I85PREP
⌘ K
CDSP (DSP-110) study materials
CDSP (DSP-110) study resources

CDSP (DSP-110) practice questions

One question at a time. Build your CDSP (DSP-110) understanding.

First 50 of 200 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1Defining business problems

A team is drafting a problem statement for a data science project. Which element most clearly distinguishes a well-framed business problem from a vague aspiration?

  1. A request for dashboards covering every department
  2. A named decision or action the analysis will inform Correct answer.
  3. A general desire to become more data-driven
  4. A list of tools the team prefers to use

Explanation

A well-framed problem ties the work to a specific decision or action that someone will take once evidence exists. Without a decision to inform, the analysis has no success criterion and cannot be scoped or evaluated. Tool choices and general aspirations describe means or moods, not problems.

Question 2Defining business problems

A stakeholder asks the team to "predict customer churn." Before scoping, what is the most useful clarifying question for defining the business problem?

  1. Which cloud region should host the model?
  2. Which machine learning library should we standardize on?
  3. What retention action would change based on the prediction? Correct answer.
  4. How many rows are in the customer table?

Explanation

The business value of a churn model depends on what intervention the prediction triggers and for whom. Asking about the retention action reveals the decision, the timing, and the target population. Library, row count, and hosting questions are technical or implementation concerns that come after the problem is framed.

Question 3Defining business problems

Which statement best describes the role of a measurable outcome in defining a business problem?

  1. It allows the team to judge whether the solution succeeded. Correct answer.
  2. It replaces the need to identify stakeholders.
  3. It specifies the technology stack for delivery.
  4. It guarantees the model will generalize to new data.

Explanation

A measurable outcome gives the team an objective test of whether the intervention worked, which is essential for evaluating a solution after deployment. It does not substitute for stakeholder identification, ensure generalization, or dictate tooling. Defining the metric early also forces clarity about the decision being supported.

Question 4Defining business problems

During problem definition, a team lists assumptions such as "customers who call support twice are at risk." Why should these assumptions be recorded explicitly?

  1. They replace the need for a baseline metric.
  2. They must be removed before any analysis begins.
  3. They can be tested and can reveal where the framing may be wrong. Correct answer.
  4. They become the final deliverable for stakeholders.

Explanation

Recording assumptions makes them testable and exposes where the problem framing rests on unverified beliefs. If an assumption fails, the team can revise the problem before investing in modeling. Assumptions are not automatically removed, are not deliverables, and do not substitute for baselines.

Question 5Defining business problems

A retail manager wants to reduce stockouts but has not specified which products or stores matter. Which reframing best turns this into an actionable data science problem?

  1. Forecast weekly demand per store-SKU pair to prioritize replenishment for high-margin items Correct answer.
  2. Hire additional warehouse staff to handle inventory
  3. Build a recommendation engine for online shoppers
  4. Collect all transaction data into a data lake

Explanation

The reframing names a unit of analysis, a prediction target, and a prioritization rule tied to business value. This makes the problem scoped, measurable, and connected to a replenishment decision. The other options are unrelated initiatives or operational actions that do not define an analytical problem.

Question 6Defining business problems

A team is deciding whether a problem is better solved with a predictive model or a simple business rule. Which consideration is most relevant at the problem-definition stage?

  1. Whether the team has used the chosen algorithm before
  2. Whether the decision varies enough across cases that a fixed rule would perform poorly Correct answer.
  3. Whether the problem involves any numeric data at all
  4. Whether the stakeholder prefers visual dashboards

Explanation

A model adds value when the right action differs across cases in ways a static rule cannot capture. If a simple rule already achieves the objective, modeling adds cost and complexity without benefit. Prior algorithm experience, data type, and visualization preferences are secondary to this fit question.

Question 7Defining business problems

A stakeholder defines success as "increase engagement." Which revision makes this a usable success metric for a data science project?

  1. Improve the app experience for all users
  2. Increase engagement substantially within the next quarter
  3. Make engagement a key performance indicator
  4. Raise the seven-day return rate of new mobile users from 18% to 24% by Q3 Correct answer.

Explanation

A usable metric specifies the population, the measure, the baseline, the target, and the timeframe. The revised statement includes all of these, making success verifiable. The other options restate the goal with adverbs or labels but provide no baseline or target.

