These 12 CBDA scenario questions mix all six exam domains in blueprint proportions: three on research questions, two each on sourcing data, analysis, reporting and decision-making, and one on organizational strategy. Each scenario asks what the analyst should do first, next or best. Answers and explanations follow the questions, each tied to a section of the Guide to Business Data Analytics.
How these CBDA scenario questions are built
The real exam asks you to judge situations, not recall terms, so every question here is a multi-step scenario. All 12 are original and written for this article; none come from exam dumps. Each is grounded in a Guide task, technique or case study, named in the answer key. Difficulty sits slightly above typical question banks, and two questions need a small calculation.
For single-domain practice first, use our first 35-question CBDA mock or the research questions drill. This set is for telling the domains apart.
CBDA exam at a glance
| Item | Detail |
|---|---|
| Format | 75 multiple-choice, scenario-based questions |
| Time | 2 hours |
| Exam fee | US$405 (IIBA members US$250); first year of membership included |
| Rewrite fee | US$195 (non-members US$350) |
| Passing score | Not published by IIBA |
Last verified: October 2026
What the questions cover
| Domain (weight) | Guide focus | Questions |
|---|---|---|
| 1. Identify the Research Questions (20%) | Existing model assumptions (2.1.8); reframing (3.15); plain-language questions (2.1.5) | 1–3 |
| 2. Source Data (15%) | Deriving predictors; sampling before full scale (2.2.6) | 4–5 |
| 3. Analyze Data (16%) | Translating metrics (2.3.7); course correction (2.3.5) | 6–7 |
| 4. Interpret and Report Results (20%) | Stakeholder analysis (2.4.7); 3-minute story (3.20) | 8–9 |
| 5. Use Results to Influence Decisions (20%) | Decision trees (3.8); integration before deployment (2.5.1–2.5.2) | 10–11 |
| 6. Organizational-Level Strategy (9%) | Centralized, decentralized or hybrid (2.6.6) | 12 |
Key ideas the set teaches:
- Problem shaping and reframing (3.15): shaping refines a problem until a solution process applies; reframing restates it in another context, such as the customer's view instead of the company's.
- Transforming source data (2.2.6): raw records often need business knowledge to become useful predictors, and methods can escalate from rules to a sample to full-scale analysis.
- Course correction (2.3.5): exploration and analysis repeat until the data can answer the question; the analyst weighs fixes on cost, feasibility and impact.
- Short-form communication (3.19, 3.20): the big idea compresses an insight into one decisive sentence; the 3-minute story moves from a hook to a few facts to an action.
- Decision analysis (3.8): a decision tree follows each option to its likely impact when insights conflict.
How to read a long CBDA scenario
- Name the domain and task. What is the team producing now: a research question, a data set, a model, a report, a recommendation or a structure? Options from other domains are the usual distractors.
- Place the stage. Phrases such as "the model has been trained" show which steps are done. Options that repeat a finished step or skip a missing one are usually wrong.
- Check the role. The business analysis professional brings domain context, translates and builds consensus; the data scientist selects and runs models. Prefer options where the analyst clarifies, translates or assesses business impact.
Then read the order word: FIRST is the earliest valid step, NEXT follows what the scenario already describes, BEST is the option most consistent with the Guide.
Practice Questions
Question 1
Meridel Freight relies on a delivery-time model built six years ago. Late deliveries caused three client penalties this quarter, but the planning team calls the misses normal variance. The operations director asks the analyst whether new analytics work is needed.
What should the analyst do NEXT?
- A. Recommend replacing the model with a commercial machine learning tool
- B. Review the assumptions and input data behind the existing model
- C. Accept the planning team's view and close the request as out of scope
- D. Start collecting weather and traffic data for a new forecasting model
Question 2
A fitness chain asks whether a price cut would reduce cancellations. The analyst reframes the problem from the member's perspective. The draft statement describes the problem, lists staff, members and finance as impacted stakeholders, and calls for a predictive approach using visit and payment data.
Which element completes the reframed problem statement?
