average squared deviation around the mean, and the standard deviation is its square root in the original units

    Variance is the average squared deviation around the mean, and the standard deviation is its square root in the original units, which is why a variance of 36 corresponds to a standard deviation of 6. The claim that variance is the square root of the standard deviation, so the standard deviation is its square and is expressed in squared units, reverses the relationship. The claim that variance is the standard deviation divided by the sample size, and it keeps shrinking toward zero as more data are gathered, confuses variance with the standard error of the mean, which divides the standard deviation by the square root of n, and the claim that variance is the standard deviation divided by the average, and it is expressed as a percentage to compare spread across datasets, describes the coefficient of variation.

  44. A quality analyst observes a strong positive correlation between the number of ceiling fans sold and the number of heat-related illnesses in a city, both rising in summer. A manager wants to conclude that fan sales cause the illnesses. What is the most statistically sound response?

    • A.Strong correlation proves causation, and a rise in fan sales is the clear cause of the illnesses
    • B.The city-wide link is real, and each resident who buys a fan is therefore more prone to fall ill
    • C.Correlation does not imply causation, and a third factor such as hot weather likely drives both
    • D.A seasonal coincidence links both trends, and this pattern does not belong in a quality summary
    Show answerHide answer

    Correct answer: Correlation does not imply causation, and a third factor such as hot weather likely drives both

    Correlation does not imply causation, and a third factor such as hot weather likely drives both: summer heat raises fan purchases and heat-related illness at the same time. The claim that strong correlation proves causation, and a rise in fan sales is the clear cause of the illnesses, is the classic error, because strength shows how closely two variables move together, not why. The claim that the city-wide link is real, and each resident who buys a fan is therefore more prone to fall ill, applies a city-level pattern to individuals (the ecological fallacy) and still treats the fans as a cause, and the claim that a seasonal coincidence links both trends, and this pattern does not belong in a quality summary, dismisses a real, explainable pattern instead of identifying its common cause.

  45. A team is deciding whether to use correlation analysis or regression analysis on two numeric variables. They want to predict monthly revenue from advertising spend, not merely measure association. Which statement correctly distinguishes the two techniques?

    • A.Correlation is used to create the equations for forecasting, while regression merely reports how tightly the two tend to move together
    • B.Correlation measures the strength and direction of the link, while regression merely reports how tightly the two tend to move together
    • C.Correlation is used to create the equations for forecasting, while regression models an outcome against a predictor to estimate values
    • D.Correlation measures the strength and direction of the link, while regression models an outcome against a predictor to estimate values
    Show answerHide answer

    Correct answer: Correlation measures the strength and direction of the link, while regression models an outcome against a predictor to estimate values

    Correlation measures the strength and direction of the link, while regression models an outcome against a predictor to estimate values, which is why regression is the technique for predicting revenue from advertising spend. The statement that correlation is used to create the equations for forecasting, while regression merely reports how tightly the two tend to move together reverses both roles. The statement that correlation measures the strength and direction of the link, while regression merely reports how tightly the two tend to move together describes correlation correctly but reduces regression to a measure of association and ignores its predictive equation, and the statement that correlation is used to create the equations for forecasting, while regression models an outcome against a predictor to estimate values describes regression correctly but gives correlation a predictive role it does not have.

  46. After running a hypothesis test comparing two marketing campaigns, an analyst obtains a p-value of 0.02 against a preset significance level (α\alpha) of 0.05. How should the analyst interpret this result?

    • A.Reject the null hypothesis and treat the result as statistically significant, because the p-value falls below the alpha level fixed in advance
    • B.Retain the null hypothesis and conclude the campaigns are equally successful, because a 2 percent chance remains that the values arose by luck
    • C.Reject the null hypothesis and treat the result as statistically significant, because a 2 percent chance remains that the values arose by luck
    • D.Retain the null hypothesis and conclude the campaigns are equally successful, because the p-value falls below the alpha level fixed in advance
    Show answerHide answer

    Correct answer: Reject the null hypothesis and treat the result as statistically significant, because the p-value falls below the alpha level fixed in advance

    Reject the null hypothesis and treat the result as statistically significant, because the p-value falls below the alpha level fixed in advance: 0.02 is smaller than 0.05, and alpha is the Type I error rate the analyst agreed to accept before testing. The option to retain the null hypothesis and conclude the campaigns are equally successful, because a 2 percent chance remains that the values arose by luck gets both the decision and the meaning of the p-value wrong. The option to reject the null hypothesis and treat the result as statistically significant, because a 2 percent chance remains that the values arose by luck reaches the right decision for the wrong reason, because a p-value is the probability of seeing data this extreme if the null hypothesis were true, not the probability that luck produced the observed values. The option to retain the null hypothesis and conclude the campaigns are equally successful, because the p-value falls below the alpha level fixed in advance reverses the decision rule.

Visualization and Reporting (38)

  1. What is the primary purpose of using a box plot in data analysis?

    • A.To show the central tendency and spread of a dataset
    • B.To show the trends and seasonality of a time series
    • C.To show the various parts and proportions of a whole
    • D.To show the nested levels and sizes of a hierarchy
    Show answerHide answer

    Correct answer: To show the central tendency and spread of a dataset

    To show the central tendency and spread of a dataset is the purpose of a box plot: the line inside the box marks the median, the box spans the interquartile range, the whiskers extend toward the smallest and largest typical values, and points beyond them are flagged as outliers, which also makes box plots easy to compare across groups. To show the trends and seasonality of a time series is the job of a line chart plotted over time. To show the various parts and proportions of a whole describes a pie or stacked chart. To show the nested levels and sizes of a hierarchy describes a treemap or a dendrogram.

  2. What is the purpose of using a scatter plot in data analysis?

    • A.To show how much one variable tracks another
    • B.To show how much one factor outweighs another
    • C.To show how much one quarter outpaces another
    • D.To show how much one product outsells another
    Show answerHide answer

    Correct answer: To show how much one variable tracks another

    To show how much one variable tracks another is the purpose of a scatter plot: each dot places one observation by its values on two numeric variables, one on the x-axis and one on the y-axis, so the pattern of dots reveals the direction, form and strength of the relationship along with any outliers, although it does not by itself prove cause and effect. To show how much one factor outweighs another is the role of a Pareto chart, which ranks categories by their contribution. To show how much one quarter outpaces another is a comparison of time periods, usually drawn as a line or column chart. To show how much one product outsells another is a category comparison suited to a bar chart.

  3. What is the primary use of a histogram in data analysis?

    • A.To show the relationship of two numerical variables
    • B.To show the totals of separate data categories
    • C.To show the distribution of one continuous variable
    • D.To show the ranking of separate defect categories
    Show answerHide answer

    Correct answer: To show the distribution of one continuous variable

    To show the distribution of one continuous variable is the primary use of a histogram: the range of values is divided into adjacent bins and the height of each bar shows how many observations fall in that bin, revealing the shape, center, spread and skew of the data. To show the relationship of two numerical variables calls for a scatter plot. To show the totals of separate data categories is the job of a bar chart, whose bars stand for distinct categories rather than adjoining numeric ranges. To show the ranking of separate defect categories describes a Pareto chart, which sorts categories from most to least frequent.

  4. In the context of data analysis, what is the role of a Pareto chart?

    • A.To flag the special cause signals in a process
    • B.To identify the vital few factors in a dataset
    • C.To group the possible root causes in a fishbone
    • D.To bin the raw continuous values in a column
    Show answerHide answer

    Correct answer: To identify the vital few factors in a dataset

    To identify the vital few factors in a dataset is the role of a Pareto chart: categories are drawn as bars in descending order of frequency or impact, with a line showing the cumulative percentage, so the few most significant factors that account for most of the problem (the 80/20 principle) stand out. To flag the special cause signals in a process is the job of a control chart, which plots a measure against control limits over time. To group the possible root causes in a fishbone describes a cause-and-effect (Ishikawa) diagram, which groups possible causes by category without ranking them. To bin the raw continuous values in a column describes a histogram of a single numeric variable.