Question 8Defining business problems

Which pair of elements is most essential when translating a business problem into a data science framing?

  1. Team size and sprint length
  2. Target variable and unit of analysis Correct answer.
  3. Programming language and IDE
  4. Cloud vendor and storage format

Explanation

Translating a business problem requires deciding what is being predicted or estimated and at what level of granularity. Those two choices determine the dataset, the validation design, and how results map back to decisions. Tooling, infrastructure, and process details are downstream choices.

Question 9Defining business problems

A marketing team wants to know which customers to target with a limited budget. The data scientist proposes predicting the probability that each customer responds to an offer. Which problem framing does this represent?

  1. Unsupervised anomaly detection
  2. Descriptive segmentation of the customer base
  3. Supervised prediction of an individual outcome Correct answer.
  4. Causal inference on the effect of the offer

Explanation

Predicting a response probability for each customer from historical labeled outcomes is a supervised prediction task. Segmentation would group customers without a labeled target, causal inference would estimate the offer's effect, and anomaly detection would flag unusual records. The targeting decision maps directly to per-customer predicted probabilities.

Question 10Defining business problems

When a business problem involves choosing between two costly strategies, why might a causal question be more appropriate than a predictive one?

  1. Predictive models cannot use historical data.
  2. Causal questions always require less data than predictive ones.
  3. Causal questions avoid the need for a success metric.
  4. The decision depends on the effect of an intervention, not just on correlation with an outcome. Correct answer.

Explanation

Choosing a strategy requires knowing what happens if you intervene, which is a causal question. A predictive model may exploit correlations that do not persist under intervention. Causal framing is not about data volume, does not preclude historical data, and still needs success metrics.

Question 11Defining business problems

A team is scoping a project and a stakeholder keeps adding new questions. Which practice best keeps the problem definition stable?

  1. Let the data scientist decide the scope alone
  2. Accept every new question to keep the stakeholder engaged
  3. Freeze all requirements before any data is examined
  4. Document the primary decision and explicitly park secondary questions for later Correct answer.

Explanation

Anchoring the scope to a primary decision lets the team defer secondary questions without losing stakeholder goodwill. Parking them preserves a record for future phases. Accepting everything causes scope creep, freezing requirements ignores what data exploration reveals, and unilateral scoping excludes stakeholders.

Question 12Defining business problems

Which situation most strongly signals that a proposed data science project lacks a genuine business problem?

  1. The target metric has not been measured before
  2. The stakeholder requests an interpretable model
  3. The data required is spread across several source systems
  4. No one can name a decision that would change based on the results Correct answer.

Explanation

If no decision would change, the analysis cannot create business value regardless of technical quality. Messy data, new metrics, and interpretability requests are common challenges that can be managed within a real problem. The absence of a decision is the fundamental gap.

Question 13Defining business problems

A hospital wants to reduce missed appointments. Which formulation best connects the business problem to a data science deliverable?

  1. Predict each appointment's no-show probability to guide reminder intensity Correct answer.
  2. Survey patients about their satisfaction with the clinic
  3. Increase the number of appointment slots available
  4. Purchase a new scheduling software system

Explanation

This formulation names a prediction target, a per-appointment unit of analysis, and a downstream action that uses the prediction. It is therefore scoped and testable. Software purchase, surveys, and capacity changes are operational or research activities without a defined analytical target.

Question 14Defining business problems

Which term describes the gap between the current state of a business process and a desired future state that a project aims to close?

  1. Feature store
  2. Data pipeline
  3. Problem statement Correct answer.
  4. Model card

Explanation

A problem statement articulates the gap between current and desired performance and frames why it matters. Data pipelines, model cards, and feature stores are technical artifacts that support analysis but do not define the business gap. A clear problem statement is the foundation for scoping and metrics.

Question 15Defining business problems

A stakeholder insists the model must be "accurate." Which response best advances problem definition?

  1. Agree and target the highest possible overall accuracy
  2. Switch to a simpler algorithm to guarantee accuracy
  3. Ask which errors are most costly and how the output will be used Correct answer.
  4. Collect more data until accuracy stops improving

Explanation

Accuracy is only meaningful relative to the costs of different errors and the way predictions drive decisions. Asking about error costs and usage converts a vague quality demand into a decision-relevant objective. Chasing overall accuracy can be harmful when classes are imbalanced or error costs are asymmetric.