- A. A list of the visualizations planned for the final report
- B. A project schedule with task owners and estimated durations
- C. The accuracy target the predictive model must reach
- D. A comparison with an alternative formulation of the problem
Question 3
A home insurer's analytics team is finalizing research questions for claims handling. A data scientist proposes: "Fit a gradient-boosted classifier to minimize log loss for the fraud label." The claims managers who will act on the results do not understand it.
What should the analyst do FIRST?
- A. Restate the question in plain business language and review it with all stakeholders
- B. Keep the wording, since the data scientist owns the research questions
- C. Train the claims managers on classification metrics before continuing
- D. Replace it with a descriptive question about past fraud losses
Question 4
Tallow Mobility suspects riders abuse promo codes by creating many accounts. The team has trip-level records: time, pickup zone, fare, promo code and device ID. No single trip-level field separates abusers from normal riders.
What should the analyst do NEXT?
- A. Ask the data scientist to try more complex algorithms on the trip fields
- B. Report that the available data cannot answer the research question
- C. Use business knowledge to derive account- and device-level measures
- D. Launch a rider survey about how people use promotions
Question 5
An electricity retailer wants to detect meter tampering and has candidate rules from field inspections. The full history holds 2.4 billion readings. Analysing all of it on a new big-data platform would take ten weeks; a 1% sample (24 million readings) takes one week on existing tools.
What is the BEST approach?
- A. Test the rules on the sample to find predictor variables, then scale up
- B. Wait for the big-data platform so the analysis uses every record
- C. Skip the rules and train a model on the full history when available
- D. Ask inspectors to collect a new data set to replace the readings
Question 6
Orchard Life is building a model that assigns applications to one of six risk classes. The data scientists prefer Cohen's kappa to plain accuracy as the evaluation measure. The underwriters who will rely on the model do not see why, and their trust is slipping.
What should the analyst do?
- A. Ask the data scientists to switch to accuracy, as underwriters prefer it
- B. Send the underwriters the statistical definition and formula of kappa
- C. Leave the metric to the data scientists and focus on reporting
- D. Explain kappa as two underwriters' agreement with chance agreement removed
Halfway there? Create a free account to get more CBDA practice questions. The CBDA exam simulator has a full 75-question set in exam mode or practice mode with explanations.
Question 7
A language-learning app wants to target a discount at likely premium buyers. A logistic regression was trained on 30,000 of 50,000 CRM records and tested on the rest. The data scientist suspects customers and free-trial leads behave differently and that income is missing.
| Data set | Records | Accuracy |
|---|---|---|
| Training | 30,000 | 93% |
| Testing | 20,000 | 64% |
What should the analyst do NEXT?
- A. Recommend launching the campaign, since training accuracy is high
- B. Help assess the cost, feasibility and impact of possible fixes with stakeholders
- C. Remove all leads so that only existing customers are modelled
- D. Ask the data scientist to retrain the same model until testing improves
Question 8
Brightwell Bakeries is being acquired. The acquirer's marketing head asks why the team chose one delivery app over five others. The team sends a table of regional user counts per app. She replies that it does not show how the choice supports the strategy.
What should the analyst do FIRST?
- A. Send a longer appendix with all source data and calculations
- B. Re-run the analysis with a larger sample of app users
- C. Assess her as a stakeholder to plan how, and at what detail, to communicate
- D. Build an interactive dashboard so she can explore the data herself
Question 9
A hospital network found that 18% of outpatient no-shows follow reminders sent less than 24 hours ahead. The analyst has five minutes with the chief operating officer, who has not seen the analysis, to get approval to change the reminder schedule.
Which outline BEST fits a 3-minute story?
- A. Purpose with the 18% hook, a few plain facts, then a request to approve
- B. Full methodology, each statistical test, then the finding at the end
- C. A walkthrough of every no-show metric, leaving conclusions to the COO
- D. The finding alone, with the approval request sent later by email
Question 10
Two analyses at Corvo Home Goods disagree: one favours a national launch, the other a two-region launch. The analyst builds a decision tree (thousands of euros).
| Option | Outcome | Probability | Result |
|---|---|---|---|
| National | Strong demand | 0.6 | +900 |
| National | Weak demand | 0.4 | -500 |
| Two regions | Strong demand | 0.8 | +300 |
| Two regions | Weak demand | 0.2 | -100 |
What should the analyst do NEXT?