  5. Which type of chart is most appropriate for displaying the relationship between two quantitative variables?

    • A.Line chart
    • B.Pareto chart
    • C.Scatter plot
    • D.Box plot
    Show answerHide answer

    Correct answer: Scatter plot

    Scatter plot is the most appropriate: each point is placed by its values on two quantitative variables, so the pattern of points shows whether they are related, in which direction, how strongly and whether the relationship is linear. Line chart connects values in order, usually over time, and is best for the trend of one measure across an ordered sequence. Pareto chart ranks categories by frequency with a cumulative-percentage line. Box plot summarizes the distribution of a single variable, or compares it across groups, through its median, quartiles and outliers.

  6. When visualizing time series data, which type of chart is typically the most effective?

    • A.Line chart
    • B.Pie chart
    • C.Stacked bar chart
    • D.Pareto chart
    Show answerHide answer

    Correct answer: Line chart

    Line chart is typically the most effective choice for time series data: time runs along the horizontal axis and the connected points make trends, seasonality and turning points easy to see. Pie chart shows how parts make up a single whole at one moment and cannot show change over time. Stacked bar chart can show each period's total and its composition, but the stacked segments make the trend over time harder to follow than a connected line. Pareto chart ranks categories by frequency with a cumulative-percentage line, so its order follows size rather than time.

  7. In data visualization, what is the primary purpose of using a heat map?

    • A.To display the density or concentration of values
    • B.To display the levels or subgroups of hierarchies
    • C.To display the shares or percentages of totals
    • D.To display the totals or counts of categories
    Show answerHide answer

    Correct answer: To display the density or concentration of values

    To display the density or concentration of values is the primary purpose of a heat map: each cell of a grid or region of a map is shaded by color intensity, so areas of high and low concentration of activity stand out at a glance. To display the levels or subgroups of hierarchies is the job of a treemap or tree diagram. To display the shares or percentages of totals is the role of a pie or donut chart. To display the totals or counts of categories is best done with a bar chart, whose bar lengths are easier to compare than shades of color.

  8. What is the main advantage of using a histogram over a bar chart for data visualization?

    • A.Comparing the counts of several discrete categories
    • B.Tracking the trend of quarterly revenue figures
    • C.Showing the distribution of one continuous variable
    • D.Summarizing the correlation of two numeric measures
    Show answerHide answer

    Correct answer: Showing the distribution of one continuous variable

    Showing the distribution of one continuous variable is what a histogram does better than a bar chart: it divides the variable's range into adjacent bins and draws the frequency in each, so the shape, center, spread and skew of the values become visible. Comparing the counts of several discrete categories is the job of a bar chart, whose separated bars stand for categories rather than intervals on a number line. Tracking the trend of quarterly revenue figures calls for a line chart, which keeps values in time order; a histogram discards that order when it bins the values. Summarizing the correlation of two numeric measures needs a scatter plot, because a histogram describes only one variable at a time.

  9. Which visualization tool is best suited for displaying hierarchical data?

    • A.Tree map
    • B.Heat map
    • C.Pie chart
    • D.Bar chart
    Show answerHide answer

    Correct answer: Tree map

    Tree map is the chart built for hierarchical data: it nests rectangles inside larger rectangles, so each branch of the hierarchy is a block whose area shows its size and whose inner blocks show its children. Heat map uses color intensity in a grid of cells to show how large values are across two dimensions, but it has no nesting to show parent and child levels. Pie chart shows the parts of a single whole at one level only, so it cannot show categories within categories. Bar chart compares values across categories on a common axis, again at a single level with no parent and child structure.

  10. When would a box plot be most appropriate in data visualization?

    • A.Monitoring the trend of quarterly earnings
    • B.Showing the composition of annual budgets
    • C.Listing the individual values of observations
    • D.Comparing the distributions of several groups
    Show answerHide answer

    Correct answer: Comparing the distributions of several groups

    Comparing the distributions of several groups is where a box plot is most appropriate: side-by-side boxes show each group's median, quartiles, range and outliers on one scale, so differences in center and spread are easy to see. Monitoring the trend of quarterly earnings is a job for a line chart, which keeps values in time order. Showing the composition of annual budgets is a part-to-whole question for a pie chart or stacked bar chart, not a summary of spread. Listing the individual values of observations is the opposite of what a box plot does, since it condenses the data into a five-number summary; a table or dot plot shows every value.

  11. In data visualization, what is the primary benefit of using a radar chart?

    • A.Showing the proportional shares of overall spending
    • B.Tracking the historical trajectory of stock prices
    • C.Summarizing the distribution of statewide exam results
    • D.Comparing the values of several quantitative variables
    Show answerHide answer

    Correct answer: Comparing the values of several quantitative variables

    Comparing the values of several quantitative variables is the main benefit of a radar chart: each variable gets its own axis radiating from a common center, and the shape traced across the axes lets several items be compared on all of the variables at once. Showing the proportional shares of overall spending is the purpose of a pie chart; a radar chart's circular layout does not divide a whole into slices. Tracking the historical trajectory of stock prices calls for a line chart with time on the horizontal axis. Summarizing the distribution of statewide exam results is done with a histogram, which describes a single variable rather than many.

  12. What type of data visualization is most effective for comparing individual values to the whole?

    • A.Line chart
    • B.Scatter plot
    • C.Box plot
    • D.Pie chart
    Show answerHide answer

    Correct answer: Pie chart

    Pie chart is the standard choice for comparing individual values to the whole: each slice's angle and area show that category's share of the total, so the parts and the whole are read together. Line chart connects values in sequence to show how a measure changes over time, and it has no single whole for the values to divide. Scatter plot shows the relationship between two numeric variables, not how parts make up a total. Box plot summarizes the distribution of a numeric variable with its median, quartiles and outliers, which says nothing about shares of a whole.

  13. Which chart type is best for showing the proportion of a total over time?

    • A.Stacked area chart
    • B.Multiple line chart
    • C.Clustered bar chart
    • D.Waterfall chart
    Show answerHide answer

    Correct answer: Stacked area chart

    Stacked area chart is best for showing the proportion of a total over time: each category's band is layered on top of the one below it along a time axis, so the chart shows how much each part contributes to the total and how both change from period to period. Multiple line chart draws each category as its own line, which shows each series over time but not its share of a combined total. Clustered bar chart places bars for each category side by side, which compares their absolute values but does not build them into a total. Waterfall chart shows how a starting value is raised or lowered by a sequence of positive and negative changes, not how a total is divided among its parts over time.

  14. Which visualization technique is most effective for displaying data with multiple variables, where each variable contributes to the final outcome?

    • A.Cumulative frequency plot
    • B.Parallel coordinates plot
    • C.Pie chart
    • D.Bar chart
    Show answerHide answer

    Correct answer: Parallel coordinates plot

    Parallel coordinates plot is built for data with many variables: each variable gets its own vertical axis, the axes stand side by side, and every record is drawn as a line crossing all of them, so patterns across the full set of variables can be followed together. Cumulative frequency plot shows the running total of frequencies for a single variable, so it cannot display several variables at once. Pie chart divides one whole into slices for a single categorical variable. Bar chart compares one measure across categories, so it shows only one variable per set of bars.

  15. In data visualization, what is the primary purpose of a bubble chart?

    • A.To show three dimensions of a dataset
    • B.To show group counts of a categorical variable
    • C.To show percentage shares of a combined budget
    • D.To show friendship links of a social network
    Show answerHide answer

    Correct answer: To show three dimensions of a dataset

    To show three dimensions of a dataset is the purpose of a bubble chart: two variables set each bubble's horizontal and vertical position and a third sets the bubble's size, so three dimensions appear in one view. To show group counts of a categorical variable is the job of a bar chart, which uses bar heights for frequencies rather than position and size. To show percentage shares of a combined budget is the purpose of a pie chart; bubble sizes in a bubble chart are not parts of one whole. To show friendship links of a social network calls for a network diagram, which connects nodes with lines; a bubble chart has no connections between its bubbles.

