Data analyst interviews in the UK usually mix three things: a technical check (SQL, Excel and often a short take-home or live task), questions about how you turn numbers into a decision, and competency questions about working with people who don't speak data. The questions below are the ones that come up most, with what the interviewer is actually listening for.
What the interview process usually looks like
A screening call with a recruiter or the hiring manager (20–30 minutes): your background and why this role.
A technical stage: live SQL or Excel questions, or a take-home dataset with a short write-up.
A final interview: walk through your take-home or a past project, plus competency questions, often using the STAR format.
Technical: SQL, Excel and data quality
1. Explain the difference between an INNER JOIN and a LEFT JOIN, and when you'd use each.
What they're testing: Whether you understand what happens to rows that don't match — the source of most silent errors in reporting.
A strong answer covers:
An INNER JOIN keeps only rows that match on both sides; a LEFT JOIN keeps every row from the left table and fills unmatched columns with NULL.
A real case: customers with no orders disappear in an INNER JOIN, which can make an average look better than it is.
How you'd check: count rows before and after the join.
2. How would you find duplicate records in a table, and what would you do about them?
What they're testing: Practical SQL (GROUP BY … HAVING COUNT(*) > 1, or ROW_NUMBER()) and judgement — not every duplicate is an error.
A strong answer covers:
The query you'd write and which columns define a duplicate.
Asking why they exist (a double-loaded file, a genuine repeat purchase) before deleting anything.
Fixing the cause upstream, not just the symptom in your report.
3. What's the difference between WHERE and HAVING?
What they're testing: Whether you know the order SQL evaluates a query in.
A strong answer covers:
WHERE filters rows before grouping; HAVING filters groups after aggregation.
A one-line example: total sales by region, showing only regions over a threshold.
4. You're given a dataset with missing values. How do you decide what to do with them?
What they're testing: Judgement: whether you think about why data is missing before choosing a fix.
A strong answer covers:
Measure first: how much is missing, in which columns, for which groups.
Whether it's missing at random or systematically (e.g. one region never fills a field) — the second changes your conclusions.
Options and their trade-offs: exclude, fill with a sensible value, or flag it — and saying which you chose in the write-up.
5. Which Excel functions do you use most, and when would you move from Excel to SQL or Python?
What they're testing: Honest tool fluency and a sense of scale.
A strong answer covers:
The ones you really use (XLOOKUP or INDEX/MATCH, pivot tables, SUMIFS) with a quick example.
The point you'd switch: data too large, a process that needs repeating, or a need for an audit trail.
Analysis and judgement
6. Walk me through an analysis you did from the question to the recommendation.
What they're testing: Your whole method, and whether your work changed anything.
A strong answer covers:
The business question in one sentence and who asked it.
The data you used and one problem you hit with it.
What you found, what you recommended, and what happened next — with the real numbers if you can share them.
Example answer:
At my last placement the operations manager asked why late deliveries had risen over the quarter. I joined the delivery log to the depot and driver tables and found the rise wasn't spread evenly — almost all of it came from one depot after a route change in week six. I showed the trend before and after the change on one chart and recommended reviewing that route. They reverted part of it, and late deliveries from that depot came back to their previous level within a month. What I'd do differently is agree the definition of 'late' with the manager at the start — I lost a day because two teams measured it differently.
7. A dashboard shows sales dropped 20% last week. What do you do first?
What they're testing: Whether you check the data before explaining the business.
A strong answer covers:
Check it's real: a data load failure, a changed definition, a bank holiday, a filter.
Then break it down: which products, regions or channels account for the drop.
Only then go to stakeholders — with what you know and what you don't yet.
8. How do you explain a finding to someone who isn't technical?
What they're testing: Communication — the skill that separates analysts who get promoted.
A strong answer covers:
Lead with the answer and what it means for them, then the evidence.
One clear chart rather than a table of numbers.
Being upfront about uncertainty without burying the message.
9. Tell me about a time your analysis was wrong or challenged.
What they're testing: Honesty and how you handle being wrong.
A strong answer covers:
A real example — interviewers distrust answers where you were never wrong.
How you found out, what you did to correct it and who you told.
The check you now do so it doesn't happen again.
Working with people
10. Tell me about a time you had to push back on a stakeholder's request.
What they're testing: Whether you can say no constructively and protect the quality of the work.
A strong answer covers:
Situation and why the request was a problem (misleading metric, unrealistic deadline).
What you proposed instead and how you explained it.
The outcome and the relationship afterwards.
11. How do you prioritise when several people need reports from you at once?
What they're testing: Organisation and transparency.
A strong answer covers:
How you judge urgency against impact.
Making the queue visible and agreeing priorities with the people involved rather than deciding alone.
Automating the repeat requests so they stop competing for time.
12. How are you using AI tools in your work as an analyst?
What they're testing: Whether you use them sensibly — and check their output.
A strong answer covers:
A specific use: drafting SQL, explaining an unfamiliar query, summarising open-text survey answers.
How you verify what the tool produces before it reaches a report.
What you wouldn't use it for (sensitive data, final numbers without checking).
Questions to ask the interviewer
Which decisions does this team's analysis feed into most often?
What does the data stack look like, and where are the known data-quality problems?
How is a good first six months measured for this role?
Practise these out loud
Reading answers isn't the same as saying them under pressure. Interview Coach writes eight questions for the data analyst role you're applying for — from the job description and your CV — scores each answer with STAR feedback and gives a model answer to compare. Your first session is free.
Type “Data Analyst” as the role, or pick a saved job. AI feedback can be wrong — use it as practice, not a prediction of how an interviewer will score you.
Common questions
Do data analyst interviews include a technical test?
Most do: either live SQL or Excel questions, or a take-home dataset with a short write-up. Ask the recruiter which, and how long you'll have.
How should I prepare for a data analyst take-home task?
Answer the question that was asked, state your assumptions and data-quality checks, keep the write-up short with one or two clear charts, and finish with a recommendation.
Do I need Python for a data analyst job?
Not always. SQL and Excel are expected almost everywhere; Python or R is common in larger or more technical teams. Check the job description and be honest about your level.