Behavioural questions for data roles, answered the way you’d say them.
10 questions for Data Analyst, Data Scientist and AI/ML Engineer roles, Easy to Hard, 3 of them reported by candidates at companies. Say your answer out loud, then open ours. Free, no sign-up.
Tell me about a time you explained a technical result to a non-technical manager
Choose a real example and tell it with STAR. In the action, show that you led with the answer and what it meant for their decision, gave the one number that mattered in their units (rupees, orders or days), said how sure you were in plain words, and kept the method for questions. In the result, say what they decided and how you checked they had understood.
Open a question to see the answer to say. Each links to its sources, and to its full explanation where we have one.
Walk me through a project on your CV
EasyBehavioural
Use the STAR structure in about two minutes: the situation and why it mattered, your task, the actions you took and the result, plus what you’d do differently. Spend most of the time on your actions: the data, the method and why you chose it over simpler options. Say “I” for your own work and “we” for the team’s, and claim only results you can show.
This is a structure, not a script: practise it with your own project until it sounds natural.
Connect three things: what drew you to the work, evidence that you’ve already acted on it, and why this role. Name a problem from your current field that data science & AI could help solve, the steps you’ve taken, such as projects, courses or analysis in your current job, and how your domain knowledge fits the role. Stay positive about your past work, and don’t make the answer about pay.
This is a structure, not a script: your own reasons and projects are what make it convincing.
What is the difference between a data analyst and a business analyst?
Easy
Both turn evidence into decisions, from different starting points. A data analyst starts from the data: querying it with SQL, cleaning it, analysing it in tools such as Python or Excel, and reporting the patterns and trends it shows. A business analyst starts from the business need: defining requirements and processes, and recommending changes that deliver value to stakeholders. Titles overlap in practice, so read each job description to see which skills it tests.
What is data science, and how is it different from data analysis?
Easy
Data analysis examines data to answer a defined question, often what happened and why, with tools such as SQL, spreadsheets and dashboards. Data science is broader: NIST defines it as extracting useful knowledge directly from data through discovery or hypothesis testing, and it adds statistics, machine learning and programming, in languages such as Python and R, to predict and automate as well as explain. Both start from clean data and a clear question.
Tell me about a time you explained a technical result to a non-technical manager
MediumBehavioural
Choose a real example and tell it with STAR. In the action, show that you led with the answer and what it meant for their decision, gave the one number that mattered in their units (rupees, orders or days), said how sure you were in plain words, and kept the method for questions. In the result, say what they decided and how you checked they had understood.
Translate jargon such as “p-value” or “AUC” into what it means for the decision, unless the manager uses the term first.
Pick a real, moderate mistake and own it, using STAR: what happened, how the error was found, how quickly you told the people relying on it, and how you fixed it. Spend most of the time on what you changed afterwards, such as reconciling totals against the source system, peer review or tests on your queries. Don’t blame others, and don’t offer a disguised strength such as working too hard.
This is a structure, not a script: use your own mistake, and keep every number you mention true.
How do you handle vague or ambiguous requirements?
MediumBehavioural
Clarify the decision behind the request before touching data: who will use the result, what they’ll do with it, and by when. Restate the requirement in writing with metric definitions, filters and the time period, get a yes, then share a rough first version early rather than a polished wrong one. In an interview, tell it as a STAR story about a real request you clarified.
A useful clarifying question: “If this number came back high, what would you do differently?”
Three teams want reports by Friday. How do you prioritise?
MediumBehavioural
Rank the requests by the decision each one feeds, its real deadline and the effort, and be open about the trade-off. Ask each requester what happens if theirs arrives next week, agree the order with your manager when priorities clash, and offer options such as a quick cut now and the full version later. Then tell everyone the plan and keep them updated, so nobody is surprised.
Tell it as a STAR story from your own work; this is a structure, not a script.
I’d cover five things. The problem and its users, and why GenAI and not rules or classic ML; then the architecture, from data, chunking and retrieval to the model, prompts and tools. Next, evaluation: my golden set, the metrics, and how the numbers moved with each change, followed by the failures, such as hallucinations, bad retrieval or injection, and what I did about them. I’d close with latency and cost per request, and what would change at ten times the users.
Every kind of company opens with this, so rehearse it out loud.
I’d keep it to about two minutes: the business problem and why it mattered, then the data, with its source, size, label and how I cleaned it. Next comes the baseline, my model and why it beat simpler options, followed by the metric, the reason for it, and how I validated without leakage. I’d finish with error analysis: where the model fails and what I’d try next.