  16. Which type of chart is best suited for displaying frequency distributions for categorical data?

    • A.Line chart
    • B.Bar chart
    • C.Frequency histogram
    • D.Frequency polygon
    Show answerHide answer

    Correct answer: Bar chart

    Bar chart is the right choice for frequency distributions of categorical data: each category gets its own separate bar whose height is the count or percentage, and the bars can be arranged in any useful order. Line chart connects points in sequence, which implies a continuous or time-ordered axis that unordered categories do not have. Frequency histogram groups a quantitative variable into adjacent numeric intervals, so its touching bars stand for ranges of numbers rather than separate categories. Frequency polygon joins the frequencies plotted at the midpoints of numeric classes with straight lines, which again assumes an ordered numeric scale that categories lack.

  17. What is the main advantage of using a stacked bar chart in data visualization?

    • A.It shows the correlation of paired measures.
    • B.It shows the composition of category totals.
    • C.It shows the spread of numeric observations.
    • D.It shows the location of regional customers.
    Show answerHide answer

    Correct answer: It shows the composition of category totals.

    It shows the composition of category totals. That is the main advantage of a stacked bar chart: each bar is one category's total, and its stacked segments are the individual values that make it up, so both the parts and their combined size can be read at once. "It shows the correlation of paired measures." describes a scatter plot, which plots two numeric variables against each other. "It shows the spread of numeric observations." describes a histogram or box plot, which summarize the distribution of one variable. "It shows the location of regional customers." describes a map, which places values in their geographic positions; a stacked bar chart has no geographic layout.

  18. Which chart type is most effective for comparing the composition of different groups at a single point in time?

    • A.Combination chart
    • B.Waterfall chart
    • C.Multiple line chart
    • D.Stacked bar chart
    Show answerHide answer

    Correct answer: Stacked bar chart

    Stacked bar chart is the most effective way to compare the composition of several groups at one point in time: each group gets one bar, and the bar is divided into segments for its parts, so both the totals and the makeup of each group can be compared side by side. Combination chart overlays two chart types, such as columns and a line, to show two different measures on shared or dual axes, not the parts that make up each group. Waterfall chart breaks a single total into a sequence of positive and negative steps, so it describes one whole rather than comparing several groups. Multiple line chart tracks several series across time, which is the wrong layout for a snapshot taken at a single point in time.

  19. What visualization tool is best for displaying data that has a geographical component?

    • A.Choropleth map
    • B.Sankey diagram
    • C.Scatter plot
    • D.Sunburst chart
    Show answerHide answer

    Correct answer: Choropleth map

    Choropleth map is the standard tool for data with a geographical component: it shades each region, such as a state or county, by the value of a measure, so regional patterns and differences stand out. Sankey diagram shows flows between stages or categories with bands whose widths are proportional to quantity, and it has no geographic layout. Scatter plot shows the relationship between two numeric variables on x and y axes rather than the location of regions. Sunburst chart shows hierarchical data as nested rings around a center, not as places on a map.

  20. In data visualization, what is a waterfall chart primarily used for?

    • A.Showing how high and low readings spread around a median
    • B.Showing how stores and products compare against a target
    • C.Showing how later gains and losses shift a starting value
    • D.Showing how site visitors and leads shrink along a funnel
    Show answerHide answer

    Correct answer: Showing how later gains and losses shift a starting value

    Showing how later gains and losses shift a starting value is what a waterfall chart is for: it begins with an initial value, adds or subtracts each positive or negative change as a floating bar, and ends with the resulting total, as when revenue is walked down to net income. Showing how high and low readings spread around a median describes a box plot, which summarizes a distribution rather than a sequence of changes. Showing how stores and products compare against a target describes a bullet chart or bar chart measured against a goal. Showing how site visitors and leads shrink along a funnel describes a funnel chart, which tracks how many items remain at each stage of a process rather than how changes add up to a total.

  21. Which type of chart is most effective for displaying data where the order of categories is important?

    • A.Clustered bar chart
    • B.Stacked area chart
    • C.Polar area chart
    • D.Ordered bar chart
    Show answerHide answer

    Correct answer: Ordered bar chart

    Ordered bar chart is most effective when the order of categories matters: the bars are arranged in a deliberate sequence, such as ascending or descending value or a natural order of ratings, so rank and sequence can be read directly. Clustered bar chart places several bars side by side within each group to compare subcategories, and its purpose is that comparison rather than a meaningful order of categories. Stacked area chart shows how parts of a total change over a continuous time axis, not a ranking of separate categories. Polar area chart arranges categories around a circle as wedges, which makes their order and relative sizes harder to read than bars on a common baseline.

  22. What is the primary benefit of using a funnel chart in data visualization?

    • A.It shows the ranking of category totals.
    • B.It shows the progression of process stages.
    • C.It shows the nesting of hierarchical levels.
    • D.It shows the distribution of numeric scores.
    Show answerHide answer

    Correct answer: It shows the progression of process stages.

    It shows the progression of process stages. That is the primary benefit of a funnel chart: each stage of a process, such as a sales pipeline or customer journey, is drawn as a band whose width shows how many items reach it, so drop-off between stages is easy to see. "It shows the ranking of category totals." describes a sorted bar chart, which ranks unrelated categories rather than following items through a process. "It shows the nesting of hierarchical levels." describes a tree map or sunburst chart, which display parent and child categories. "It shows the distribution of numeric scores." describes a histogram, which groups one numeric variable into bins instead of tracking stages.

  23. In data visualization, what is the primary use of a box-and-whisker plot?

    • A.Highlighting the average of a population
    • B.Summarizing the distribution of a dataset
    • C.Displaying the composition of a portfolio
    • D.Visualizing the geography of a territory
    Show answerHide answer

    Correct answer: Summarizing the distribution of a dataset

    Summarizing the distribution of a dataset is the primary use of a box-and-whisker plot: it shows the minimum, first quartile, median, third quartile and maximum, with outliers plotted separately, so the center, spread and skew of the data are visible at a glance. Highlighting the average of a population is a common misreading, because the line inside the box marks the median rather than the mean, and the plot's purpose is the whole spread of values. Displaying the composition of a portfolio is a part-to-whole question for a pie chart or stacked bar chart. Visualizing the geography of a territory calls for a map such as a choropleth map, since a box plot has no geographic layout.

  24. A retail operations team asks an analyst to build a single screen that consolidates daily sales, inventory turnover, and customer satisfaction so managers can monitor performance at a glance. What is the analyst being asked to build?

    • A.A wireframe
    • B.A dashboard
    • C.A worksheet
    • D.A histogram
    Show answerHide answer

    Correct answer: A dashboard

    A dashboard is a single visual screen that consolidates several metrics, such as sales, inventory turnover and satisfaction, so managers can monitor performance at a glance. A wireframe is only the layout mockup drawn before a dashboard is built, a worksheet is a spreadsheet tab of cells rather than a visual monitoring screen, and a histogram is one chart showing the distribution of a single variable.

  25. An analyst wants to highlight a metric such as monthly recurring revenue that the business has identified as critical to its strategic goals. Which term describes this type of measurable value used to gauge progress toward a business objective?

    • A.A quality management system (QMS)
    • B.A service level agreement (SLA)
    • C.A key performance indicator (KPI)
    • D.A critical success factor (CSF)
    Show answerHide answer

    Correct answer: A key performance indicator (KPI)

    A key performance indicator (KPI) is a measurable value that shows how effectively an organization is progressing toward a specific business objective, such as monthly recurring revenue. A critical success factor (CSF) is a condition that must go right for a goal to be met, which KPIs are then chosen to measure; a service level agreement (SLA) is a contract that fixes the level of service a provider must deliver; and a quality management system (QMS) is the set of processes an organization uses to control quality, not a single measured value.

  26. A finance director needs a report that refreshes automatically and lets viewers filter by region and drill into individual transactions while viewing it. Which report or dashboard type best meets this need?

