Weekly Breakdown · CBDA · Guide to Business Data Analytics
Welcome to Weekly Breakdown, a new issue every Monday from certificate.tips. Each issue takes one certification exam, builds one realistic case study and runs it through five or six exam-style questions. Try each question before you read the breakdown: we explain why the right answer is right, why the other options are wrong, and which section of the official guide says so. All questions are original and based on the official guides. We do not publish exam dumps.
This first issue is a bigger one: three cases and 15 questions, one case each for the Identify the Research Question, Source Data and Interpret and Report Results domains. The question of the week, the COO who wants a churn model on day one, is Question 1.
Case 1. Pedalway: rides are down 18%
Pedalway runs a bike-share service in a mid-size city: 1,200 bikes, 140 docks and about 21,000 members. Between April and July monthly rides fell 18%, and revenue from single rides fell with it.
Leadership has just paid for a new analytics platform and wants results fast. You are the business data analyst on the initiative. Guide sections behind this case: 2.1.1 Define Business Problem or Opportunity, 2.1.2 Identify and Understand the Stakeholders, 2.1.5 Formulate Research Questions and 3.15 Problem Shaping and Reframing.
- Meeting 1. The COO asks for a churn prediction model by the end of the month.
- Meeting 2. Your first stakeholder list has the COO, the marketing lead and the finance lead.
- Meeting 3. The team agrees on a problem statement. Finance asks if it is the research question.
- Meeting 4. A data engineer proposes a research question full of model terms.
- Meeting 5. A product designer suggests looking at the drop from the rider's side.
Question 1. Meeting 1: the COO wants a churn model on day one
In the first meeting the COO says rides are down 18% and, since the analytics platform is already paid for, asks for a churn prediction model by the end of the month. What should the analyst do FIRST?
- A. Start building the churn model, since the COO is the sponsor and owns the budget.
- B. Facilitate discussions with the right stakeholders to define the business problem first.
- C. Ask the data team to export all ride and membership data so analysis can begin.
- D. Write three predictive research questions about churn and send them for approval.
Answer: B.
Section 2.1.1 says defining the business problem or opportunity is often the first step. It warns that the urgency to see results from an analytics investment creates a tendency to jump into a solution focus, which can lead to a misdiagnosis of the business problem. Problem discovery runs in parallel with identifying the right stakeholders. A is wrong: the sponsor's preferred solution is exactly the solution focus the Guide warns about. C is wrong: pulling all data starts from the data, not from the problem. D is wrong: research questions are derived from a business problem that is not defined yet.
Read more: CBDA Research Questions Drill: Framing Problems and Stakeholders
Question 2. Meeting 2: who belongs on the stakeholder list
The analyst's first stakeholder list has the COO, the marketing lead and the finance lead. Which question BEST reflects the data view of stakeholders that the Guide asks for?
- A. Who has the largest budget for the analytics initiative?
- B. Who creates the ride and dock data, who sees it, and who decides on the insights?
- C. Which stakeholders attended the last quarterly business review?
- D. Who has the most experience with machine learning tools?
Answer: B.
Section 2.1.2 says an analytics initiative needs a data view of the business problem: who is creating the relevant information, who is exposed to the data, and which decision-makers are influenced by the insights. At Pedalway that adds people the first list missed, such as the dock maintenance crew who log faults and the app team. A is wrong: budget is not part of the data view. C is wrong: attendance at a past meeting says nothing about the data. D is wrong: analytics experience matters for communication, but it is not how relevant stakeholders are found.
Question 3. Meeting 3: problem statement or research question
The team agrees on this statement: "Ridership fell 18% between April and July, and single-ride revenue fell with it." The finance lead asks if this is the research question. Which response is MOST accurate?
- A. Yes. Once the problem is agreed, it serves as the research question.
- B. No. The business problem is a higher-level description and may lead to several research questions.
- C. No. A research question can be written only after all the data is collected.
- D. Yes, as long as the statement includes a numeric target for July.
Answer: B.
Section 2.1.1 notes that a business problem or opportunity is usually a higher-level description than a research question. The outcome of this task is a business problem that may lead to one or more research questions, and further analysis is then done to formulate the right analytics questions. A is wrong: it skips the step of formulating questions from the problem. C is wrong: Section 2.1.5 says the question is formulated before the detailed analytics work. D is wrong: a number in the statement does not turn a problem into a question.
Question 4. Meeting 4: the wording of the question
A data engineer proposes: "Train a gradient boosting classifier on dock telemetry and membership features to output churn probability." Which version is the BEST research question to take to stakeholders?
