Business Analyst interview
Business Analyst interview practice, out loud
A BA screen is really three questions in a trench coat: can you form a hypothesis before you open a query editor, can you structure a messy business question into parts, and will you chase an insight until something actually changes. None of that shows up on a résumé — it shows up when someone asks a follow-up.
This rehearsal is a spoken interview built on a real business-analyst job description. It moves from an analysis that changed a decision, to a flat-revenue mystery, to SQL reasoning you narrate without a keyboard, to defending a number to a sceptic.
The report tells you whether you decomposed before you queried, whether you noticed the rows that quietly drop out of a join, and whether you sized the opportunity in money or in percentage points.
Business Analyst interview at a glance
- Rounds
- Often a data or aptitude test, then an analytical case, then a stakeholder round
- Length
- 45–60 minutes per interview; a written test is scheduled separately
- What decides it
- Decomposing before querying, and sizing the answer in money
- Tested more than syntax
- Joins, aggregation, and the rows that fall out of one
- Near-identical loops
- Business analyst, data analyst and operations analyst at the early-career level
- Asked of nearly everyone
- Explain a number to somebody who does not believe it
What actually happens in a Business Analyst interview
A written or online data test
Many employers screen with a spreadsheet or query exercise before an interview. It is usually mechanical rather than clever — a lookup, an aggregation, a percentage change — and it is timed. Ask whether the process includes one when you are shortlisted, because it is the only stage where you cannot think aloud your way through a gap.
An analysis you actually ran~10 min
What the question was, what you found, and what changed as a result. Interviewers use this to separate people who produce dashboards from people who produce decisions. A finding with no consequence attached to it invites the follow-up about who read it and what they did.
The vague business question~15 min
Something is flat or falling and you are given almost no context. The expected first move is to break the number into its components and pick a segmentation, before any mention of tables. Candidates who begin by describing which query they would write have skipped the part being assessed.
Data reasoning out loud~15 min
You are given two tables and a question, and you narrate the logic rather than typing. The interesting part is always the edge case: which rows the join drops, what a null does to your average, whether a period boundary excludes the very customers the question is about.
Defending a number to a sceptic~10 min
Somebody tells you your figure is wrong. What is scored is whether you find out what specifically they doubt before you re-explain, and whether you would actually change the number if their objection turns out to be right. Repeating the derivation louder is the common failure.
Sizing and prioritisation~10 min
Given a funnel with several drop-offs and limited time, which one do you work on. The answer is expected in currency rather than in percentage points, because the steepest percentage fall and the largest amount of lost revenue are frequently different steps.
What this interview assesses
Analytical Rigour
Do you decompose a question before querying, join and aggregate correctly, think about the rows that go missing, and sanity-check a result before you present it?
Business Judgement
Do you size opportunities in money rather than percentages, separate activity from output, and chase the largest addressable loss instead of the most visible one?
Stakeholder Communication
Can you explain a derivation to a sceptic, engage the objection instead of restating the number, and change your mind when the objection is right?
Sample Business Analyst interview questions
A feel for the kind of questions you’ll face. The real interview reacts to your answers with live follow-ups — these are examples, not the exact set.
1.Tell me about an analysis you ran that changed a decision. What did you find?
What lands: Anchor on a real question with a decision riding on it. State the finding with actual numbers and name what changed.
2.Revenue is flat this month, but the team is busier than ever. How do you find out what's going on?
What lands: Decompose revenue into volume, price and mix, segment by channel or cohort, and separate activity from output.
3.You've got an orders table and a customers table. How would you find the customers whose spend dropped the most?
What lands: Join on the right key, aggregate by period — and notice that customers with zero orders this quarter fall out of an inner join.
4.Explain a number to me that I don't believe.
What lands: Understand the specific objection first, walk the derivation step by step, and adjust if they turn out to be right.
5.One week to find where we're losing the most money in the funnel. How do you spend it?
What lands: Size each drop-off in rupees, not percentage points. The biggest gap and the biggest loss are rarely the same step.
The job description it’s built around
The free taster rehearses against this realistic Business Analyst posting. In a full rehearsal you can paste the exact job you’re targeting instead.
Read the sample job description
Business Analyst (0–3 yrs) · Fintech · Mumbai About the role We are a lending fintech scaling fast across India. Business Analysts here sit with the growth and operations teams, own the numbers, and turn them into actions: which channel to double down on, which process is leaking money, which cohort is going bad. What you'll do - Build and maintain dashboards for growth, risk and operations metrics - Pull and clean data with SQL and Excel/Google Sheets; automate recurring reports - Analyse funnels, cohorts and unit economics; find and size opportunities - Work with product and ops to design process fixes, then measure whether they worked - Prepare crisp weekly business reviews for leadership - Document requirements for data and product changes and follow them through What we're looking for - 0–3 years in analytics, consulting or operations - Strong SQL and advanced Excel/Sheets (pivots, lookups, basic modelling) - Structured problem-solving — you form hypotheses before opening a query editor - Clear communication with both business and technical teams - Bias for ownership: you chase an insight until something changes Nice to have - Python or a BI tool (Power BI/Tableau/Looker/Metabase) - Exposure to lending, payments or marketplace metrics
Business Analyst interview — the specifics worth knowing
- The data round tests joins, aggregation and edge cases rather than syntax, which is why it can be conducted out loud without a keyboard at all.