    • A.A dynamic dashboard
    • B.A static dashboard
    • C.A recurring report
    • D.A compliance report
    Show answerHide answer

    Correct answer: A dynamic dashboard

    A dynamic dashboard connects to refreshing data and lets viewers filter by region and drill into individual transactions while they view it. A static dashboard is a fixed snapshot that cannot be filtered or drilled into once it is produced, a recurring report refreshes on a timetable but arrives as a fixed output with no interactivity, and a compliance report presents required figures in a fixed format for regulators or auditors rather than for exploration.

  27. An executive requests a visualization that shows a single current value, such as this quarter's customer satisfaction score, relative to a target range using a needle or arc display. Which chart type is most appropriate?

    • A.A bullet chart
    • B.A gauge chart
    • C.A bubble chart
    • D.A pie chart
    Show answerHide answer

    Correct answer: A gauge chart

    A gauge chart shows one current value against a target range using a needle or colored arc, which is exactly what the executive described. A bullet chart compares a single value with its targets using horizontal bars rather than a dial, a bubble chart plots three variables as circles positioned on two axes and sized by the third, and a pie chart shows how parts make up a whole as slices of one circle, so none of them shows one value on a needle-and-arc dial.

  28. An analyst exports raw sales records into a tool and creates a summary that groups revenue by region in rows and by product category in columns, with totals at the intersections. What feature is the analyst using?

    • A.A fact table
    • B.A lookup table
    • C.A bridge table
    • D.A pivot table
    Show answerHide answer

    Correct answer: A pivot table

    A pivot table regroups raw records by placing fields in rows, columns and values, producing cross-tabulated totals such as revenue by region and product category. A fact table stores the detailed measures of a star schema at their original grain rather than summarizing them for display, a lookup table maps codes to descriptive values, and a bridge table resolves many-to-many relationships between dimensions.

  29. A marketing manager occasionally needs a one-off answer to an unplanned question, such as which campaign drove sign-ups during a single unexpected traffic spike. Which report type best describes this request?

    • A.An annual report
    • B.An ad-hoc report
    • C.A canned report
    • D.A status report
    Show answerHide answer

    Correct answer: An ad-hoc report

    An ad-hoc report is created on demand to answer a specific, one-time question that is not part of any regular reporting schedule. An annual report is produced once a year on a fixed cycle, a canned report is a predefined standard report that runs the same way every time, and a status report tracks progress at regular intervals, so none of them fits a single unplanned request.

  30. A business intelligence team wants non-technical staff to build their own filtered views and simple charts without submitting requests to the analytics team. Which reporting approach are they enabling?

    • A.Standardized reporting
    • B.Operational reporting
    • C.Centralized reporting
    • D.Self-service reporting
    Show answerHide answer

    Correct answer: Self-service reporting

    Self-service reporting gives business users governed access to data and tools so they can build filtered views and simple charts themselves, without filing requests with the analytics team. Standardized reporting delivers the same predefined reports to everyone, operational reporting tracks day-to-day activity on a set schedule, and centralized reporting keeps report building with a central analytics or IT team, which is the bottleneck the team wants to remove.

  31. An analyst is choosing between a bar chart and a histogram for two different tasks. Which statement correctly describes when to use each?

    • A.Use a bar chart to compare values across distinct categories and a histogram to show the frequency distribution of a continuous variable in bins
    • B.Use a bar chart to display the spread of one numeric measure and a histogram to contrast the subtotals of separately labeled groups side by side
    • C.Use a bar chart to compare values across distinct categories and a histogram to contrast the subtotals of separately labeled groups side by side
    • D.Use a bar chart to display the spread of one numeric measure and a histogram to show the frequency distribution of a continuous variable in bins
    Show answerHide answer

    Correct answer: Use a bar chart to compare values across distinct categories and a histogram to show the frequency distribution of a continuous variable in bins

    Use a bar chart to compare values across distinct categories and a histogram to show the frequency distribution of a continuous variable in bins: bar charts compare separate groups and usually leave gaps between the bars, while a histogram groups a numeric variable into ranges whose bars touch. The advice to use a bar chart to display the spread of one numeric measure and a histogram to contrast the subtotals of separately labeled groups side by side reverses the two tools. The advice to use a bar chart to compare values across distinct categories and a histogram to contrast the subtotals of separately labeled groups side by side gets the bar chart right but misuses the histogram, and the advice to use a bar chart to display the spread of one numeric measure and a histogram to show the frequency distribution of a continuous variable in bins gets the histogram right but misuses the bar chart.

  32. A senior analyst reviews a colleague's quarterly revenue chart and notes that the vertical axis starts at 90 rather than 0, making a small increase look dramatic. According to reporting best practices, what is the main concern?

    • A.A raised baseline is harmless because readers can see the values
    • B.A truncated axis can corrupt stored data by rounding off figures
    • C.A raised baseline can hide growth that real values would reveal
    • D.A truncated axis can mislead viewers by overstating differences
    Show answerHide answer

    Correct answer: A truncated axis can mislead viewers by overstating differences

    A truncated axis can mislead viewers by overstating differences: starting the value axis at 90 instead of 0 makes a small increase look dramatic, which is why reporting guidance warns against it. The claim that a raised baseline is harmless because readers can see the values ignores that people judge the relative height of the bars, not the labels. The claim that a truncated axis can corrupt stored data by rounding off figures is wrong because the underlying values are untouched, and the claim that a raised baseline can hide growth that real values would reveal reverses the effect, since a raised baseline magnifies change rather than hiding it.

  33. An analyst is preparing a presentation that walks executives from a problem, through key findings, to a clear recommendation using a logical sequence of visuals and narration. What practice is the analyst applying?

    • A.Auditing a data set
    • B.Building a data model
    • C.Telling a data story
    • D.Sketching a data flow
    Show answerHide answer

    Correct answer: Telling a data story

    Telling a data story is the practice described: it arranges findings into a narrative with context, insight and a recommendation so executives can follow the analysis and act on it. Auditing a data set checks it for quality and compliance problems, building a data model designs the tables and relationships that store data, and sketching a data flow maps how data moves between systems, so none of them is a narrative that leads an audience to a recommendation.

  34. When designing a dashboard, an analyst places the most important summary KPIs in the top-left area and supporting detail lower on the page. What dashboard design principle is being followed?

    • A.Arranging content by visual hierarchy and reading order
    • B.Clustering related metrics by proximity and alignment
    • C.Maximizing the data-ink ratio with minimal chart clutter
    • D.Harmonizing the visuals with one consistent color scheme
    Show answerHide answer

    Correct answer: Arranging content by visual hierarchy and reading order

    Arranging content by visual hierarchy and reading order puts the highest-priority metrics where the eye lands first, the top left in left-to-right reading cultures, with supporting detail below. Clustering related metrics by proximity and alignment groups items that belong together but does not decide which item comes first, maximizing the data-ink ratio with minimal chart clutter removes decoration that carries no information, and harmonizing the visuals with one consistent color scheme keeps the styling coherent; none of these explains putting the summary KPIs at the top left.

  35. A compliance team requires a report delivered every month in an identical format showing required regulatory metrics, with no interactive elements. Which report type best fits this requirement?

    • A.A forecasting report
    • B.An analytical report
    • C.A recurring report
    • D.An ad-hoc report
    Show answerHide answer

    Correct answer: A recurring report

    A recurring report is generated on a fixed schedule in a consistent format, which suits a monthly regulatory requirement with no interactive elements. A forecasting report projects future values rather than restating required metrics each month, an analytical report digs into a question in depth and changes with each investigation, and an ad-hoc report answers a single unplanned question, so none of them fits a standing monthly compliance deliverable.

  36. An analyst must choose colors for a dashboard that will be viewed by an audience that includes colorblind users. Which design choice best supports accessibility?

