For hiring managers
AI interview questions for data and analytics candidates
Most analysts now use AI to write SQL, draft cleaning code, and explain results. The interview has to find out how well. A query that runs can still count the wrong rows. A clear summary can still claim more than the data shows. Strong candidates know where those errors hide and check for them as a habit.
The twelve questions below cover the six skills Emeri measures, with two questions for each skill. Every question lists what a strong answer shows and what a weak answer shows. Two interviewers can then score the same answer the same way.
How to use these questions
- Ask for a real example
- Each question works best when the candidate describes work they have done. When an answer stays general, ask what they did the last time, and what happened next.
- Ask every candidate the same core questions
- Pick four to six questions before the interviews start. The same questions make answers comparable across candidates.
- Score the behavior, not the vocabulary
- A candidate who names many tools or uses the right terms has not shown the skill. Score against the strong and weak notes under each question.
- Pair the interview with a short work sample
- If the job uses AI, let the candidate use it in the sample. Ask for the requests they wrote and a note on what they changed. An interview shows how a candidate talks about the work. A sample shows the work.
Asking well
Asking for what you need clearly enough to get useful results. What this skill looks like for data analysts.
Question 1
A product manager asks how many customers use the new export feature. Before you ask an AI tool to write the query, what goes into your request?
Business questions arrive without definitions. This question shows whether the candidate turns a loose question into a request the model can answer without guessing.
A strong answer
Defines "use," such as at least one export event in the period. Says whether to count accounts or individual users. States the date range, the time zone, and which test or internal accounts to exclude. Asks the model to list the assumptions it made so each one can be checked.
A weak answer
Passes the product manager’s sentence to the tool as written. Or names the events table and nothing else, then corrects the query by hand when the count looks odd.
Question 2
A stakeholder asks whether the site redesign improved conversion. How would you use an AI tool to plan the analysis before anyone writes a query?
The first framing of an analysis question is often the wrong one. This question tests whether the candidate uses AI to test the approach, not only to produce code.
A strong answer
Asks the model for several ways to answer the question, such as a before-and-after comparison or a comparison against unchanged pages. Asks what else changed in the same period that could explain a shift. Chooses the method themself and can say why the others were weaker.
A weak answer
Asks the tool whether the redesign improved conversion and accepts the explanation it gives. Or goes straight to a before-and-after query with no thought about seasonality or other launches.
Giving good context
Giving AI the background, data, and limits it needs. What this skill looks like for data analysts.
Question 3
Our event data has known problems, such as one app version that logged some events twice for a month. How do you make sure an AI tool accounts for issues like that?
A model cannot know about data problems nobody has written down. This question tests whether the candidate supplies that knowledge as part of the request.
A strong answer
Includes the known issue in the request, with the affected dates and the rule for removing duplicates. Keeps a written list of data problems and table grains to reuse across requests. Checks the result for the affected period to confirm the rule was applied.
A weak answer
Expects the model to spot the duplicates on its own. Or remembers to fix the problem by hand afterward and sometimes forgets.
Question 4
You want AI help writing code to clean a messy customer file. What do you show the tool, and what do you hold back?
Cleaning code is only as good as the model’s picture of the data. This question tests whether the candidate shows the right detail without sending personal data.
A strong answer
Shares column names and types, and a sample of the real inconsistent values with names and emails replaced by fakes. States which values are valid, the target format, and what to do with values that fit no rule. Holds back the full file.
A weak answer
Describes the mess in general terms, such as "the dates are inconsistent," so the code misses real cases. Or uploads the whole file, customer contact details included, because that is faster.
Checking the work
Spotting mistakes, checking sources, and knowing what to trust. What this skill looks like for data analysts.
Question 5
An AI tool wrote a query for monthly revenue by region. It runs without errors. What do you check before you trust the result?
A query that runs is not a query that is correct. This question tests whether the candidate has concrete methods for finding silent errors.
A strong answer
Reads the logic before reading the output. Compares row counts before and after each join to catch duplicated rows. Checks date boundaries, filters, and rows dropped by an inner join. Reconciles the total against a trusted figure, such as the finance report or an existing dashboard.
A weak answer
Checks that the numbers fall in a plausible range and moves on. Or spot-checks one region and assumes the others are right.
Question 6
An AI-written summary of an A/B test says the new checkout flow "caused a clear lift in conversion." What do you review before that sentence reaches stakeholders?
AI summaries often state results with more certainty than the test supports. This question tests whether the candidate checks both the numbers and the claim.
A strong answer
Confirms the figures in the summary match the test output. Checks the sample size, the confidence interval, and whether the test ran for its planned length. Checks that each group received the expected share of traffic. Rewrites the sentence with the measured effect and its uncertainty.
A weak answer
Keeps the sentence because the test was randomized. Or checks that the headline number matches and does not look at the uncertainty around it.
Putting it together
Turning prompts, tools, and steps into a repeatable workflow. What this skill looks like for data analysts.