Question 16Defining business problems

Which of the following best characterizes a "proxy metric" chosen during problem definition?

  1. An indirect measure used when the true outcome is hard or slow to observe Correct answer.
  2. A metric that replaces the need for stakeholder input
  3. A technical benchmark used only for model tuning
  4. A metric that is guaranteed to equal the true business outcome

Explanation

A proxy metric stands in for a true outcome that is difficult, expensive, or delayed to measure, such as using clicks to approximate purchase intent. Proxies are useful but must be validated against the outcome they represent. They are not guaranteed equivalents, do not replace stakeholders, and are distinct from tuning benchmarks.

Question 17Defining business problems

A team discovers that the business problem as stated would require data the organization cannot legally use. What is the most appropriate next step in problem definition?

  1. Proceed and hope the legal issue is resolved later
  2. Reframe the problem around data that is legally available and still serves the decision Correct answer.
  3. Use the restricted data but avoid documenting its source
  4. Abandon data science entirely for this domain

Explanation

Legal and ethical constraints are part of the problem definition, so the team should reframe around permissible data while preserving the decision the work must support. Proceeding on hope or hiding provenance creates serious risk. Abandoning the domain entirely is an overreaction when a compliant reframing may exist.

Question 18Defining business problems

Which question best tests whether a proposed project has an identifiable stakeholder decision owner?

  1. What accuracy threshold will the model achieve?
  2. Who will act differently depending on the model's output? Correct answer.
  3. How many features will the model require?
  4. Which team will maintain the model after launch?

Explanation

The decision owner is the person or role whose choices change based on the results. Asking who will act differently identifies that owner and tests whether the project is connected to real decisions. Maintenance ownership, feature count, and accuracy thresholds are important but do not establish decision ownership.

Question 19Defining business problems

A team is comparing two framings of the same issue: predicting which machines will fail versus estimating the effect of a maintenance schedule on failure rates. Which distinction does this comparison illustrate?

  1. Classification versus regression
  2. Clustering versus dimensionality reduction
  3. Prediction versus causal estimation Correct answer.
  4. Batch versus streaming processing

Explanation

Predicting which machines fail is a predictive task using observed associations, while estimating how a schedule changes failure rates is a causal question about an intervention. The distinction is about the type of question, not the algorithm family or data-processing mode. Choosing correctly determines the study design and validation approach.

Question 20Defining business problems

What is the primary purpose of writing a problem statement before collecting data?

  1. To satisfy documentation requirements for auditors
  2. To align stakeholders on the objective, scope, and success criteria Correct answer.
  3. To demonstrate technical sophistication to leadership
  4. To lock in the final model architecture

Explanation

A problem statement aligns everyone on what is being solved, what is in and out of scope, and how success will be judged. This alignment reduces rework and keeps the team focused. It is not meant to fix model architecture, serve auditors, or showcase technical skill.

Question 21Defining business problems

A stakeholder says the goal is to "reduce fraud losses." Which additional piece of information is most critical for defining a solvable data science problem?

  1. The color scheme of the fraud dashboard
  2. The definition of a fraudulent transaction and the action taken on a flagged case Correct answer.
  3. The number of engineers on the team
  4. The vendor of the payment processing system

Explanation

Fraud is a label that must be operationally defined, and the response to a flag determines the costs of false positives and false negatives. Without both, the team cannot build a meaningful target or evaluate performance. Dashboard design, staffing, and vendor choice are secondary to this definitional work.

Question 22Defining business problems

Which of the following is the clearest example of an output that is actionable for a business decision?

  1. A ranked list of accounts with estimated revenue potential and a suggested contact order Correct answer.
  2. A 40-page report on the history of the industry
  3. A correlation matrix of all available variables
  4. A list of every data quality issue found

Explanation

An actionable output tells the decision maker what to do next, in this case whom to contact and in what order. It links analysis to action. Historical reports, raw correlation matrices, and data quality logs may inform work but do not by themselves prescribe a decision.

Question 23Defining business problems

A team defines a project around "improving delivery times." Which change most improves the problem definition?