    • A.Use colorblind-safe palettes and add labels or patterns so color is not the sole signal
    • B.Color bad values red and good ones green so the dashboard status is clear at a glance
    • C.Raise saturation of hues that are not distinct so each stands out more on the dashboard
    • D.Rely on a short legend or tooltips so colorblind viewers can match the hue to its group
    Show answerHide answer

    Correct answer: Use colorblind-safe palettes and add labels or patterns so color is not the sole signal

    The accessible choice is to use colorblind-safe palettes and add labels or patterns so color is not the sole signal, because direct labels and patterns let every viewer read the data even when two hues look alike. The advice to color bad values red and good ones green so the dashboard status is clear at a glance is the classic pitfall, because red and green are the hues most often confused. The advice to raise saturation of hues that are not distinct so each stands out more on the dashboard still leaves the meaning in hue alone, and the advice to rely on a short legend or tooltips so colorblind viewers can match the hue to its group still forces viewers to tell the colors apart.

  37. An analyst builds a dashboard that pulls from an operational database and needs to decide how often the underlying data should update so users see timely numbers without overloading the source system. Which design component is being decided?

    • A.The data retention period
    • B.The query timeout limit
    • C.The data refresh interval
    • D.The export row limit
    Show answerHide answer

    Correct answer: The data refresh interval

    The data refresh interval is the design decision described: it sets how often the dashboard pulls new data, balancing timely numbers against the load placed on the operational source. The data retention period sets how long records are kept before deletion, the query timeout limit sets how long a single query may run before it is stopped, and the export row limit caps how many rows a user can download, so none of them controls how current the displayed numbers are.

  38. Before building any visualization, an analyst meets with stakeholders to clarify what decisions the report must support and which metrics matter to them. Why is this step considered a reporting best practice?

    • A.It translates business requirements first, so the report answers the right questions for the audience
    • B.It settles data ownership first, so the report holds the stakeholders accountable for its errors
    • C.It lets stakeholders choose the chart styling for each panel, so the report looks the way they prefer
    • D.It documents each meeting for the audit team, so the report satisfies the record-keeping requirements
    Show answerHide answer

    Correct answer: It translates business requirements first, so the report answers the right questions for the audience

    The step is a best practice because it translates business requirements first, so the report answers the right questions for the audience and is built around the decisions and metrics that matter, which prevents rework. The idea that it settles data ownership first, so the report holds the stakeholders accountable for its errors, is wrong because the analyst still has to validate the data. The idea that it lets stakeholders choose the chart styling for each panel, so the report looks the way they prefer, confuses requirements with cosmetic preferences, and the idea that it documents each meeting for the audit team, so the report satisfies the record-keeping requirements, mistakes a side record for the purpose of the meeting.

Data Governance (33)

  1. What does the term "Data Governance" primarily refer to?

    • A.Checking the accuracy, completeness, validity, and timeliness of data
    • B.Designing the models, relationships, databases, and pipelines of data
    • C.Managing the availability, usability, integrity, and security of data
    • D.Interpreting the trends, patterns, outliers, and correlations of data
    Show answerHide answer

    Correct answer: Managing the availability, usability, integrity, and security of data

    "Managing the availability, usability, integrity, and security of data" is correct: data governance is the overall management of those four qualities through policies, standards, assigned roles such as owners and stewards, and decision rights across the organization. "Checking the accuracy, completeness, validity, and timeliness of data" is incorrect: measuring those dimensions is data quality assessment, one activity that governance oversees, not the meaning of governance itself. "Designing the models, relationships, databases, and pipelines of data" is incorrect: that is data architecture and engineering, which design how data is modeled, related, stored, and moved. "Interpreting the trends, patterns, outliers, and correlations of data" is incorrect: that is data analysis; governance sets the rules for how data is managed and used but does not analyze it.

  2. What is the primary purpose of a Master Data Management (MDM) system in a data environment?

    • A.To keep master data encrypted, logged, and restricted across the organization
    • B.To keep master data consistent, accurate, and uniform across the organization
    • C.To keep master data copied, archived, and recoverable across the organization
    • D.To keep master data cached, indexed, and denormalized across the organization
    Show answerHide answer

    Correct answer: To keep master data consistent, accurate, and uniform across the organization

    "To keep master data consistent, accurate, and uniform across the organization" is correct: master data management creates and maintains a single trusted version (a golden record) of core entities such as customers, products, and suppliers, and keeps that version synchronized across systems and departments. "To keep master data encrypted, logged, and restricted across the organization" is incorrect: encryption, audit logging, and access restrictions are security and identity and access management controls; MDM relies on them but exists to make master data consistent, not to secure it. "To keep master data copied, archived, and recoverable across the organization" is incorrect: copying, archiving, and recovery are backup and disaster recovery functions, which protect data against loss rather than reconciling it into one consistent version. "To keep master data cached, indexed, and denormalized across the organization" is incorrect: caching, indexing, and denormalizing are database performance techniques; they speed access but do nothing to resolve conflicting or duplicate master records.

  3. In the context of data environments, what is the primary role of a data steward?

    • A.Overseeing and managing the quality and lifecycle of data
    • B.Administering and securing the storage and backup of data
    • C.Defining and approving the classification and use of data
    • D.Building and maintaining the pipelines and models of data
    Show answerHide answer

    Correct answer: Overseeing and managing the quality and lifecycle of data

    "Overseeing and managing the quality and lifecycle of data" is correct: a data steward is responsible for the day-to-day oversight of data assets, including their definitions, their quality, and their care through each stage of the lifecycle, so the data stays accurate, consistent, and fit for use. "Administering and securing the storage and backup of data" is incorrect: that is the data custodian, usually an IT role, which runs the technical environment (storage, backup, and security controls) that holds the data. "Defining and approving the classification and use of data" is incorrect: that is the data owner, the accountable business leader who classifies the data and approves who may access and use it. "Building and maintaining the pipelines and models of data" is incorrect: that is the data engineer, who builds the pipelines and structures that move and store data; a steward oversees the data's quality rather than the plumbing.

  4. What does the term "Data Lifecycle Management" (DLM) primarily involve?

    • A.Tracing data from its origin to its destination
    • B.Handling data from its creation to its deletion
    • C.Transferring data from its source to its target
    • D.Encrypting data from its sender to its receiver
    Show answerHide answer

    Correct answer: Handling data from its creation to its deletion

    "Handling data from its creation to its deletion" is correct: Data Lifecycle Management governs data through every stage, including creation, storage, use, sharing, archiving, and finally secure deletion, usually guided by retention policies. "Tracing data from its origin to its destination" is incorrect: tracing where data came from and where it flows is data lineage, which documents movement and transformation rather than managing retention and disposal over the data's life. "Transferring data from its source to its target" is incorrect: moving data from a source to a target is data migration or integration, as in ETL; it is a single movement, not the management of data across its entire life. "Encrypting data from its sender to its receiver" is incorrect: encrypting data between a sender and a receiver protects data in transit; it is a security control, not lifecycle management.

  5. Which data governance framework component primarily focuses on ensuring data availability, usability, consistency, data integrity, and data security?

    • A.Data Architecture
    • B.Data Integration
    • C.Data Warehousing
    • D.Data Stewardship
    Show answerHide answer

    Correct answer: Data Stewardship

    Data Stewardship is the governance component focused on keeping data available, usable, consistent, accurate and secure in day-to-day practice: data stewards are accountable for the quality, definitions and proper use of the data in their domains and for applying governance policies to it. Data Architecture designs the structures, models and flows that hold and move data, the blueprint rather than the ongoing care of the data itself. Data Integration combines data from different sources into a unified view, a technical process rather than the governance role that oversees usability and integrity. Data Warehousing stores integrated historical data for reporting and analysis, which supports availability for analytics but does not govern consistency, integrity or security across the organization.

  6. What is the primary purpose of a Data Dictionary in the context of data governance?

    • A.To trace the lineage of data transformations
    • B.To schedule the loading of data warehouses
    • C.To document the definitions of data elements
    • D.To record the history of data modifications
    Show answerHide answer

    Correct answer: To document the definitions of data elements

    To document the definitions of data elements is the purpose of a data dictionary: it is a descriptive list of the organization's data elements, giving each one's name, meaning, data type, format, allowed values and relationships, so everyone uses the data the same way. To trace the lineage of data transformations is the job of data lineage tools, which follow data from source to destination through each change. To schedule the loading of data warehouses is handled by ETL and workflow tools, not by a reference document of definitions. To record the history of data modifications is the role of audit logs and change tracking, which capture who changed what and when.

