Data analyst interview questions — and the CV claims behind them
The SQL, stakeholder and judgement questions data analysts actually get, what each is really probing, and which line on your CV an interviewer will press hardest.
What this interview is actually testing
Whether you can turn a vague business question into a defined one — and whether the impact on your CV is impact your analysis actually caused.
The questions, and what each one is really asking
1.Tell me about an analysis that changed a decision.
What it probes: The whole job in one question. Analyses that changed nothing are the norm; the interviewer wants to know whether you can tell the difference and what you did to make one land.
2.A stakeholder asks why revenue dropped last week. Where do you start?
What it probes: Whether you decompose before you query. Analysts who open a notebook immediately are showing you how they will handle an ambiguous request.
3.How do you handle a request you think is the wrong question?
What it probes: Whether you have the standing and the tact to reframe. Answering the literal question badly and answering the real question rudely are both failure modes.
4.Walk me through a SQL query you're proud of.
What it probes: Less the SQL than what you were solving. Interviewers are checking whether complexity was necessary or self-inflicted.
5.Tell me about a time your analysis was wrong.
What it probes: Whether you have a process for catching yourself — sanity checks, reconciliation against a second source, someone who reviews your work before it reaches a director.
6.How do you present a result a stakeholder won't like?
What it probes: Whether you have delivered bad news to someone senior and survived it. This is often the real difference between an analyst and a senior analyst.
7.What's a metric your company measures badly?
What it probes: Whether you think about definitions at all. Almost every organisation has one, and noticing it is a strong signal.
The CV claim they'll press hardest
A business outcome attributed to your analysis — "identified £2M in savings", "drove a 15% uplift".
Analysts almost never own the lever. You found the thing; someone else decided to act, and someone else again implemented it. The follow-up is "what happened after you presented it?", and the honest answer is often that a version of it was actioned six months later in a different form. That answer is completely respectable. What is not survivable is quoting a figure from a business case you contributed one slide to as though you delivered it, because the interviewer will ask who owned the P&L and there is only one true answer.
Analyst interviews test one thing more than any other, and it is not technical: whether you can convert a vague business question into a defined one before you touch the data.
That is why so many of these interviews open with a deliberately underspecified prompt. “Revenue dropped last week — why?” contains no time comparison, no segment, no definition of revenue, and no indication of whether the drop is real or a reporting artefact. The candidate who starts describing joins has already answered the question being asked, which was never about SQL.
Rule out the boring explanations first
The strongest answer to a metric-change question starts with the unglamorous causes, in roughly this order: is the pipeline healthy, did a definition change, is the comparison window fair, is there a calendar effect.
Analysts who have been burned by a broken ETL job do this instinctively. Analysts who have only worked with clean teaching datasets skip straight to a business hypothesis, and it shows.
Only after the boring explanations are eliminated does decomposition become useful — volume versus price, new versus existing, one segment versus all. And only after that does a narrative belong in the room.
The attribution problem, which is worse here than anywhere
Analysts rarely own the lever they identify. You found the leak; a director decided whether to fix it; another team implemented some modified version of the fix two quarters later.
This puts an uncomfortable pressure on the CV. “Identified £2M in savings” is a defensible sentence. “Delivered £2M in savings” usually is not, and the interviewer will find the seam with one question: who owned that budget?
The honest version is more impressive than candidates expect. “I found it, I built the case, it went to the ops director, a reduced version was implemented the following year and I do not know the final realised number” tells an interviewer that you understand how decisions actually move through an organisation — which is the skill that separates an analyst who gets listened to from one who produces excellent dashboards nobody opens.
An AI-written CV bullet will not give you that version. Asked to make an analyst’s work sound impactful, a generator reaches for the strongest verb available and hands you a claim you then have to walk back in the room.
Preparing
Take your CV and, for each impact number, write down: what you personally produced, who decided, what was actually implemented, and what you know about the result. Where those four diverge, adjust the CV line rather than plan a defence.
Then pick two analyses — one that changed a decision, one that was wrong — and prepare both. The second is the question candidates fumble and interviewers remember.
Our free claim checker flags the lines most likely to draw a follow-up, and the STAR guide covers the answer shape.
Questions about the interview itself
- What questions are asked in a data analyst interview?
- Usually a technical portion (SQL, sometimes Python or a spreadsheet exercise), a case-style question about diagnosing a metric change, and behavioural questions about stakeholders and communication. The stakeholder half is weighted more heavily than candidates expect — most analyst roles fail on communication rather than on query-writing.
- How do I answer 'why did revenue drop?' in an interview?
- Do not start querying. Start by splitting the question: is it volume or price, new or existing customers, one segment or all of them, a real drop or a reporting artefact. Ask what the data pipeline looked like that week. Interviewers are watching for whether you rule out the boring explanations — a broken job, a changed definition, a bank holiday — before reaching for a business narrative.
- Do I need Python for a data analyst interview?
- SQL is close to universal; Python varies by team and is frequently listed on the job description but not tested. Read the posting for what the analysis stack actually is, and prepare in proportion. Claiming a language on your CV that you would not want to be tested on live is a poor trade — 'familiar with' is a phrase interviewers accept without any loss of credibility.