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 We do not publish exam dumps.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 issue has three cases and 15 questions: one case on the current and future state in the Identify the Research Questions domain, one on data sets, data mapping and loading in the Source Data domain, and one on the statistics behind the Analyze Data domain. Issues 1 and 2 used other tasks of the same Guide. The question of the week, the VP who already knows why sales fell, is Question 1.
Case 1. Larkfield Pharmacy: online sales are down 12%
Larkfield Pharmacy runs 58 stores and an online shop for health and beauty products. Over the last two quarters online sales fell 12%, from 2.5 million to 2.2 million dollars a quarter.
The VP of digital already has a view on why. You are the business data analyst on the initiative. Guide sections behind this case: 2.1.3 Assess Current State, 2.1.4 Define Future State and 2.1.6 Plan Business Data Analytics Approach.
- Week 1. The VP says the drop is caused by low customer engagement.
- Week 2. Three years of order history show a pattern.
- Week 3. The team writes down the future state.
- Week 6. The VP wants daily reports and two more KPIs.
- Week 7. The team plans the first stage of the work.
Question 1. Week 1: the VP already knows the cause
In the kickoff the VP of digital says online sales fell 12% because customers are not engaged, and asks the analyst to find out which engagement campaigns to run. What is the BEST response from the analyst?
- A. Accept the diagnosis, since the VP owns the online channel and knows its customers.
- B. Treat the claim as a hypothesis and check the data for evidence of the link.
- C. Formulate research questions on which engagement campaigns lift online sales the most.
- D. Start a customer engagement survey so the campaign plan is ready this quarter.
Answer: B.
Section 2.1.3 says the tools, techniques and models of a current state assessment are evidence-driven. Its own example is almost this case: a business leader stating that internet sales fell due to a lack of customer engagement does not conclusively link the two, so analysts look for concrete evidence from data and analysis to substantiate that hypothesis. A is wrong: the seniority of the person making a claim is not evidence. C is wrong: it builds research questions on a cause nobody has checked yet. D is wrong: it jumps to a solution and still assumes the cause is engagement.
Read more: Identify the Research Questions: CBDA Domain 1, All Tasks
Question 2. Week 2: the dip that comes back every year
The current state assessment includes three years of order history. Online sales dipped in the same two quarters every year, and about two thirds of this year's 12% drop matches the usual dip. How should this finding affect the research question?
- A. Refine it: study the seasonal pattern and how much of the drop it explains.
- B. Keep it as agreed, since the research question was already approved.
- C. Set the finding aside, since historical data says little about the future.
- D. Close the initiative, since a seasonal drop needs no further analysis at all.
Answer: A.
Section 2.1.3 says a current state analysis may reveal constraints, insights and assumptions that affect the nature of the research question, and its results set a baseline for the change. In the Guide's bicycle example, a clear seasonal trend in historical data turns the research question into a study of seasonal patterns. B is wrong: the Guide shows current state findings refining the original question into a more relevant one. C is wrong: historical data is exactly what a current state analysis studies. D is wrong: about a third of the drop is still unexplained.
Question 3. Week 3: a future state nobody can measure
The team agrees on this future state: "A better online shop that our customers love." Which addition would MOST improve it?
- A. Measurable objectives, such as quarterly online sales and repeat orders.
- B. A list of every internal and external data source the team will need.
- C. The name of the forecasting model the data team will build.
- D. A detailed schedule of the analytics work for the next 12 months.
Answer: A.
Section 2.1.4 says defining the future state includes making it clear and understandable, achievable with the resources available, shared by key stakeholders, and backed by measurable objectives that show the vision is met. Analysts facilitate discussions so decision-makers select the most appropriate measures, often a mix of strategic and operational KPIs. Nobody can measure how much customers love a shop. B is wrong: data sources are planned in the Source Data domain, not in the future state. C is wrong: a model is a solution choice, not a measure of success. D is wrong: a schedule says when work happens, not what success looks like.
Question 4. Week 6: the vision grows
Six weeks in, the VP asks to move the online sales report from monthly to daily and to track two more KPIs. The future state was agreed in week 3. What should the analyst do?
- A. Decline the request, since the future state was agreed and is now fixed.
- B. Restart the initiative with a new current state assessment.
- C. Update the future state with the stakeholders and record the change.
- D. Add the daily report quietly, so the team does not lose any time.
Answer: C.
Section 2.1.4 says defining the future state is continuous and iterative: the future state of an analytics initiative evolves throughout the life cycle of the engagement, and analysts manage and record the changes. The Guide even names tracking more KPIs and moving reports from monthly to daily or weekly as typical parts of a growing analytics vision. A is wrong: a frozen future state contradicts the iterative nature of the task. B is wrong: a wider vision does not make the current state assessment invalid. D is wrong: an unrecorded change breaks the shared vision the stakeholders agreed on.