  7. In data governance, what is the role of a Data Steward?

    • A.To oversee and maintain the quality of data
    • B.To store and protect the data in secure systems
    • C.To sponsor and fund the governance of data
    • D.To analyze and interpret trends in the data
    Show answerHide answer

    Correct answer: To oversee and maintain the quality of data

    To oversee and maintain the quality of data is the role of a data steward: stewards are responsible for the accuracy, consistency and proper definition of data in their area and for making sure it follows governance policies. To store and protect the data in secure systems describes a data custodian, the technical role that manages the systems, backups and security controls where data lives. To sponsor and fund the governance of data describes an executive sponsor or governance council, which sets direction and provides resources rather than managing data quality day to day. To analyze and interpret trends in the data describes a data analyst, who uses the data to answer business questions rather than being accountable for its quality.

  8. Which concept in data governance is primarily concerned with the accuracy and consistency of data over its lifecycle?

    • A.Data Lineage
    • B.Data Quality
    • C.Data Privacy
    • D.Data Catalog
    Show answerHide answer

    Correct answer: Data Quality

    Data Quality is the governance concept concerned with the accuracy and consistency of data throughout its lifecycle, together with dimensions such as completeness, timeliness and validity, so that data stays fit for its intended use. Data Lineage records where data came from and how it moved and changed between systems; it helps explain a value but does not itself keep the data accurate. Data Privacy governs how personal and sensitive data is collected, used and shared. Data Catalog is an inventory of an organization's data assets and their metadata that helps people find data, not a measure of whether the data is correct.

  9. What is the main objective of implementing Data Classification policies in an organization?

    • A.To archive data by inactivity and staleness
    • B.To sample data by location and demographics
    • C.To label data by sensitivity and importance
    • D.To deduplicate data by name and birthdate
    Show answerHide answer

    Correct answer: To label data by sensitivity and importance

    To label data by sensitivity and importance is the main objective of data classification policies: each data set is assigned a level such as public, internal, confidential or restricted according to how sensitive and critical it is, so the right security measures and access controls can be applied. To archive data by inactivity and staleness describes storage tiering and retention practice, which moves old data to cheaper storage rather than rating its sensitivity. To sample data by location and demographics describes stratified sampling for analysis, not a governance policy. To deduplicate data by name and birthdate is a data cleansing step that removes repeated records, which improves quality but assigns no sensitivity level.

  10. In the context of data governance, what does 'Data Lineage' refer to?

    • A.The process of converting raw data into meaningful information
    • B.The permissions that data users receive from request to approval
    • C.The process of checking raw data against validation rules
    • D.The lifecycle that data follows from creation to retirement
    Show answerHide answer

    Correct answer: The lifecycle that data follows from creation to retirement

    The lifecycle that data follows from creation to retirement is what data lineage documents: where the data originates, how it moves and changes between systems, and where it ends up, giving a traceable history of the data over its life. The process of converting raw data into meaningful information describes data processing, which carries out that conversion rather than recording where the data came from and what was done to it. The permissions that data users receive from request to approval describes access provisioning, which governs who may use data, not where the data has been. The process of checking raw data against validation rules describes data validation, a quality control applied to data rather than a record of its path.

  11. Which principle of data governance ensures that data is accessible to authorized users and applications when required?

    • A.Data Confidentiality
    • B.Data Availability
    • C.Data Validation
    • D.Data Encryption
    Show answerHide answer

    Correct answer: Data Availability

    Data Availability is the principle that data is accessible to authorized users and applications when it is required, supported by measures such as redundancy, backups and reliable systems so that business processes and decisions are not held up. Data Confidentiality is the related principle of keeping data away from people who are not authorized to see it; it restricts access rather than guaranteeing that access is there when needed. Data Validation checks that data meets defined rules for type, format and range before it is accepted, which concerns correctness rather than access. Data Encryption is a protective technique that encodes data so only holders of the decryption key can read it, which supports confidentiality rather than availability.

  12. What does the principle of 'Data Integrity' ensure in a data governance framework?

    • A.Data stays accurate and consistent across its lifecycle
    • B.Data stays encrypted and private across its systems
    • C.Data stays reachable and online across its networks
    • D.Data stays traceable and documented across its pipeline
    Show answerHide answer

    Correct answer: Data stays accurate and consistent across its lifecycle

    Data stays accurate and consistent across its lifecycle is what the principle of data integrity ensures: data remains complete, correct and trustworthy from creation through every change, transfer and use, protected from unintended or unauthorized alteration. Data stays encrypted and private across its systems describes confidentiality, which keeps data from unauthorized viewers but does not guarantee the data is correct. Data stays reachable and online across its networks describes availability, the principle that authorized users can get to data when they need it. Data stays traceable and documented across its pipeline describes data lineage, which records where data came from and how it moved rather than ensuring that its values are right.

  13. In data governance, what is the main purpose of implementing a Master Data Management (MDM) system?

    • A.To maintain a reference list of field definitions for database tables
    • B.To build a historical store of transactions for reporting
    • C.To provide a single source of truth for key business entities
    • D.To keep a raw archive of social media posts for later analysis
    Show answerHide answer

    Correct answer: To provide a single source of truth for key business entities

    To provide a single source of truth for key business entities is the purpose of master data management: it reconciles, links and synchronizes master data such as customers, products and suppliers across systems so the organization has one trusted, consistent record for each of them. To maintain a reference list of field definitions for database tables is the role of a data dictionary or metadata repository, which describes data rather than mastering business records. To build a historical store of transactions for reporting describes a data warehouse, which keeps transaction history for reporting and analysis. To keep a raw archive of social media posts for later analysis describes a data lake, which holds data in its original form without creating an agreed master record.

  14. Which element in data governance is primarily responsible for defining who can take what action, on what data, in what situations?

    • A.Data Lineage Records
    • B.Data Access Controls
    • C.Data Quality Metrics
    • D.Data Retention Rules
    Show answerHide answer

    Correct answer: Data Access Controls

    Data Access Controls define who, meaning which users or systems, can take what action, such as view, edit or delete, on what data and under what conditions, through mechanisms such as role-based permissions; they are central to data security and compliance. Data Lineage Records document where data came from and how it moved and changed between systems, which explains the data's history but grants or denies no action to anyone. Data Quality Metrics measure properties such as accuracy, completeness and timeliness, which says nothing about who is allowed to use the data. Data Retention Rules set how long data is kept and when it is disposed of, not who may act on it.

  15. What is the primary goal of implementing a Data Retention Policy in an organization?

    • A.To decide how much data is compressed before backups
    • B.To decide how long data is retained before disposal
    • C.To decide how well data is encrypted before transfer
    • D.To decide how closely data is checked before loading
    Show answerHide answer

    Correct answer: To decide how long data is retained before disposal

    To decide how long data is retained before disposal is the primary goal of a data retention policy: it sets retention periods for different types of data, based on legal, regulatory and business needs, and states when and how the data is then disposed of. To decide how much data is compressed before backups is a storage optimization choice; lower storage cost can be a side benefit of retention rules, but it is not their purpose. To decide how well data is encrypted before transfer belongs to data security policy. To decide how closely data is checked before loading is a data validation and quality concern in the ETL process, not a question of how long records are kept.

  16. In data governance, what does the term 'Data Sovereignty' refer to?

    • A.The business unit in which data is managed
    • B.The retention period in which data is stored
    • C.The geographic location in which data is kept
    • D.The security tier in which data is classified
    Show answerHide answer

    Correct answer: The geographic location in which data is kept

    The geographic location in which data is kept is what data sovereignty turns on, because data is subject to the laws and governance requirements of the country where it physically resides; organizations operating in several jurisdictions must therefore track where each data set is stored. The business unit in which data is managed describes data ownership and stewardship within an organization, not the national laws that apply. The retention period in which data is stored describes a data retention policy, which concerns how long data is kept rather than where. The security tier in which data is classified describes data classification by sensitivity, which is unrelated to the country hosting the data.