- How we know: Narrated data rounds are common precisely because the reasoning is the assessable part. Ask whether the round is written or spoken — both exist, and the preparation differs.
- The classic trap in a spend-decline question is that customers who ordered nothing this period disappear from an inner join, so the very people the question is about are missing from the result.
- How we know: This follows directly from how an inner join behaves on absent rows. You can confirm it on any two tables you have access to in a few minutes.
- Sizing an opportunity in money rather than in percentage points is the habit interviewers most often name as the difference between candidates, because the steepest drop is rarely the largest loss.
- How we know: A large percentage fall on a small base can be worth less than a small fall on a large one. Work an example with real volumes and the ordering usually changes.
- Business analyst, data analyst and operations analyst loops overlap almost entirely at the early-career level, so preparation transfers between them and the title alone should not decide which openings you apply to.
- How we know: Compare the postings side by side: decomposition, querying and stakeholder communication appear in all three. Read the responsibilities rather than the job title.
- A finding with no decision attached to it is treated as incomplete work. The follow-up is always who acted on it, and an answer that ends at the report is the answer the follow-up exists to expose.
- How we know: These postings describe analysis that supports decisions rather than reporting. Read what the role is expected to produce and who consumes it.
- When a stakeholder disputes your number, finding out what specifically they doubt scores higher than re-deriving it, because the objection is often about a definition rather than the arithmetic.
- How we know: Disagreements about a metric frequently turn out to be disagreements about who is being counted. Ask how the business defines the metric and compare it with how the query does.
- Where a timed written test exists it is usually mechanical rather than difficult, and it is the one stage where thinking aloud cannot cover a gap in tooling.
- How we know: Varies by employer. Ask the recruiter whether the process includes a written or online data exercise and what tool it is set in.
- Sanity-checking a result before presenting it is asked about directly, because a wrong number that reaches a decision-maker is more expensive than a slow one.
- How we know: Ask what review a recurring report goes through before circulation. A team that has been burned once will describe a specific check.
Common mistakes — and what to do instead
Answering a vague business question by naming the tables you would open.
Break the metric into its parts and choose a segmentation first, then say what you would pull to test the most likely component. The structure is the part being marked.
Writing the join without saying which rows it will drop.
Name the edge case out loud — the customers with no orders this period, the nulls, the boundary date — before you move on. That sentence is frequently the whole point of the question.
Ranking funnel problems by the size of the percentage drop.
Convert each step into the amount actually being lost and rank on that. A steep fall late in a small funnel can be worth less than a shallow one at the top.
Responding to a sceptical stakeholder by walking the whole derivation again.
Ask which part they think is wrong, address that, and change the number if they are right. Most of these disputes are about a definition, and re-deriving does not touch a definition.
Preparation checklist
- Ask whether the process includes a timed written or online data exercise, and in which tool
- Practise decomposing a flat metric into volume, price and mix before mentioning any table
- Say the dropped rows out loud whenever you describe a join
- Convert one funnel you know into money lost per step, and see whether the ranking changes
- Prepare one analysis that changed a decision, and know who acted on it
- Rehearse asking a sceptic what specifically they doubt before you re-explain
- Have a named sanity check you run before you show anybody a number
Business Analyst interview — FAQs
Do I need to write live SQL?
No — you narrate the logic out loud, the way a real analyst screen tests whether you'd write the right query. It's about the joins, the aggregation and the edge cases, not syntax.
Is this for data analysts too?
Yes. Business analyst, data analyst and operations-analyst screens overlap heavily at the 0–3 year level, and the questions here target that shared core of decomposition, SQL reasoning and stakeholder communication.
How much does industry matter?
The reasoning transfers across fintech, e-commerce and SaaS. The curated posting is a lending fintech; the full rehearsal lets you paste the exact JD you're targeting.
What will the feedback actually tell me?
Where your reasoning was sharp, where you jumped to a query without a hypothesis, and the single habit — often 'size it in money' — most likely to change your next interview.
What is actually assessed in a business analyst interview?
Whether you structure a question before you touch data, and whether the answer you produce is one a business could act on. The analytical round hands you something vague — a revenue line that is flat while the team is busier than ever — and the interviewer is watching whether you decompose it into volume, price and mix before proposing anything to query. The data round rarely tests syntax: it tests joins, aggregation and the rows that quietly disappear, which is why the classic prompt involves customers whose spend fell to zero and who therefore vanish from an inner join. The third thing measured is sizing, and it is the one candidates most often fail: an opportunity described in percentage points rather than in money is not yet an analysis, because the largest percentage drop and the largest amount of lost revenue are usually different steps.
Related interviews
How to rehearse this exact thing
The flat metric with almost no context
You get the vague version, the way it actually arrives, and the report shows whether you structured the question or jumped to the data.
The join where the answer is the missing rows
You narrate the logic aloud with no editor to catch you. The report flags whether you named the rows that fall out before moving on.
The stakeholder who does not believe your number
The objection arrives without saying which part is doubted. The feedback covers whether you asked, or simply explained it all over again.
Ready to rehearse for real?
Start a free five-minute Business Analyst interview now. You’ll get a spoken interview with live follow-ups and a feedback report that quotes your own answers back.