Question 5. Week 7: the first stage of the work
The research question turns out to be more complex than expected. For the first stage the team plans a static extract from the web shop with no data pipelines, low governance, only key stakeholders engaged and a flexible data architecture. The goal is to see if the approach works at all. Which stage is this?
- A. A proof of concept stage, focused on the feasibility of the analytics approach.
- B. A pilot stage, focused on integration and quality issues at limited scale.
- C. A production stage, focused on business value for customers.
- D. A discovery stage, focused on mapping the current state of the business model.
Answer: A.
Section 2.1.6 says that when research questions or solution approaches are more complex than anticipated, analytics initiatives run in stages of maturity. The proof of concept stage focuses on the feasibility of the analytics approach and has low formality: static and limited data without pipelines, low governance, only key stakeholder engagement and a flexible solution and data architecture. B is wrong: a pilot uses dynamic data integrated from most known sources and some governance. C is wrong: production needs a deployed governance structure and an integrated architecture. D is wrong: the Guide does not list a discovery stage, and the current state is assessed in Section 2.1.3.
Case 2. Northpine Fitness: one warehouse for two gym chains
Northpine Fitness runs 31 gyms and has just bought Ridgeway Gyms, a chain of 12 clubs with about 19,000 members on a different membership system. Across both chains 23% of new members cancel within 90 days.
The agreed research question: which groups of new members cancel within 90 days, and what do they have in common? Now Ridgeway's data has to join the Northpine data warehouse. Guide sections behind this case: 2.2.2 Determine the Data Sets, 3.6 Data Mapping, 3.10 Extract, Transform, and Load (ETL) and 3.13 Interface Analysis.
- Week 1. Ridgeway's check-in records come from three door systems.
- Week 2. A vendor offers wearable data at a high price.
- Week 3. The data map has a name column that is too short.
- Week 4. Two teams need the same data for different uses.
- Week 5. A profile update in the Ridgeway app does something unexpected.
Question 6. Week 1: three door systems
Ridgeway's check-in records come from three different door systems. The team finds duplicate check-ins and timestamps that disagree by up to 40 minutes for the same visit. In the five Vs assessment, which V does this finding MOST concern?
- A. Volume
- B. Velocity
- C. Veracity
- D. Value
Answer: C.
Section 2.2.2 says veracity refers to the trustworthiness of the data, the uncertainties and inconsistencies it presents. Duplicate check-ins and timestamps that disagree are inconsistencies, so they make the check-in data less trustworthy for a question about how often new members visit before they cancel. A is wrong: volume is the amount of data produced and the size of the data sets. B is wrong: velocity is the speed at which data is generated and how often it must be collected. D is wrong: value is about driving the analytics from real, valuable business goals.
Question 7. Week 2: data that costs more than it helps
One research question asks if workout intensity predicts early cancellation. A vendor offers wearable data for it, but the licence costs more than the team expects the answer to be worth, and only 8% of members use a wearable. What should the analyst do?
- A. Buy the data, since more data sets always improve the quality of a model.
- B. Advise on cost versus benefit, and propose dropping that question if needed.
- C. Collect intensity data from scratch at every gym to avoid external bias.
- D. Ask the data scientist to estimate intensity from the check-in times.
Answer: B.
Section 2.2.2 says analysts support data scientists by analyzing the costs versus benefits of different data sets, from a cost, value, timing, risk and feasibility perspective. This matters most when the data must come from an external third party. Certain research questions may need to be dropped when the data required to answer them is too expensive. A is wrong: more data is not a reason to pay more than the answer is worth. C is wrong: the Guide calls collecting your own data ideal, but notes there are frequently not enough resources to do so. D is wrong: check-in times do not measure intensity, so it swaps one gap for a guess.
Read more: CBDA Source Data Questions: Collection Plans, Data Sets, Sampling
Question 8. Week 3: a column that is too short
In the data mapping specification, the source attribute Member_Name in Ridgeway's system is Varchar(80). The database administrator proposes a target attribute of Varchar(50) to save space. What should the analyst say?
- A. Accept it, since most member names are shorter than 50 characters.
- B. Make the target at least Varchar(80), since a smaller target can lose data.
- C. Keep Varchar(50) and trim longer names during the load.
- D. Leave the size open until the first test load shows which names fail to load.
Answer: B.