  1. Adding more delivery vehicles to the fleet
  2. Specifying which customer segments, time window, and baseline delivery time are targeted Correct answer.
  3. Renaming the project with a more modern title
  4. Requesting a larger budget for analytics

Explanation

Specifying the population, the metric, and the current baseline turns a broad goal into something measurable and scoped. Fleet changes and budget requests are actions or resources, not problem definitions. A new title does not change substance.

Question 24Defining business problems

In problem definition, what does establishing a baseline accomplish?

  1. It determines which algorithm must be used.
  2. It proves the proposed model will succeed.
  3. It eliminates the need for a validation dataset.
  4. It provides a reference point against which improvement can be measured. Correct answer.

Explanation

A baseline records current performance so that any change can be attributed and quantified. Without it, claims of improvement are unverifiable. A baseline does not guarantee model success, replace validation data, or dictate algorithm choice.

Question 25Defining business problems

A data scientist is asked to "find insights in the data" with no further direction. Which response best reflects sound problem-definition practice?

  1. Ask for a larger dataset before doing anything
  2. Work with the requester to identify a decision and a question worth answering Correct answer.
  3. Begin exploratory analysis immediately and report whatever is found
  4. Refuse the request because it is not a real project

Explanation

Open-ended insight requests become productive when tied to a decision and a question, which gives exploration a purpose and a stopping rule. Starting without direction often yields unfocused results, refusing outright is unhelpful, and requesting more data does not address the missing objective.

Question 26Defining business problems

A business unit asks a data science team to build a model that identifies customers likely to churn. Which element is missing if the request contains no statement of what action will be taken based on the model output?

  1. A defined decision or intervention that the prediction will trigger Correct answer.
  2. An estimate of the model's expected accuracy before any data is collected
  3. A complete data dictionary for all available customer attributes
  4. A clear description of the modeling algorithm to be used

Explanation

A data science problem is only actionable when the output connects to a decision or intervention. Without knowing what will be done differently for predicted churners, the team cannot choose an appropriate target, threshold, or evaluation metric. The algorithm and data details are implementation choices that follow from the decision context.

Question 27Defining business problems

A retail executive states the goal as 'improve customer experience.' Which reformulation best converts this into a testable business problem for a data science project?

  1. Collect more customer feedback data across all channels to understand experience drivers
  2. Increase the average customer satisfaction score from 3.8 to 4.2 within two quarters by reducing checkout wait times Correct answer.
  3. Build a machine learning model that predicts customer sentiment from social media posts
  4. Hire additional customer service staff to handle complaints more quickly

Explanation

A testable business problem specifies a measurable outcome, a target change, a time frame, and a lever the organization can influence. This option converts a vague aspiration into a bounded problem with a baseline, goal, and candidate intervention. The other options describe activities or tools rather than a defined problem with a success condition.

Question 28Defining business problems

A logistics company wants to reduce late deliveries. Which formulation best connects the business problem to a data science deliverable?

  1. Purchase newer delivery vehicles to improve overall fleet reliability
  2. Survey customers about their satisfaction with delivery timing
  3. Create a dashboard showing historical on-time delivery rates by region
  4. Predict the probability that each shipment will arrive late so dispatchers can proactively reroute high-risk shipments Correct answer.

Explanation

This option links a business outcome (fewer late deliveries) to a predictive output (per-shipment risk) and a specific operational action (rerouting). It defines both what the model produces and how it will be used. The other options are capital investment, data collection, or descriptive reporting without a decision link.

Question 29Defining business problems

During problem definition, a stakeholder says the model must be 'as accurate as possible.' Which response best advances the definition of the business problem?

  1. Propose using the most complex algorithm available to ensure high accuracy
  2. Set a target of 99% accuracy and proceed with data collection
  3. Ask what decision the model will support and what error costs are associated with false positives versus false negatives Correct answer.
  4. Agree and begin collecting the largest possible dataset to maximize accuracy

Explanation

Accuracy is not a business objective; the relevant question is which kinds of errors are costly and how the output will be used. Clarifying the decision and the asymmetric costs of mistakes lets the team choose an appropriate metric and threshold. The other responses accept an underspecified goal and jump to technical choices prematurely.

Question 30Defining business problems

A team is defining a project to reduce equipment downtime in a factory. Which question is most important to answer before selecting a modeling approach?