  17. Which process in data governance is responsible for identifying, assessing, and mitigating risks related to data?

    • A.Data Risk Management
    • B.Data Breach Response
    • C.Data Quality Assessment
    • D.Data Compliance Auditing
    Show answerHide answer

    Correct answer: Data Risk Management

    Data Risk Management is the governance process that identifies, assesses and mitigates risks related to data, such as breaches, loss, misuse and compliance failures, by ranking those risks and putting controls in place before harm occurs. Data Breach Response is the set of actions taken after a breach has happened to contain it, notify affected parties and recover, so it reacts to one realized risk rather than managing data risks in advance. Data Quality Assessment measures how accurate, complete and consistent data is, which can reveal problems but is not a risk process. Data Compliance Auditing checks whether data practices meet laws and internal policies at a point in time; its findings feed risk management but do not replace it.

  18. What is the primary purpose of conducting Data Audits in an organization?

    • A.To check that data practices follow governance policies
    • B.To check that archived data backups restore successfully
    • C.To check that financial data matches accounting journals
    • D.To check that data pipelines satisfy performance targets
    Show answerHide answer

    Correct answer: To check that data practices follow governance policies

    To check that data practices follow governance policies is the primary purpose of a data audit: it evaluates whether data is collected, stored, used and protected in line with the organization's governance policies and with external regulations. To check that archived data backups restore successfully describes a backup or disaster recovery test, which confirms recoverability rather than policy compliance. To check that financial data matches accounting journals describes a financial audit or account reconciliation, which concerns the accuracy of financial statements. To check that data pipelines satisfy performance targets describes performance monitoring of data processing systems, not an audit of governance compliance.

  19. Which component of data governance is primarily concerned with defining policies and procedures for sensitive data handling?

    • A.Data Masking
    • B.Data Lineage
    • C.Data Privacy
    • D.Data Quality
    Show answerHide answer

    Correct answer: Data Privacy

    Data Privacy is the governance component that defines the policies and procedures for handling sensitive and personal data: what may be collected, how it may be used and shared, how long it is kept, and how individuals' rights and legal requirements are met. Data Masking is one technique for protecting sensitive values by hiding or substituting them; it carries out privacy rules rather than defining them. Data Lineage documents where data comes from and how it moves between systems. Data Quality concerns the accuracy, completeness and consistency of data, not the rules for handling sensitive information.

  20. What is the main objective of Change Management in the context of data governance?

    • A.To ensure changes to data systems go smoothly and efficiently
    • B.To copy changed data to target tables quickly and incrementally
    • C.To grant new employees access to data promptly and securely
    • D.To restore data to earlier versions safely and reliably
    Show answerHide answer

    Correct answer: To ensure changes to data systems go smoothly and efficiently

    To ensure changes to data systems go smoothly and efficiently is the main objective of change management in data governance: proposed changes to systems, processes or technologies are requested, assessed, approved, tested and communicated so they are put in place with minimal disruption and risk. To copy changed data to target tables quickly and incrementally describes change data capture, an integration technique whose similar name is easy to confuse with change management. To grant new employees access to data promptly and securely describes access provisioning, part of access management. To restore data to earlier versions safely and reliably describes backup and recovery, which deals with loss rather than with controlling planned changes.

  21. In data governance, what role does a Data Custodian play?

    • A.Establishing the meaning and quality of data
    • B.Approving the release and sharing of data
    • C.Interpreting the trends and patterns of data
    • D.Protecting the security and privacy of data
    Show answerHide answer

    Correct answer: Protecting the security and privacy of data

    Protecting the security and privacy of data is the role of a data custodian: the custodian is the technical role that stores, maintains and safeguards data assets, applying controls such as access permissions, encryption and backups to guard against unauthorized access and breaches. Establishing the meaning and quality of data describes a data steward, who is responsible for definitions and data quality rather than technical custody. Approving the release and sharing of data describes a data owner, the accountable business role that decides who may use the data. Interpreting the trends and patterns of data describes a data analyst, who uses data rather than safeguarding it.

  22. Which element of data governance ensures that external data sources are reliable and appropriate for use in decision-making?

    • A.Data Source Integration
    • B.Data Source Inventory
    • C.Data Source Validation
    • D.Data Source Mapping
    Show answerHide answer

    Correct answer: Data Source Validation

    Data Source Validation is the governance element that checks external data sources for credibility, accuracy, completeness and relevance before their data is used in decision-making, so only reliable and appropriate sources feed analysis. Data Source Integration combines data from several sources into a unified view, which assumes the sources are already trusted rather than vetting them. Data Source Inventory lists the available sources and their owners and metadata so people can find them, without judging whether they are reliable. Data Source Mapping matches the fields of a source to the fields of a target system, a design step for moving data rather than a test of the source's quality.

  23. What is the purpose of implementing Data Lifecycle Management (DLM) in an organization?

    • A.To trace data from its original source to its current location
    • B.To move data from its legacy systems to its successor platform
    • C.To copy data from its production server to its standby replica
    • D.To manage data from its initial creation to its final disposal
    Show answerHide answer

    Correct answer: To manage data from its initial creation to its final disposal

    To manage data from its initial creation to its final disposal is correct: Data Lifecycle Management (DLM) sets the policies that govern data at each stage of its life, from creation and storage through use, sharing and archiving to deletion, so it is handled efficiently and in line with retention and compliance rules at every step. To trace data from its original source to its current location is incorrect: that describes data lineage, which documents where data came from and how it moved and changed, rather than managing each stage of its life. To move data from its legacy systems to its successor platform is incorrect: that is data migration, a project that transfers data from one system to another. To copy data from its production server to its standby replica is incorrect: that is replication for backup and disaster recovery, a protection task rather than the purpose of lifecycle management.

  24. Which practice in data governance involves defining clear roles and responsibilities for data-related activities?

    • A.Data policy creation
    • B.Data role definition
    • C.Data retention rules
    • D.Data privacy reviews
    Show answerHide answer

    Correct answer: Data role definition

    Data role definition is correct: this governance practice spells out who is accountable for what, for example data owners, data stewards and data custodians, so that every data-related activity has a clearly responsible person or team. Data policy creation is incorrect: policies state the rules for how data must be handled, but writing them does not assign who is responsible for each activity. Setting data retention rules is incorrect: retention rules decide how long data is kept before it is archived or deleted, not who is responsible for it. Running data privacy reviews is incorrect: privacy reviews assess how personal data is collected and used in order to find privacy risks; they do not define roles and responsibilities.

  25. In the context of data governance, what is the main focus of a Data Compliance Officer?

    • A.Ensuring data use follows laws and regulations
    • B.Setting shared data quality rules and definitions
    • C.Running data storage servers and nightly backups
    • D.Designing data models and warehouse table layouts
    Show answerHide answer

    Correct answer: Ensuring data use follows laws and regulations

    Ensuring data use follows laws and regulations is correct: a Data Compliance Officer makes sure the way the organization collects, stores, uses and shares data meets the legal and regulatory requirements that apply to it, such as privacy and industry regulations, and oversees audits and fixes when gaps are found. Setting shared data quality rules and definitions is incorrect: that is the work of data stewards, who maintain business definitions and quality standards. Running data storage servers and nightly backups is incorrect: that is the technical job of data custodians or database administrators. Designing data models and warehouse table layouts is incorrect: that is the work of a data architect or data modeler.

  26. A data analyst profiles a customer table and finds that 4,000 of 50,000 rows are missing an email address. The analyst is asked to report which data quality dimension this gap most directly measures. Which dimension is it?