Question 7
In your analysis work, which calculations should happen in code and which can happen in a conversation with an AI tool? Give an example.
Numbers produced in a chat cannot be rerun or audited. This question tests whether the candidate keeps reported figures in reproducible code.
A strong answer
Says any figure that will be reported comes from a query or notebook that can be rerun. Uses AI to write, explain, and review that code. Uses the conversation for planning and wording, not for arithmetic on pasted tables. Gives a real example.
A weak answer
Pastes a result table into the chat and asks for growth rates or averages. Cannot say how they would reproduce a figure a month later.
Question 8
A teammate must rerun your AI-assisted analysis next quarter while you are away. What would you leave for them?
Analysis that only its author can rerun cannot be repeated when that person is away. This question tests whether the candidate builds work that other people can repeat.
A strong answer
Leaves the code in a shared repository, with the data sources and metric definitions written down. Includes the requests that produced key parts of the code. Adds checks inside the notebook, such as row counts and null checks, that stop the run when a result is wrong. Notes which steps need human judgment.
A weak answer
Leaves a notebook with the outputs and no notes. Or points the teammate to an old chat history and expects them to follow it.
Getting real work done
Using AI to finish real work, not just make a first draft. What this skill looks like for data analysts.
Question 9
Tell us about a stakeholder question you answered with AI help. What did you send back, and which decisions were yours?
Stakeholders need an answer they can act on, not a query result. This question tests whether the candidate owns the judgment behind the answer.
A strong answer
Describes a short answer that states the finding, its size, and how confident they are in it. Separates the parts clearly. AI drafted code or wording. The candidate chose the method, checked the numbers, and wrote the limits of the analysis.
A weak answer
Sent the result table or the AI-written explanation with few changes. Cannot separate their own decisions from the tool’s output.
Question 10
You need a rough estimate by tomorrow morning of how many customers a pricing change would affect. How would you use AI, and what would you tell the requester about the estimate?
Fast estimates are common analyst work. This question tests whether the candidate uses AI to move quickly while staying honest about precision.
A strong answer
Uses AI to draft the query and list the assumptions behind it. Checks the result against a known total, such as the current customer count by plan. Reports a range with the main assumptions stated. Labels it as an estimate and says what a fuller analysis would add.
A weak answer
Sends a single precise number with no assumptions attached. Or declines to give any figure until a full analysis is possible.
Keeping it safe
Protecting private information and using good judgment. What this skill looks like for data analysts.
Question 11
A query result you are working with includes customer names and order histories. You want an AI tool to help explain an odd pattern in it. How do you proceed?
Analysts handle personal data every day, and pasting results into a chat is an easy habit to form. This question tests whether the candidate keeps that data inside approved systems.
A strong answer
Does not paste the rows. Shares the schema and an aggregated or synthetic version of the pattern instead. Asks the tool for code to run inside the company environment. Follows the company’s approved-tool list for any task that needs real data.
A weak answer
Pastes the rows because the tool is for work use. Or deletes the name column and keeps email addresses and customer IDs, which still identify people.
Question 12
A dashboard you are building with AI help breaks results down by team, location, and role. Some groups have only three or four people. What do you do?
Aggregate data can still identify individuals when groups are small. This question tests whether the candidate sees that risk before the dashboard is published.
A strong answer
Sets a minimum group size and hides or merges groups below it. Checks whether combining two filters could reveal a single person. Raises the design with the privacy or HR owner. Explains the rule on the dashboard so readers know why some groups are missing.
A weak answer
Treats the data as anonymous because it holds no names. Or applies the AI tool’s suggested breakdown without asking who could be identified from it.
For what each skill looks like in this role at every band from E1 to E4, read AI skills for data analysts. These questions are written for interviews. They are not taken from the Emeri assessment.
Questions
What AI skills should I look for when hiring a data analyst?
Look for all six skills Emeri measures: asking well, giving good context, checking the work, putting it together into workflows, getting real work done, and keeping it safe. For analysts, give extra weight to checking the work and giving good context. An incorrect query result looks as finished as a correct one, so the candidate’s checking habits matter most.
Should data candidates be allowed to use AI in a take-home analysis?
If the job uses AI, yes. Ask for the code, the requests they wrote, and a short note on how they checked the numbers. A dataset with one documented quirk, such as duplicate rows, shows whether the candidate reads the documentation and checks the output.
Are these questions taken from the Emeri exam?
No. These questions are written for interviews and are never exam items. The Emeri assessment uses its own questions and scenarios, graded against the same six skills.
How do I confirm a candidate’s Emeri credential?
Every Emeri credential has a public verification page. It shows the band, the assessment date, and a profile across the six skills. The employers page explains how to read one. No account is needed.
Evidence beyond the interview
Emeri assesses the same six skills with graded work, and each credential has a public verification page. Read how to check a candidate’s credential, or see how teams use Emeri assessments.