- A. Which factors are linked to members riding less often since April?
- B. Train the classifier and report the AUC to the COO every week.
- C. Build a dashboard of dock telemetry for the engineering team.
- D. Use a neural network instead, since it fits large data sets better.
Answer: A.
Section 2.1.5 says good analytics questions are clearly stated and do not use technical language, and they are reviewed with all stakeholders to reach consensus. The plain-language version asks what drives the change, much like the Guide's example of asking which factors influence customer experience. B is wrong: it restates the technical method and adds a model metric. C is wrong: a dashboard is a deliverable, not a question. D is wrong: swapping one algorithm for another is still a solution, not a question.
Question 5. Meeting 5: a new point of view
So far the team has asked "How do we stop members from leaving?" A product designer suggests looking at it from the rider's commute instead: weather, distance to the nearest working dock and travel time compared with the bus. Which technique is the team applying?
- A. Problem shaping, refining the problem until a solution process can be applied
- B. Problem reframing, restating the problem in a different context
- C. Root cause analysis with 5 Whys to narrow down a single cause
- D. Hypothesis testing to reject a null hypothesis about churn
Answer: B.
Section 3.15 defines problem shaping as refining the problem to a state where a solution process can be applied, and problem reframing as restating the problem in a different context. The Guide's own example moves a health insurer from its own view to the customer's perspective, the same move Pedalway makes from the operator to the rider. A is wrong: shaping refines the same problem rather than changing its context. C is wrong: the Guide calls 5 Whys reductive, it narrows the problem instead of reframing it. D is wrong: hypothesis testing comes later, in the analysis.
Case 2. Oakline Home: which deliveries are late?
Oakline Home sells furniture online and ships about 38,000 orders a quarter through two carriers. Complaints about late deliveries doubled in a year. The agreed research question: which carriers and regions cause the most late deliveries?
Now the team needs the data. Guide sections behind this case: 2.2.1 Plan Data Collection, 2.2.3 Collect Data, 2.2.4 Validate Data and 3.4 Data Dictionary.
- Week 1. Delivery dates exist in two places, and one of them is a cleaned spreadsheet.
- Week 2. The two carriers code delivery status in different ways.
- Week 3. The carrier file for February arrives, but it covers January.
- Week 3. The head of logistics asks who signs off the data sources.
- Week 4. Two departments mean different things by a late order.
Question 6. Week 1: two sources of delivery dates
Delivery dates exist in the central data warehouse and in a spreadsheet the operations team already cleaned and summarized. The spreadsheet is easier to use, but nobody can explain how three of its columns were calculated. What should the analyst do?
- A. Use the spreadsheet, because cleaned data saves time and cost.
- B. Assess data quality and consider a direct pull from the warehouse, even if it means more work.
- C. Use both sources and average the delivery dates where they differ.
- D. Skip delivery dates and rely on customer complaint counts instead.
Answer: B.
Section 2.2.1 describes this exact choice: data from a centrally managed data warehouse or from a peripheral secondary source where the data has already been manipulated. An assessment of data quality may be needed. A direct pull means more work and overhead, but that can be acceptable if the secondary source is questionable. A is wrong: ease of use does not outweigh unexplained calculations. C is wrong: averaging hides the discrepancy instead of resolving it. D is wrong: complaint counts cannot answer which carriers and regions are late.
Read more: Sourcing and Validating Data: Building the Right Dataset
Question 7. Week 2: two carriers, two codes
Carrier 1 reports delivery status as numeric codes (1, 2, 3). Carrier 2 uses letters (D, F, R). Both feeds will be merged into one data set. What does the Guide expect the analyst to do?
- A. Keep both feeds as they are and let the model learn the codes.
- B. Identify the need to reconcile the data elements across the two sources.
- C. Drop the carrier with letter codes, since numeric codes are easier to query.
- D. Ask the carriers to stop sending status codes until the analysis ends.
Answer: B.
Section 2.2.3 says that when data comes from different sources, analysts check whether the sources represent the same data in the same way. Its example is one source coding gender with numbers and another with letters, which creates the need to reconcile data elements. Some discrepancies also need domain knowledge, because a program cannot spot them. A is wrong: leaving codes unreconciled carries the problem into the analysis. C is wrong: dropping a carrier removes half of the data the question needs. D is wrong: stopping the feed blocks collection instead of fixing it.
Question 8. Week 3: the February file
The analyst requested carrier data for February. The file that arrives on 3 March covers 1 to 31 January. Which data quality characteristic does this issue affect?
- A. Uniqueness
- B. Completeness
- C. Timeliness
- D. Accuracy
Answer: C.