  1. What maintenance action will be taken when a failure risk is predicted, and who decides? Correct answer.
  2. Which machine learning library will the team use for implementation?
  3. What is the average age of the machines in the factory?
  4. How many sensors are installed on each machine?

Explanation

The value of a predictive maintenance model depends on the action it triggers and the decision owner who will act on it. Establishing this link ensures the model output is actionable and helps define the required lead time and risk threshold. The other questions are technical or descriptive details that can be settled later.

Question 31Defining business problems

A data scientist is asked to 'find useful patterns in the transaction data' with no further context. Which response best reflects sound problem definition practice?

  1. Request additional data sources to increase the chance of finding meaningful patterns
  2. Build a model to predict transaction amount as a first exploratory step
  3. Run clustering on all available variables and present the segments to stakeholders
  4. Ask which business process or decision the patterns should inform and what outcome would count as useful Correct answer.

Explanation

Without a decision context or success criterion, 'useful patterns' cannot be evaluated or prioritized. Sound problem definition requires clarifying the business process, the decision owner, and the outcome that would make a finding valuable. Exploratory analysis may follow, but it should be guided by a defined purpose.

Question 32Defining business problems

A subscription business wants to reduce revenue lost to involuntary churn caused by failed credit card payments. Which problem definition is most actionable for a data science team?

  1. Survey customers who cancelled about their reasons for leaving
  2. Predict which payment attempts are likely to fail so the billing system can retry at an optimal time or prompt the customer to update payment details Correct answer.
  3. Analyze all reasons customers cancel and build a model to predict any cancellation
  4. Offer discounts to all customers who show any sign of reduced usage

Explanation

This option narrows the problem to a specific, addressable mechanism (payment failure) and defines a concrete action (retry timing or payment update prompt). It links the prediction to an operational lever that can prevent the revenue loss. The other options address broader churn or apply untargeted remedies.

Question 33Defining business problems

A hospital wants to reduce emergency department overcrowding. Which formulation best connects the business problem to a data science deliverable?

  1. Install more beds in the emergency department to increase capacity
  2. Analyze historical admission data to describe past overcrowding patterns
  3. Build a model to predict which patients will have the longest length of stay
  4. Forecast patient arrival volume by hour so staffing levels can be adjusted in advance Correct answer.

Explanation

Forecasting arrival volume by hour directly supports a staffing decision that can mitigate overcrowding. It defines a predictive output, a time granularity, and an operational action. The other options either describe the past, predict an outcome without a clear intervention, or propose a physical capacity change that does not require data science.

Question 34Defining business problems

A team is scoping a project and the business sponsor keeps adding new questions to the problem statement. Which practice best keeps the problem definition stable?

  1. Document a primary decision and success metric, and route additional questions to a separate backlog for future phases Correct answer.
  2. Expand the data collection effort to cover every possible question the sponsor raises
  3. Accept all new questions to ensure stakeholder satisfaction and address them later
  4. Refuse to meet with the sponsor until the original scope is completed

Explanation

A stable problem definition requires an agreed primary decision, success metric, and scope boundary. Capturing additional questions in a backlog preserves them without destabilizing the current project. Accepting every addition or expanding data collection indiscriminately leads to scope creep and delays.

Question 35Defining business problems

A stakeholder defines the project goal as 'increase customer lifetime value.' Which additional piece of information is most critical for defining a data science problem?

  1. The names of all departments involved in customer management
  2. The specific business lever the organization can pull to influence lifetime value and the timeframe over which to measure it Correct answer.
  3. The total number of customers in the database
  4. The software platform used for the customer relationship management system

Explanation

Lifetime value is an outcome that can be influenced through many levers, such as retention, upsell, or acquisition quality. Defining the problem requires knowing which lever the organization can act on and over what horizon, so the data science work can target the right prediction or analysis. The other details are contextual but do not determine the problem framing.

Question 36Defining business problems

A data science team proposes predicting which loan applicants will default. Before building the model, which question most directly tests whether the project has an identifiable decision owner?

  1. What is the historical default rate in the loan portfolio?
  2. Who will review the model's risk scores and what action will they take for high-risk applicants? Correct answer.
  3. Which credit bureau data sources are available for the model?
  4. Which evaluation metric will be used to measure model performance?