- A. Recommend two regions, because its expected value of 260 is higher
- B. Hold any recommendation until the two analyses agree
- C. Recommend national, because its best case is the highest
- D. Recommend national (340 vs 220) and state the assumptions and downside
Question 11
Vesta Grocer ran a proof of concept to detect abnormal discount codes. On static data, a mix of models reached an F1 score the business accepted. The sponsor wants a recommendation on deployment.
What should the analyst do BEFORE recommending deployment?
- A. Raise the F1 target and rerun the proof of concept with new models
- B. Analyse how the solution fits current processes and its return in production
- C. Ask IT to deploy now, because the business accepted the results
- D. Present the F1 score to executives as proof the solution is ready
Question 12
Kestrel Pensions launched a retirement app a year ago, and all its data expertise sits in the app team. Other departments want the app's customer data, including income and net worth. Data maturity is low, hiring budget is tight, and leaders want quick progress.
What is the BEST recommendation?
- A. Build a fully centralized analytics function now with a new team
- B. Let each department build its own analytics team and standards
- C. Set standards and light governance in the app team, centralize later
- D. Pause all data sharing until an enterprise data strategy is approved
Ready for a full run? The CBDA exam simulator has 75 questions across all six domains, with a timed exam mode and an explained practice mode. Create a free account to start. certificate.tips is our product.
Answer Key and Explanations
1. Correct answer: B. The Guide advises verifying the principles an existing model was built on. In its investment bank case, experts called losses a cost of business, and examining model assumptions showed that news and macro data were ignored. Source: Guide to Business Data Analytics, 2.1.8 Case Study for Identify the Research Questions Why the others are wrong: A jumps to a solution; C ignores a real problem; D collects data before the problem is understood.
2. Correct answer: D. A reframed problem statement has four parts: the problem, impacted stakeholders, key needs (approach, data, analytics type), and a comparison to an alternative formulation. The draft lacks only the comparison. Source: Guide to Business Data Analytics, 3.15 Problem Shaping and Reframing Why the others are wrong: A belongs to reporting; B to planning; C is a success criterion set during analysis.
3. Correct answer: A. Good research questions avoid technical language and are reviewed with all stakeholders for consensus. The data scientist restates them mathematically later, during Perform Data Analysis. Source: Guide to Business Data Analytics, 2.1.5 Formulate Research Questions Why the others are wrong: B skips consensus; C burdens the wrong group; D changes intent without agreement.
4. Correct answer: C. In the Guide's telecom fraud case, raw call records were not useful as they were. Domain knowledge turned them into subscriber-level measures, such as night-time calls, which became the real predictors. Source: Guide to Business Data Analytics, 2.2.6 Case Study for Source Data Why the others are wrong: A treats a data gap as a model problem; B gives up early; D adds a slow, self-reported source.
5. Correct answer: A. The same case escalates from rule discovery to sampled analysis to big-data analysis. Sampling tests the hypothesis and finds predictors faster, at some cost in accuracy, and the full-scale step builds on those predictors. Source: Guide to Business Data Analytics, 2.2.6 Case Study for Source Data Why the others are wrong: B delays learning by ten weeks; C discards domain rules; D replaces a suitable source.
6. Correct answer: D. In the Guide's insurance case, the analyst learned the data scientists' terms and translated them, explaining kappa as two underwriters classifying the same applications, corrected for chance agreement. The measure must fit the business problem and be understood. Source: Guide to Business Data Analytics, 2.3.7 Case Study for Analyze Data Why the others are wrong: A chooses by preference; B gives detail without translation; C drops the bridging role.
7. Correct answer: B. A large train–test gap suggests weak or confounded predictors. The Guide lists three fixes: more data, a randomized or A/B experiment, or methods suited to confounding. The analyst evaluates their cost-benefit, feasibility and impact with stakeholders. Source: Guide to Business Data Analytics, 2.3.5 Assess the Analytics and System Approach Taken Why the others are wrong: A trusts results that do not generalise; C drops a needed group; D repeats a failing approach.
