InterviewWare
EN中文
Start trainingStart training

Role practice · the whole loop

Data scientist interview questions on experiments, metrics and models

Prepare for data scientist interviews on experiments, metrics, causal claims and model evaluation, and on explaining uncertainty to people who want a simple answer.

Practice this interviewSample questions

Free during early access. Practice runs in the app, which needs an account.

Prep sheetData scientist

  1. 01Experiment design
  2. 02Metric definition
  3. 03Causal inference
  4. 04Model evaluation

Next repAI interview coach

01Common prompts

Sample data scientist interview questions

These data scientist interview questions are practice prompts written or adapted by InterviewWare, not questions reported by any employer. They are grouped by round, with the signals an interviewer listens for and the follow-up that usually comes next.

“Practice” opens the question in the workspace once you sign in.

QuestionWhat the interviewer listens forLikely follow-upPractice
R1 · Background & motivation
Q01Early career Tell me about an analysis that changed a business decision.Problem framing · ownership · decision impactHow do you know the decision changed because of the analysis?
Q02Early career How do you validate an unfamiliar dataset?Data quality · lineage · missingnessHow would you detect selection bias?
R2 · Role depth & case
Q03Experienced An experiment increases clicks but decreases retention. What would you investigate?Metric hierarchy · cohorts · causalityHow would you distinguish novelty effects from durable harm?
Q04Experienced Design an experiment for a feature with network effects.Randomization · interference · guardrailsWhat is the appropriate unit of randomization?
Q05Experienced Daily active users dropped 8% overnight. How do you investigate?Data validation · segmentation · root causeHow do you rule out a logging or pipeline problem first?
Q06Senior / lead Define a north-star metric for a two-sided marketplace and the guardrail metrics around it.Metric design · trade-offs · incentivesHow could teams game this metric, and how would you notice?
Q07 How would you evaluate a classifier on an imbalanced dataset?Why accuracy misleads; precision, recall and precision-recall curves tied to the cost of each kind of error; a threshold chosen with the business.Fraud is 0.5% of transactions and every false alarm costs a support call. Where do you set the threshold?
Q08 A churn model scores very well offline but poorly after launch. What do you check first?Target leakage from features computed after the outcome, a train/test split that ignores time, and drift between training data and live data.How would you set up validation so this can’t happen again?
R3 · Judgment & collaboration
Q09Senior / lead A stakeholder wants a causal claim from observational data. How do you respond?Uncertainty · causal reasoning · communicationWhat can you responsibly recommend today?
Q10Senior / lead How would you explain a model failure to a nontechnical team?Clarity · limitations · next stepsWhich decision should change as a result?
Q11Senior / lead Your analysis contradicts what leadership expects to hear. How do you present it?Candor · evidence · framingWhat would you do if you were asked to rerun it until the result agrees?

02Try one question

Answer a data scientist interview question

Draft an answer to one of the questions above. It stays in this browser until you ask for feedback.

Practice question · Data scientist

An experiment increases clicks but decreases retention. What would you investigate?

0 words
Get feedback on this answer

Feedback needs a free account (one step, no card). Your draft is kept only in this browser and filled in after you sign up; nothing is sent until you submit it.

03Worked example

A new checkout button lifted conversion in a two-week test. Should we ship it?

Mid-level outline, answering a hiring manager, about two minutes. Illustrative sample written by InterviewWare, not a real candidate’s answer.

  1. 01

    Check the test

    First I’d confirm the traffic split matched the plan and that the metric definition didn’t change mid-test.

  2. 02

    Read the result

    Then I’d look at the confidence interval, not just the p-value, and whether the lift holds in the second week.

  3. 03

    Guardrails

    I’d check refunds and support tickets as guardrails, because a more prominent button could also cause accidental orders.

  4. 04

    Recommend

    If those hold, I’d ship to everyone and keep a small holdout for a month to confirm the effect lasts.

04Coaching notes

How to answer data scientist interview questions

What works

  1. T1Start with the decision the analysis must support, then choose the method.
  2. T2Check the data before the model: definitions, logging changes, missing values and who is excluded.
  3. T3For every experiment, fix the randomization unit, the primary metric and a guardrail before looking at results.
  4. T4Say what you can’t conclude as clearly as what you can.
How to practice with an AI mock interview

Common mistakes in data scientist interviews

  1. M1Stopping an A/B test the day a daily peek finally looks significant.
  2. M2Treating a correlation in observational data as a causal effect because a stakeholder needs one.
  3. M3Reporting accuracy on an imbalanced problem.
  4. M4Ignoring novelty effects and ending a test after a few days.
  5. M5Answering “what should we do?” with a lecture on methods.

05Tools that help

Tools for data scientist interview prep

  1. AI interview coach

    Polish the answer you give when someone asks for a causal claim the data can’t support.

    One answerFollow-up

  2. Technical interview lab

    Warm up on Python drills like top-K frequent or sliding window before a coding screen.

    JS · PythonPublic tests

  3. Interview question bank

    Browse the whole data track by round, including prompts not published here, with what each one assesses.

    Role tracksCompany prep

06Company prep

Company notes for data scientist interviews

Summarised from each company’s public hiring pages. The drills are InterviewWare practice prompts, not questions these companies use, and InterviewWare is not affiliated with any of them.

All company prep sources

More directions

Explore another direction

Early access · free

Log your first rep today.

Bring a resume or a job description — or just the role you want. The log starts with your first answer.

Start trainingLimits & pricing