Myntra Data Scientist Mock Interview: Rounds and Question Types

The process typically starts with a recruiter call and an online assessment or take-home covering statistics, probability and applied ML, sometimes with SQL. On-site or virtual rounds then usually span four areas:

• Statistics and probability: hypothesis testing, distributions, A/B design and pitfalls.
• ML case study: frame a Myntra-style problem end to end - recommendations, size prediction, churn or demand forecasting.
• SQL / data manipulation: window functions, joins, cohort and funnel queries.
• Product and behaviour: metrics, experimentation judgement, and past projects.

The signal Myntra wants is a scientist who can move from a fuzzy business question to a crisp problem statement, a defensible model choice, and an evaluation a product team would trust in production.

A common prompt is build a model to predict whether a returned item was due to a sizing issue or improve product recommendations on the home feed. Structure your answer carefully:

• Clarify the objective and the decision it drives - ranking, classification or forecasting.
• Define the label and features - behavioural signals, catalogue attributes, historical returns, seasonality.
• Start with a baseline, then justify a stronger model and its trade-offs.
• Pick an evaluation metric aligned to the business - precision at K for recommendations, calibrated probability for returns.
• Address leakage, cold start and the offline-online gap.

Weak answers jump straight to a fashionable algorithm. Strong answers spend time on the objective, the label, and how success will be measured in a live experiment.

Expect crisp statistics questions: explain p-values and confidence intervals without hand-waving, describe when to use a t-test versus a proportion test, and identify why an A/B test might show a false lift - peeking, novelty effect, sample-ratio mismatch, or interference between users. Myntra runs a high volume of experiments, so experimentation judgement is weighted heavily.

Be ready to design a test end to end: pick the metric, reason about the minimum detectable effect and duration at a conceptual level, choose the randomisation unit, and describe your guardrail metrics. Candidates who can say why a result is or is not trustworthy stand out from those who only know the mechanics of a significance test. This blend of SQL fluency, clean framing and business communication is what employers screen for across Data Analyst / Data Science Jobs, so the reps transfer well beyond a single firm.

SQL is a common screening and on-site component. Practise cohort retention, funnel conversion, and per-user aggregations using window functions like ROW_NUMBER, RANK and cumulative sums. Be able to write a query that computes, say, repeat-purchase rate by acquisition month, or the second purchase interval per user.

Narrate your logic as you build the query - interviewers score reasoning, not just the final statement. A strong candidate states the grain of each table, chooses joins deliberately, and checks for duplicate rows before trusting an aggregate. Rushing to a SELECT without thinking about the join fan-out is the classic way to produce a confident but wrong number.

One more thing Myntra interviewers value is product intuition layered on the modelling: a recommendation model that ignores freshness, margin or inventory can quietly hurt the business even with strong offline metrics. Be ready to discuss how you would balance relevance against business constraints, how you would set up guardrail metrics, and how you would monitor a model in production for concept drift once the season and the catalogue change underneath it. Then split your prep across the four pillars and finish with spoken reps:

• Statistics: revise hypothesis testing, the CLT, MLE intuition and A/B failure modes; explain each aloud in plain language.
• ML: rehearse three end-to-end case studies grounded in recommendation, forecasting and classification.
• SQL: solve ten to fifteen medium-hard queries focused on cohorts and windows.
• Behaviour: prepare project stories where your analysis changed a decision.

Then rehearse out loud. InterviewPrep's free AI voice mock interview generates a session from your CV and a Myntra Data Scientist job description, and scores your answers, pace and filler words so you can tighten how you explain trade-offs before the real loop.