Explanation

An identifiable decision owner is the person or role who will act on the model output. Asking who reviews the scores and what they will do directly tests whether the prediction connects to a real decision process. The other questions concern data, descriptive statistics, or evaluation, none of which establish decision ownership.

Question 37Defining business problems

A team is comparing two framings of the same business issue: predicting which customers will respond to a promotion versus estimating the incremental effect of the promotion on sales. Which consideration most strongly favors the causal framing?

  1. The business needs to know whether the promotion itself caused the change in sales in order to decide whether to continue it Correct answer.
  2. The causal framing requires less data than the predictive framing
  3. The predictive framing always produces higher accuracy than the causal framing
  4. The causal framing avoids the need to define a target variable

Explanation

A predictive model identifies customers likely to respond, but response may occur without the promotion. A causal question estimates the incremental effect of the promotion, which is the relevant input for a decision about whether to continue or expand it. This distinction determines which framing fits the decision.

Question 38Defining business problems

A team discovers that the business problem as originally stated would require customer data that the organization is not legally permitted to use. What is the most appropriate next step in problem definition?

  1. Replace the restricted data with synthetic data and report results as if they were real
  2. Abandon all data science efforts for the business unit permanently
  3. Proceed with the project using the data anyway and address legal concerns after the model is built
  4. Re-scope the problem to rely only on legally permissible data, or redefine the decision so that it does not require the restricted data Correct answer.

Explanation

When legal constraints block a data source, the problem definition must be revised so that the project remains feasible and compliant. Re-scoping to permissible data or redefining the decision preserves the possibility of a valid project. Proceeding unlawfully, abandoning all efforts, or misrepresenting synthetic results are not acceptable.

Question 39Defining business problems

A team defines a project around 'improving delivery times.' Which change most improves the problem definition for a data science engagement?

  1. Expand the scope to include all logistics operations across the company
  2. Collect data on every possible factor that might affect delivery
  3. Add a requirement to use deep learning for the solution
  4. Specify the target reduction in late deliveries, the customer segment affected, and the operational decision the model will inform Correct answer.

Explanation

A strong problem definition specifies a measurable target, the affected population, and the decision the model will support. These elements make the problem bounded and testable. Adding technology constraints, broadening scope, or collecting undirected data does not sharpen the definition.

Question 40Defining business problems

A stakeholder insists that the model must be 'accurate.' Which response best advances problem definition?

  1. Ask what decision the model will support and what error costs are associated with false positives versus false negatives Correct answer.
  2. Set a target of 99% accuracy and proceed with data collection
  3. Propose using the most complex algorithm available to ensure high accuracy
  4. Agree and begin collecting the largest possible dataset to maximize accuracy

Explanation

Accuracy is not a business objective; the relevant question is which kinds of errors are costly and how the output will be used. Clarifying the decision and the asymmetric costs of mistakes lets the team choose an appropriate metric and threshold. The other responses accept an underspecified goal and jump to technical choices prematurely.

Question 41Data preparation and data quality

A data engineer needs to profile a new column before loading it into a warehouse. Which activity best describes data profiling?

  1. Encrypting sensitive columns to meet regulatory compliance requirements
  2. Designing the physical schema and indexes for a target warehouse table
  3. Applying transformations to normalize column values for downstream reporting
  4. Examining a dataset's structure, value distributions, and quality characteristics to understand its content Correct answer.

Explanation

Data profiling is an exploratory activity that examines structure, data types, value distributions, uniqueness, null rates, and patterns to build understanding of a dataset before it is used. It is distinct from transformation, schema design, or security controls, which act on the data rather than characterize it.

Question 42Data preparation and data quality

A column stores customer email addresses. During profiling, an analyst finds 4% of rows contain values with no '@' symbol. Which data quality dimension does this finding most directly violate?

  1. Validity Correct answer.
  2. Uniqueness
  3. Timeliness
  4. Completeness

Explanation

Validity concerns whether values conform to the required format, type, or domain rules for that attribute. Email addresses lacking '@' fail the format rule, so the finding is a validity defect rather than an issue of how current, how unique, or how fully populated the data is.

Question 43Data preparation and data quality

Which statement correctly distinguishes a data domain from a data type?