    • A.Availability
    • B.Traceability
    • C.Validity
    • D.Completeness
    Show answerHide answer

    Correct answer: Completeness

    Completeness is the dimension measured here, because it describes whether all required values are present and is usually scored as the share of non-null values; 4,000 missing emails out of 50,000 rows is a 92 percent completeness rate for that field. Validity asks whether the values that are present follow a defined format or rule, availability asks whether authorized users can reach the data when they need it, and traceability asks whether the origin and changes of a value can be followed, so none of them directly quantifies missing values.

  27. During a data governance review, a customer's birth year is stored as 1990 in the CRM but as 1992 in the billing system for the same person. Which data quality dimension does this conflict most directly violate?

    • A.Accuracy
    • B.Availability
    • C.Consistency
    • D.Traceability
    Show answerHide answer

    Correct answer: Consistency

    Consistency is the dimension violated, because it requires the same data value to match across every system and record where it appears, and two systems holding different birth years for one customer is a textbook consistency failure. Accuracy concerns whether a value matches the real-world fact; one of the two years may well be inaccurate, but the conflict between systems is what the review found. Availability concerns whether users can reach the data when needed, and traceability concerns whether the origin and changes of a value can be followed, so neither describes two systems that disagree.

  28. A marketing dataset contains the columns full name, home address, Social Security number, and email address. A data analyst must flag which fields qualify as personally identifiable information (PII) for governance purposes. Which statement best describes PII?

    • A.Only identifiers that a government agency issues such as a Social Security or passport number
    • B.Any detail that can identify a specific individual directly or once linked with other sources
    • C.Only details marked as confidential in a company database, such as a salary or a staff number
    • D.Any field stored together with other customer data, whether or not it can point to a person
    Show answerHide answer

    Correct answer: Any detail that can identify a specific individual directly or once linked with other sources

    PII is any detail that can identify a specific individual directly or once linked with other sources, so name, home address, Social Security number and email can all qualify. Defining it as only identifiers that a government agency issues such as a Social Security or passport number is too narrow, because address and email also count when they can single someone out. Defining it as only details marked as confidential in a company database, such as a salary or a staff number, confuses a confidentiality label with identifiability, and defining it as any field stored together with other customer data, whether or not it can point to a person, is too broad, because a field that cannot identify anyone is not PII.

  29. A healthcare analytics team wants to share a patient dataset with an external research vendor but must remove the ability to re-identify individuals permanently, with no key kept to reverse the process. Which de-identification technique fits this requirement?

    • A.Pseudonymization
    • B.Field encryption
    • C.Anonymization
    • D.Tokenization
    Show answerHide answer

    Correct answer: Anonymization

    Anonymization fits, because it irreversibly removes identifying information so the data can no longer be traced back to an individual, and nothing is kept that could reverse it. Pseudonymization replaces identifiers with substitutes but keeps a separate mapping that allows re-identification, field encryption protects selected columns but can be reversed by anyone holding the decryption key, and tokenization swaps values for tokens that a secure vault can map back, so all three are reversible and fail the permanent-removal requirement.

  30. An organization processes the personal data of customers located in the European Union. Under the GDPR principle of data minimization, what should the organization do?

    • A.Encrypt and store the personal data so any extended collection becomes acceptable
    • B.Tell customers what the personal data is for and share it only with their approval
    • C.Collect and keep only the personal data needed for the stated processing purpose
    • D.Ensure the personal data stays accurate and current by correcting any errors found
    Show answerHide answer

    Correct answer: Collect and keep only the personal data needed for the stated processing purpose

    Data minimization means the organization should collect and keep only the personal data needed for the stated processing purpose, limiting processing to what is adequate, relevant and limited to that purpose. The idea to encrypt and store the personal data so any extended collection becomes acceptable treats a security safeguard as permission to over-collect, which minimization does not allow. The idea to tell customers what the personal data is for and share it only with their approval describes the separate transparency and consent duties, and the idea to ensure the personal data stays accurate and current by correcting any errors found is the separate accuracy principle, not a limit on how much data is collected.

  31. A new analyst asks how data governance differs from data management on their team. Which statement best captures the distinction?

    • A.Governance defines the policies, roles, and decision rights for data, while management is the day-to-day work of collecting, storing, and processing it
    • B.Governance runs the pipelines, servers, and overnight loads for data, while management sets the standards, rules, and user permissions that apply to it
    • C.Governance defines the policies, roles, and decision rights for data, while management sets the standards, rules, and user permissions that apply to it
    • D.Governance runs the pipelines, servers, and overnight loads for data, while management is the day-to-day work of collecting, storing, and processing it
    Show answerHide answer

    Correct answer: Governance defines the policies, roles, and decision rights for data, while management is the day-to-day work of collecting, storing, and processing it

    Governance defines the policies, roles, and decision rights for data, while management is the day-to-day work of collecting, storing, and processing it; governance is the oversight layer and management is the execution layer. The version in which governance runs the pipelines, servers, and overnight loads for data, while management sets the standards, rules, and user permissions that apply to it, inverts the two. The version in which governance defines the policies, roles, and decision rights for data, while management sets the standards, rules, and user permissions that apply to it, gives both of them the policy role and leaves no one doing the work, and the version in which governance runs the pipelines, servers, and overnight loads for data, while management is the day-to-day work of collecting, storing, and processing it, gives both the execution role and leaves no one setting policy.

  32. Before loading a web sign-up file into the customer database, an analyst configures a rule that rejects any record whose email field does not match a valid email pattern and whose age field is not a number between 0 and 120. This screening of incoming values against defined rules is best described as which activity?

    • A.Data validation
    • B.Data enrichment
    • C.Data governance
    • D.Data cleansing
    Show answerHide answer

    Correct answer: Data validation

    This is data validation, because it checks incoming values against defined rules, formats and ranges, such as a valid email pattern or an age between 0 and 120, and rejects records that fail before they enter the system. Data cleansing corrects or removes bad values after they have been found, data enrichment adds attributes from other sources, and data governance sets the policies and responsibilities for data as a whole rather than checking each incoming record.

  33. A governance team maps the stages that a dataset moves through, from initial creation, to storage, active use, archival, and finally secure destruction. What is this end-to-end sequence of stages called?

    • A.The data lineage
    • B.The data lifecycle
    • C.The data inventory
    • D.The data hierarchy
    Show answerHide answer

    Correct answer: The data lifecycle

    This end-to-end sequence is the data lifecycle, the stages data passes through from creation and storage, through active use, to archival and secure destruction, which shapes governance decisions such as retention and disposal. The data lineage traces where specific data came from and how it was transformed as it moved between systems, the data inventory is a catalog of the datasets an organization holds, and the data hierarchy describes how data is organized from fields into records, files and databases, so none of them names the create-to-destroy stages.

References

  1. 1.CompTIA. “CompTIA Data+ Certification V2 (DA0-002) - Official Exam Details.” CompTIA, accessed Nov. 9, 2026. ↑
  2. 2.CompTIA. “CompTIA Data+ Certification V1 (DA0-001) - Retiring Version.” CompTIA. ↑
  3. 3.CompTIA. “CompTIA Data+ Certification - Overview.” CompTIA. ↑
  4. 4.CompTIA. “CompTIA Data+ Frequently Asked Questions.” CompTIA, accessed Nov. 9, 2026. ↑
  5. 5.CompTIA. “CompTIA Continuing Education Renewal Fees and CEU Requirements.” CompTIA. ↑
  6. 6.Career Employer. “CompTIA Data+ practice-test performance data.” careeremployer.com, updated daily, CC BY 4.0. ↑
Career Employer

Career Employer is the ultimate resource to help you get started working the job of your dreams. We cover topics from general career information, career searching, exam preparation with free study materials, career interviewing, and becoming successful in your career of choice.

Follow Us:

All Posts

Career Employer’s Editorial Process

Here at Career Employer, we focus a lot on providing factually accurate information that is always up to date. We strive to provide correct information using strict editorial processes, article editing, and fact-checking for all of the information found on our website. We only utilize trustworthy and relevant resources. To find out more, make sure to read our full editorial process page here.