Section 2.2.4 lists five characteristics of high-quality data. Timeliness might be assessed by checking whether the data received is for the period requested. A January file sent for a February request fails that check, however clean the rows are. A is wrong: uniqueness is about duplicate records. B is wrong: completeness is about missing values such as nulls in required fields. D is wrong: accuracy compares data with what the source intended, for example front-end numbers against the database.
Read more: CBDA Data Quality Questions: Validation, Privacy, and Ethics
Question 9. Week 3: who signs off
The head of logistics asks who approves the data sources and sets the acceptance criteria for judging whether the delivery data is accurate. What is the BEST answer?
- A. The data engineer who built the ETL job, as part of technical testing
- B. Key business stakeholders with the authority to approve sources, with the analysts
- C. The carriers, because they produce the delivery status data
- D. The analyst alone, since validation at this stage is high-level
Answer: B.
Section 2.2.4 splits validation in two. Business validation means business stakeholders approve the data sources and set the acceptance criteria for accuracy. It is performed by key stakeholders with the authority to approve sources and the knowledge to assess accuracy, in collaboration with the analysts. Technical validation is the testing of data quality. A is wrong: technical testing is the other half of validation, not approval. C is wrong: producing data does not give approval rights. D is wrong: validation is high-level at this point, but the analyst does not approve sources alone.
Question 10. Week 4: two meanings of late
The warehouse system marks an order late when it ships after the promised date. Customer service marks it late when it arrives after the promised date. Their reports differ by 9 percentage points. Which technique BEST addresses this?
- A. A data dictionary that standardizes the definition of the data element across sources
- B. A survey of customers asking what they mean by a late delivery
- C. A 3-minute story that explains the gap to the head of logistics
- D. A heat map of late orders by region and by week of the quarter
Answer: A.
Section 3.4 says a data dictionary is used to standardize the definition of a data element and enable a common interpretation within one data source or across several. In analytics it can also track business rules for derived data, such as the rule that decides when an order counts as late. B is wrong: customers do not define the company's data elements. C is wrong: a story explains results, it does not fix a definition. D is wrong: a heat map of an inconsistent field only shows the inconsistency.
Case 3. Greenbasket: reporting a loyalty app promotion
Greenbasket is a supermarket chain with 60 stores. It ran a 10-week promotion in its loyalty app, and the analysis is done. Now the results must reach the CFO, the marketing director and 60 store managers.
Guide sections behind this case: 2.4.3 Determine Communication Needs of Stakeholders, 2.4.4 Derive Insights from Data and 3.19 The Big Idea.
- Day 1. The analyst plans communication for three audiences.
- Day 2. A striking pattern appears in stores near universities.
- Day 3. A data scientist wants to open the marketing readout with a technical chart.
- Day 4. The analyst drafts a one-sentence big idea for the CFO.
- Day 5. Marketing wants four insights in one big idea.
Question 11. Day 1: three audiences
Before building any slides, the analyst records for the CFO, the marketing director and the store managers what information matters most to each, how and how often they want updates, who decides and what biases they may carry. What is the PRIMARY purpose of this work?
- A. To customize communications to each group so the message is clearly understood
- B. To decide which group should receive the results first
- C. To confirm the data sources before the analysis is repeated
- D. To replace the stakeholder analysis done at the start
Answer: A.
Section 2.4.3 says determining the communication needs of stakeholders enables customization of communications to individual stakeholders or groups so the message is clearly understood. Its list of stakeholder preferences matches the analyst's notes: relevant information, how and how often to receive it, who the decision-makers are and what biases they carry. B is wrong: order of delivery is not the purpose. C is wrong: confirming data sources belongs to Validate Data. D is wrong: the Guide uses the stakeholder analysis as an input, it does not replace it.
Question 12. Day 2: a striking pattern
Stores near universities had the highest app redemptions and the highest discount cost. The marketing director wants to forecast next year's promotion budget per store from the number of students nearby. What should the analyst do?
- A. Agree, because a strong pattern in the data is enough for a forecast.
- B. Apply statistical judgment and look for stronger evidence before modelling on the pattern.
- C. Remove the university stores from the data set as outliers.
- D. Report the pattern only to the CFO, since it affects budget.
Answer: B.
Section 2.4.4 gives the example of ride density matched with a city's crime rate: a fascinating correlation, but demand should not be modelled on the crime rate without stronger evidence of a relationship. Analysts use sound statistical judgment and explanatory analysis to turn data patterns into useful insights. A is wrong: a correlation alone is not evidence of a relationship to model on. C is wrong: removing the stores hides the pattern instead of testing it. D is wrong: the issue is the evidence, not the audience.