  1. A data domain applies only to numeric columns, while a data type applies only to text columns
  2. A data domain defines storage format, while a data type defines the allowed value set
  3. A data domain defines allowed values for an attribute, while a data type defines the storage format and operations Correct answer.
  4. A data domain and a data type are synonyms in relational modeling and are used interchangeably

Explanation

A domain is the set of permissible values for an attribute, sometimes enforced by constraints or reference tables. A data type governs representation, storage, and permissible operations. They are complementary concepts, not synonyms, and neither is restricted to numeric or text columns.

Question 44Data preparation and data quality

A retail company wants a single trusted source for product master data shared across sales, inventory, and finance systems. Which practice addresses this requirement?

  1. Increasing the retention period of transaction logs
  2. Implementing master data management with defined stewardship and matching rules Correct answer.
  3. Creating separate product tables in each system with no synchronization
  4. Archiving historical product records to cold storage

Explanation

Master data management establishes governance, stewardship, and matching or deduplication rules so that shared reference entities such as products have one authoritative definition. Isolated tables, archival, and log retention do not create a governed shared source of truth.

Question 45Data preparation and data quality

In a star schema, a fact table typically stores which kind of content?

  1. Unstructured documents and binary attachments
  2. Business rule definitions and validation constraints
  3. Descriptive attributes used for filtering and grouping
  4. Numeric measures and foreign keys referencing dimension tables Correct answer.

Explanation

Fact tables hold quantitative measures at a defined grain plus foreign keys to dimension tables. Descriptive attributes belong in dimensions, while rules and unstructured content are not characteristic of fact table design.

Question 46Data preparation and data quality

Which technique reduces the storage footprint of a large fact table while preserving query performance for aggregations?

  1. Converting all measures to variable-length text
  2. Partitioning the table by date and using columnar compression Correct answer.
  3. Adding more nullable columns to the fact table
  4. Removing foreign keys to the dimension tables

Explanation

Partitioning by a common filter such as date enables partition pruning, and columnar compression reduces storage while supporting efficient aggregate scans. Adding nullable columns or converting measures to text increases size, and dropping dimension keys breaks the model.

Question 47Data preparation and data quality

A pipeline must load only records changed since the previous run. Which approach supports this requirement?

  1. Full extract and truncate-reload of the entire source each run
  2. Incremental extraction using a change timestamp or change data capture Correct answer.
  3. Random sampling of source rows on each execution
  4. Manual export of the source into spreadsheets

Explanation

Incremental extraction uses a reliable change indicator such as an update timestamp, sequence, or change data capture stream to select only new or modified records. Full reloads move all data each time, while manual exports and random samples cannot guarantee complete change coverage.

Question 48Data preparation and data quality

What is the primary purpose of a data quality rule in a pipeline?

  1. To schedule jobs in the orchestration tool
  2. To document the business glossary for analysts
  3. To define the physical storage layout of target tables
  4. To express an expected condition that records are tested against and to trigger action on failure Correct answer.

Explanation

A data quality rule encodes an expected condition, such as a range or format check, that records are evaluated against. Failures can then route records to quarantine, halt the pipeline, or raise alerts, which is distinct from glossary documentation, storage layout, or scheduling.

Question 49Data preparation and data quality

A source system stores dates as text in the format 'DD/MM/YYYY', while the target expects a DATE type. Which preparation step is required before loading?

  1. Concatenating the text with a time zone offset string
  2. Aggregating the text values into monthly totals
  3. Hashing the text values to preserve privacy
  4. Parsing the text with an explicit format mask and casting it to a date type Correct answer.

Explanation

Converting text to a date requires parsing with the correct format mask so that day and month are not transposed, then casting to the target date type. Aggregation, concatenation, or hashing do not produce a valid typed date value.

Question 50Data preparation and data quality

An analyst joins a customer table to an orders table on customer_id and finds that total order revenue is inflated after the join. Which cause is most likely?

  1. The join used an inner join instead of a left join
  2. The revenue column uses a decimal data type
  3. The orders table contains multiple rows per customer, causing fan-out when aggregated Correct answer.
  4. The customer table has a primary key on customer_id

Explanation

Joining a parent to a child table produces one row per child record, so aggregating revenue after the join can double-count or multiply values. The fix is to aggregate the child table to the parent grain before joining, rather than changing join type or data type.