Section 3.6 says the data type and size of the target attribute are defined so they can accommodate the source attribute data. The target can match the source size or be larger to allow for future expansion, but it should never be smaller than the source, as that may result in missing data. Data mapping needs careful attention to detail to avoid data loss. A is wrong: most names fitting still means some names are cut. C is wrong: trimming is exactly the data loss the Guide warns about. D is wrong: attribute sizes are set in the map, before any data moves.
Read more: Data Mapping for CBDA: Guide Technique 3.6 Explained
Question 9. Week 4: one source, two uses
Two teams will use the loaded data. Operations wants a weekly cancellations dashboard. The data science team wants to retrain a cancellation prediction model every quarter. Which load approach fits these needs BEST?
- A. Incremental loads for the dashboard, full extractions for the model.
- B. Full extractions for the dashboard, incremental loads for the model.
- C. Incremental loads for both, since they are faster to run every time.
- D. Full extractions for both, since only full loads keep data accurate.
Answer: A.
Section 3.10 says some data warehouses allow only an incremental load, which checks the delta between existing data and adds the changed data. This allows faster generation of business intelligence reports, such as the weekly dashboard. A full extraction from the sources replaces the data on the target system and allows for better predictive and prescriptive analytics tasks, such as retraining the model. B is wrong: it swaps the strengths of the two load types. C is wrong: speed suits the dashboard, but the model is better served by a full extraction. D is wrong: the Guide does not call incremental loads inaccurate, and they suit BI reports.
Question 10. Week 5: the profile update
While studying how the Ridgeway member app saves data, the team finds that an address change does not update the member record. It creates a new record flagged as the most recent one, so a simple count shows some members twice. Which technique uncovered this, and where should the finding go?
- A. Interface analysis, with the finding documented in the data mapping.
- B. Data flow diagram, with the finding stored in the stakeholder map.
- C. Survey and questionnaire, with the finding added to the member survey.
- D. Business visualization, with the finding shown in a dashboard.
Answer: A.
Section 3.13 says that during data sourcing an interface can be analyzed to correlate interface elements to database attributes and discover underlying data or business rules. The Guide uses this very example: an update to a customer's profile may create a new record with an identifier for the most recent one. The findings are documented through data mapping, which helps the people responsible for business reporting. B is wrong: a stakeholder map holds people, not data rules. C is wrong: members cannot tell how the system stores their records. D is wrong: a dashboard would show the double count, not explain it.
Case 3. Saltmarsh Cinemas: does the loyalty app work?
Saltmarsh Cinemas runs 26 cinemas. Six months ago it launched a loyalty app, and 48,000 members have signed up. The marketing director wants another 400,000 dollars for the app's next phase.
The research question: did members visit more often after the app launch? Guide sections behind this case: 2.3.4 Perform Data Analysis, 3.9 Descriptive and Inferential Statistics and 3.12 Hypothesis Formulation and Testing.
- Week 1. The director says no test is needed.
- Week 2. The data scientist restates the question.
- Week 3. The team writes the null hypothesis.
- Week 4. A pilot offer ran in only 9 cinemas.
- Week 6. The result is in, and the director drafts a slide.
Question 11. Week 1: experience against evidence
The marketing director says: "I have run cinemas for 20 years, and I know the app works." The director adds that no test is needed. What is the BEST response?
- A. Accept the view, since experience-based decisions are reliable in a mature business.
- B. State the claim as a hypothesis and test it, so the intuition becomes measurable.
- C. Show the 48,000 sign-ups as proof that the app brings members back.
- D. Ask the director to rate their confidence from 1 to 10 and record it.
Answer: B.
Section 3.12 says hypothesis formulation and testing is used so business hypotheses are formulated in a rigorous manner to avoid purely empirical decisions. It transforms intuitive assessments into a verifiable and measurable assessment, and it is a scientific way to verify hunches, intuition and experience-based decisions. With 400,000 dollars at stake, the claim is worth testing. A is wrong: experience-based decisions are exactly what the technique is meant to verify. C is wrong: sign-ups show adoption, not a change in how often members visit. D is wrong: a confidence rating is still a subjective view, not evidence.
Question 12. Week 2: from business language to a model
Data quality issues were resolved during exploratory analysis. Using those results, the data scientist picks the method and restates "Did members visit more often after the launch?" as a comparison of mean monthly visits per member before and after the launch. Which task is this?
- A. Explore Data, testing the data set for quality issues.
- B. Perform Data Analysis, turning the question into a mathematical model.
- C. Prepare Data, cleaning and transforming the visit records.
- D. Formulate Research Questions, agreeing on the question with stakeholders.
Answer: B.