Read more: Correlation vs Causation on the CBDA Exam: Reading a Sales Lift
Question 13. Day 3: which chart for marketing
A data scientist plans to open the marketing readout with an error residue graph that shows prediction error falling as predictors are added to the redemption forecast. What is the BEST advice?
- A. Keep it, because it proves the forecast is technically sound.
- B. Use it to choose the number of variables, and show marketing how promo spend relates to revenue.
- C. Replace it with a table of all model coefficients for full transparency.
- D. Show both charts side by side so each person can pick one.
Answer: B.
Section 2.4.4 uses this exact example. An error residue graph is a technical visualization, useful for choosing the optimum number of variables in a forecast. A marketing stakeholder will likely be more interested in a visualization that shows how ad spend relates to overall revenue. The communication needs from Section 2.4.3 shape which visual to use. A is wrong: technical soundness is not what marketing needs to see. C is wrong: a coefficient table is even more technical. D is wrong: two charts side by side push the interpretation onto the audience.
Question 14. Day 4: drafting the big idea
The analyst drafts a one-sentence big idea for the CFO. Which draft BEST matches the Guide's description of a big idea?
- A. Tracking the results of our loyalty app promotion this quarter.
- B. Moving 20% of the flyer budget to app offers would lift basket size by 3% by Q2.
- C. Our loyalty app data shows many interesting trends across 60 stores.
- D. We recommend further analysis before any budget decision is made.
Answer: B.
Section 3.19 says a big idea is a single, powerful statement that compels a decision, and that it articulates a point of view, conveys what is at stake and is one complete sentence. The Guide contrasts a vague line about tracking campaign success with a specific statement linking more digital investment to more revenue. A is wrong: it mirrors the Guide's example of what is not a big idea. C is wrong: it has no point of view and no action. D is wrong: it postpones the decision instead of recommending one.
Read more: CBDA Interpret and Report Results: Insight and Storytelling Quiz
Question 15. Day 5: four insights, one sentence
The results include four insights at different levels: a chain-wide budget shift, a regional pricing change, a store-level display issue and an app bug. The marketing director wants all four in one big idea. Which limitation of the big idea does this show?
- A. Multi-level insights are difficult to communicate through the big idea.
- B. The big idea needs a technical audience to interpret it.
- C. The big idea cannot be used in a 3-minute story.
- D. The big idea is useful only before the data is analyzed.
Answer: A.
Section 3.19 lists three limitations: the big idea may not be useful when evidence is lacking or stakeholders are biased, without stakeholder analysis it may lead to conflict, and multi-level insights are difficult to communicate through it. Four insights at chain, region, store and app level are a multi-level set. B is wrong: needing technical knowledge is a limitation of technical visuals in Section 3.18. C is wrong: the Guide lists use as a closing summary for a 3-minute story as a strength. D is wrong: the big idea conveys an insight after analysis.
What this case teaches
- Define the business problem with the right stakeholders before anyone picks a model. Urgency to show results is the trap, and a business problem can lead to several plain-language research questions.
- Source data top-down and check it twice: business stakeholders approve sources and acceptance criteria, technical checks cover accuracy, completeness, consistency, uniqueness and timeliness.
- Match the message to the audience: business visuals for business stakeholders, stronger evidence before modelling on a correlation, and one complete sentence for the big idea.
Practice the full exam
Our CBDA practice exam has 75 scenario questions across all six domains of the Guide, and it is free right now. Last verified: October 2026.
FAQ
What is Weekly Breakdown?
A weekly series from certificate.tips. Every Monday we take one certification exam, build a realistic case and break down exam-style questions on it with references to the official guide.
Are these real CBDA exam questions?
No. All 15 questions are original and written from the text of the IIBA Guide to Business Data Analytics. We do not publish exam dumps.
Which Guide sections does this issue cover?
Sections 2.1.1, 2.1.2, 2.1.5 and 3.15 in case 1, Sections 2.2.1, 2.2.3, 2.2.4 and 3.4 in case 2, and Sections 2.4.3, 2.4.4 and 3.19 in case 3.
Sources
- IIBA: Guide to Business Data Analytics
- IIBA: Certification in Business Data Analytics (CBDA)
- IIBA: CBDA Certification Handbook
- Elizabeth Ricks, storytelling with data: What toddlers can teach us about data storytelling (2020)
- Mufaddal Haidermota on Medium: Business Data Analytics Certification (2023)
The case and company are fictional. Not affiliated with Scrum.org or IIBA.