Data Scientist interview questions

The questions (10)

1. 01Walk me through an analysis where the data said something the team didn't want to hear.
2. 02How do you decide whether a metric movement is a signal or noise?
3. 03Tell me about a time your model was technically correct but business-wrong.
4. 04Design an experiment to test whether a new feature increases retention.
5. 05How would you detect that an A/B test is being polluted by interference?
6. 06Walk me through how you'd build the metrics tree for our product.
7. 07What's the most surprising thing you've learned from a dataset in the last six months?
8. 08How do you handle a stakeholder who doesn't understand confidence intervals?
9. 09Describe a time you stopped a launch based on data the team disagreed with.
10. 10Walk me through your last end-to-end project — data → model → recommendation → outcome.

How to prepare

• —Practice articulating the business question *before* the statistical question. 'They want to know if launch X drove churn' beats jumping to 'I'd run a chi-square'.
• —Have a clean answer for the experimentation pitfalls — peeking, novelty effects, network effects, segment heterogeneity. These come up in every product DS loop.
• —Show calibration. When asked for a quick estimate, give a range with a stated confidence. Pretending to point-estimate to two decimals reads as junior.
• —Prepare for the 'metrics tree' question for the company's actual product. Walking in with a sketched tree of how you'd measure their North Star is a near-instant signal.
• —Don't lead with model architecture. Senior DS interviewers want to hear about hypothesis, data, validation, decision — model choice is one slide of five.

Get answers tailored to your DS resume.

Generic prep helps. Tailored prep wins. JobJab pulls from your actual resume + the posting you're targeting and writes a strong, first-person answer to each question — in about 25 seconds per question.