Business analysis for product teams

12 questions drawn from the parts of this series. Answers here count toward each post's own quiz.

1. Question 1 of 12·IntermediateA client says "build us a mobile app so the director can approve invoices". Which approach best fits the idea that "a requirement is a hypothesis"?

   Record it exactly as stated and start estimating, because the client knows their work bestReject the request because the client is describing a solution instead of a problemTreat "a mobile app makes approval faster" as a hypothesis, find the problem behind it and verify it (for example measure which step the waiting time is in) before writing codeQuickly build a mobile app and wait for client feedback to learn the real requirement
2. Question 2 of 12·IntermediateWhy add shadowing (sitting next to someone doing real work) instead of relying on interviews alone?

   Because interviews take more time than shadowingBecause interviewees usually lie on purpose to hide problemsBecause shadowing produces a complete list of requirements, so no other techniques are neededBecause when people recount their work they describe the ideal version, while doing it in front of you reveals workarounds, real waiting time and interruptions
3. Question 3 of 12·StarterIn the 3 Cs model, which part of a user story carries the most value, and why is a polished As a / I want / so that template not enough?

   Card, because the more detailed the template, the less misunderstanding between PO, dev and testerConversation, because real requirements surface when the PO/BA, dev and tester sit down together; the template is only a shellConfirmation, because long enough acceptance criteria remove the need to talkCard, because the card is the only thing kept once the story is in the sprint
4. Question 4 of 12·IntermediateA team wants to perfect "find an order" (search, filters, guest lookup) before building refunds. Why does the post say this is the wrong direction?

   Because finding orders is a backend concern and should not be a user-facing storyBecause the story is too small to estimate and should become a spikeBecause building features one by one is always slower than building in parallelBecause shipping one perfect vertical column while nobody can get a refund burns sprints with nothing for users; ship a thin horizontal slice (walking skeleton) across the whole journey
5. Question 5 of 12·IntermediateA process splits into two branches with a parallel (AND) gateway but merges them with an exclusive (XOR) gateway. What is the consequence?

   The process hangs forever because XOR waits for a branch that never arrivesNothing, since the two gateway types are equivalent when mergingEverything after the merge runs twice, for example two POs sent to the supplierOnly the faster branch executes and the other is cancelled
6. Question 6 of 12·IntermediateThe company blames slow director sign-off for the two-week purchase time and plans a mobile approval app. What does the as-is analysis show?

   The director's share is small; the larger waits come from the rework loop, accountant batching and forgotten forwarded emails, so the app would solve the wrong problemThe director is indeed the main bottleneck but an app would not help since he travels oftenEvery step is equally slow, so each person needs to work fasterThe ISO paper process is correct and the issue is only that staff do not follow it
7. Question 7 of 12·StarterIn Event Storming, what is the core difference between a command and a domain event?

   A command is created by users and an event by the system, but they are otherwise the sameAn event is an intent that can be rejected, while a command is a fact that already happenedA command is an intent that can be rejected, while an event is a fact that already happened and so is never rejectedA command is written in the past tense for readability, while an event is written as an imperative
8. Question 8 of 12·IntermediateA dev team wants to jump straight to design level because it is "close to the code", and they only own one module. What main risk does the post describe?

   Design-level sticky notes cannot be used without a color legendThey may design a beautiful model of the wrong business, missing flows such as corporate package visits that a big-picture session would have exposedDesign level only works with 12 or more participantsAggregates will be split too small because there are no pivotal events yet
9. Question 9 of 12·StarterIn a PRD's opening section, which statement is a real problem rather than a solution?

   "We need to send appointment reminders via Zalo""The no-show rate for online bookings is double that of phone bookings, and receptionists spend two hours every afternoon calling to remind patients""Launch the reminder feature before next quarter""The system must remind quickly and be friendly to patients"
10. Question 10 of 12·IntermediateWhy should a performance NFR be written with a percentile (like p95) rather than average response time?

    Because percentiles are easier to compute than averages when there are few requestsBecause averages cannot be measured on the server, only percentiles canBecause p95 is always lower than the average, so the SLA is easier to meetBecause averages hide the long tail: 95% of requests at 100 ms and 5% at 4 s average about 300 ms and look fine, yet one in twenty users waits 4 seconds
11. Question 11 of 12·IntermediateWhat is wrong with defining conversion as "orders this week divided by checkouts this week", and which fix does the post recommend?

    Nothing is wrong; just count by session so the numerator is always smaller than the denominatorThe numerator and denominator are different sets of people; anchor the time window to each person's own checkout timeStart the week on Sunday so the numerator and denominator match the marketing team's calendarJust exclude internal users, since they are what makes the numerator larger than the denominator
12. Question 12 of 12·IntermediateA dashboard uses occurred_at::date on a timestamptz column and the BI connection is set to UTC. What is the consequence for a Vietnamese shop?

    Orders from midnight to 7 AM Vietnam time land on the previous day, skewing daily charts systematicallyThe query fails because timestamptz cannot be cast to dateOnly orders during daylight saving time are shifted, since UTC has no DSTThe numbers stay correct because timestamptz always stores and shows the buyer's local time