Section 2.3.4 says performing data analysis is the deep analysis done once data quality issues are resolved through exploratory analysis. It uses the results of the exploratory analysis to determine the best mathematical methods. The original research question in business language is transformed into a mathematical question, which is translated into a model. A is wrong: exploration tested the data set, and that work is already done. C is wrong: preparation gets the data ready, it does not restate the question. D is wrong: the question was agreed with stakeholders earlier, and the Guide points ahead to Section 2.3.4 for its mathematical restatement.
Read more: CBDA Analyze Data: Domain 3, All Five Tasks Explained
Question 13. Week 3: writing the null hypothesis
The team writes the hypotheses for the test. Which statement is the null hypothesis?
- A. Mean monthly visits per member are the same before and after the app launch.
- B. Mean monthly visits per member are higher after the app launch.
- C. The app caused a rise of at least 20% in monthly visits per member.
- D. Members who use the app are more loyal than members who never installed it.
Answer: A.
Section 3.12 says the null hypothesis, H0, is the default position of a claim that results in no change of the current state. The alternate hypothesis, H1, is the negation of the default position. Here no change means the app made no difference to visits, so H0 says the means are equal, and the test decides whether H0 can be rejected. B is wrong: this is the alternate hypothesis, the director's claim. C is wrong: it claims both a cause and a size, which the default position does not. D is wrong: loyalty is a different variable from visits and is not what the test checks.
Read more: CBDA Statistics Questions: Hypothesis Tests, P-Values, Regression
Question 14. Week 4: a test for a small sample
A new midweek offer in the app ran as a pilot in only 9 cinemas. The data scientist wants to check if mean weekly ticket sales in these 9 cinemas differ from a hypothesized value. Which test does the Guide point to for so few observations?
- A. A z-test, since it deduces the range a sample mean can take.
- B. A t-test, since it is used instead of a z-test for small samples.
- C. A chi square test, since it is mostly used to test sample variances.
- D. An F-test, since it compares population variances to sample variances.
Answer: B.
Section 3.9 says a z-test deduces what range an observed sample mean can take, and a t-test is used instead of a z-test when the number of observations in a sample is less. Nine cinemas is a small sample, so the t-test fits. The Guide adds that different tests apply based on the business situation. A is wrong: the z-test fits a question about a mean, but not with so few observations. C is wrong: the Guide says a chi square test is mostly used to test the variation of statistics such as sample variances. D is wrong: the F-test compares variances, and the question is about a mean.
Question 15. Week 6: the result and the slide
The test rejects the null hypothesis at a 95% confidence level: mean monthly visits rose after the launch. The director drafts a slide: "The app is proven to bring members back more often." What should the analyst recommend?
- A. Keep the slide, since rejecting the null hypothesis proves the claim.
- B. Report the rise with its 95% confidence level, not as a certainty.
- C. Remove all numbers from the slide, since statistics confuse stakeholders.
- D. Wait to report until the test reaches a 100% confidence level.
Answer: B.
Section 3.12 lists the limits of the technique. Hypothesis testing contains errors that are probabilistic in nature, so the results do not indicate certainty of outcome, and confidence intervals with appropriate significance levels must be included when stating the outcome of a test. A test shows whether a claim is statistically sound, but it does not replace conclusions from other analysis. A is wrong: a rejected null hypothesis still leaves a chance of error. C is wrong: the Guide advises simple examples to explain how the tests work, not hiding the results. D is wrong: no sample test removes all uncertainty, so the report would never come.
Read more: P-Value Meaning: What p = 0.03 Really Says on the CBDA Exam
What this case teaches
- Treat a leader's explanation as a hypothesis and check it in the current state. What the data shows, like a seasonal dip, can reshape the research question, and the future state needs measurable objectives that are updated and recorded as the vision grows.
- Judge data sets before you move them: veracity, cost versus benefit, a target attribute never smaller than its source, the right load type for each use, and data rules that only interface analysis reveals.
- Turn hunches into hypotheses. The null hypothesis is the no-change position, small samples call for a t-test, and every result is reported with its confidence level, never as proof.
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: November 2026.
FAQ
What is Weekly Breakdown?
A weekly series from certificate.tips. Every Monday we take one certification exam, build realistic cases and break down exam-style questions on them 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 2.1.3, 2.1.4 and 2.1.6 in case 1, Sections 2.2.2, 3.6, 3.10 and 3.13 in case 2, and Sections 2.3.4, 3.9 and 3.12 in case 3.
Sources
- IIBA: Guide to Business Data Analytics
- IIBA: Certification in Business Data Analytics (CBDA)
- IIBA: CBDA Certification Handbook
- NIST/SEMATECH e-Handbook of Statistical Methods: Two-Sample t-Test for Equal Means
- Wikipedia: Extract, transform, load, an overview of load strategies
The case and company are fictional. Not affiliated with Scrum.org or IIBA.