8. Correct answer: C. In the Guide's ice cream retailer case, an acquirer's director rejected an adoption table. The analyst first ran a quick stakeholder analysis, chose an active, transparent approach, then built a simple decision matrix at the right detail. Source: Guide to Business Data Analytics, 2.4.7 Case Study for Interpret and Report Results Why the others are wrong: A adds detail when detail was the problem; B assumes an analysis flaw; D shifts interpretation to her.
9. Correct answer: A. A 3-minute story opens with intent and a story hook, keeps analysis to simple facts ending in one finding, and closes with a stakeholder action such as a decision. Source: Guide to Business Data Analytics, 3.20 3-Minute Story Why the others are wrong: B buries the finding; C omits the message; D drops the decision prompt.
10. Correct answer: D. National: 0.6 × 900 − 0.4 × 500 = 340. Two regions: 0.8 × 300 − 0.2 × 100 = 220. In Domain 5, decision analysis evaluates conflicting insights, and a tree follows each decision to its logical impact; stating the 500 downside lets leaders judge risk. Source: Guide to Business Data Analytics, 3.8 Decision Modelling and Analysis Why the others are wrong: A miscalculates; B leaves the conflict open; C ignores probabilities.
11. Correct answer: B. The Guide's pricing anomaly example says analytics results alone do not justify deployment. The analyst checks integration with current processes, production ROI, cheaper alternatives, process changes such as real-time blocking, and deployed behaviour. Source: Guide to Business Data Analytics, 2.5.2 Develop Implementation Plan (example under 2.5.1 Recommend Actions) Why the others are wrong: A reworks an accepted result; C skips integration analysis; D sells a metric as a business case.
12. Correct answer: C. In the Guide's retirement product case, low maturity made decentralization risky, and budget and hiring time ruled out a central team. The analyst recommended standards and light governance in the product team, short-term experts, and later centralization. Source: Guide to Business Data Analytics, 2.6.6 Case Study for Guide Organizational-Level Strategy for BDA Why the others are wrong: A ignores budget and speed; B risks silos and misuse of sensitive data; D stops value delivery.
How to score this set
- 11–12: You read domain and stage reliably; move to timed full-length practice.
- 8–10: Look for a shared domain or stem type in your misses; FIRST/NEXT misses usually mean a skipped step.
- Below 8: Reread the Guide case studies, then try single-domain sets such as our business decision-making quiz.
Twelve questions cannot predict a result, and IIBA does not publish a passing score. For pacing, take the full 75-question CBDA exam simulator after you sign up free.
What candidates say about scenario questions on the CBDA
The author of our CBDA certification overview flagged about 20 of 75 questions and found the exam much harder than question banks, with more statistical reasoning and analytical decision-making than definitions. Mufaddal Haidermota, certified in December 2022, wrote on Medium that per-domain mocks and full-length tests anchored his preparation. Techcanvass candidate Ivan credited mind maps and suggested two to three hours of daily study. Trainer LN Mishra of Adaptive US advises practitioners to take timed mocks and review every wrong answer.
FAQ
What makes CBDA scenario questions different from definition questions?
They describe a business situation and ask for a judgement, such as the analyst's next step. You must identify the domain, the stage and the analyst's role. Distractors are often valid actions from another domain or stage.
How should I split study time across the six domains?
IIBA weights Domains 1, 4 and 5 at 20% each, Domain 3 at 16%, Domain 2 at 15% and Domain 6 at 9%. Practise in similar proportions, then add time wherever your mixed-set scores are lowest.
Does the CBDA expect business analysts to run statistical models?
The Guide describes the analyst as more likely to support the data scientist than to run models. The analyst brings domain knowledge, translates results and helps define success criteria, which still requires basic statistics.
Sources
- Certification in Business Data Analytics (CBDA) — IIBA, checked October 2026
- Certification Fees — IIBA, checked October 2026
- CBDA Certification Handbook — IIBA, checked October 2026
- Business Data Analytics Certification — Mufaddal Haidermota, Medium, January 2023
- CBDA Certification Journey: Ivan — Techcanvass, 2024
- Mastering the CBDA Exam — LN Mishra, Adaptive US, October 2024