Flipkart Analytics Engineer Interview Questions (2026)

1. 1Online Coding Screen60 minMedium

   Timed online test or telephonic round with 2-3 DSA problems to qualify for the onsite loop; campus entry often via Flipkart GRiD.

2. 2Machine Coding Round90 minHard

   Build a small working application (e.g. parking lot, Snake & Ladder, splitwise) with clean object-oriented design, working demo, and extensibility within a strict time box.

3. 3PS/DS Round60 minHard

   Two hard problem-solving/data-structures questions where optimal complexity and bug-free code are expected.

4. 4System Design Round60 minHard

   HLD of an e-commerce-scale system such as flash-sale inventory, cart service, or order pipeline for Big Billion Days traffic.

5. 5Hiring Manager Round50 minMedium

   Project deep dives and situational behavioral questions assessing ownership, decision-making at scale, and team fit.

6. 6HR Round30 minEasy

   Compensation, level mapping, ESOPs, and joining logistics.

1. Q1

   Design an A/B test for a new Flipkart Marketplace ranking or recommendation change. Define hypothesis, primary metric, guardrails, randomization unit, and launch decision rule

   MediumStatistics & Experimentation RoundA/B TestingFlipkart-specific

   Context: Context: Flipkart wants to grow event-sale conversion without increasing cancellations and returns.

   How to answer: A strong answer will define a clear, testable hypothesis for the ranking/recommendation change, such as 'The new algorithm will increase user engagement (e.g., add-to-cart rate) without negatively impacting revenue or user experience.' The primary metric should directly reflect the hypothesis, like 'Add-to-Cart Rate per Session' or 'Conversion Rate.' Guardrail metrics are crucial for Flipkart's marketplace, including 'Average Order Value,' 'GMV per User,' 'Session Duration,' and 'Number of Complaints,' to detect unintended negative consequences. The randomization unit should be 'User ID' to ensure consistent experience, and the launch decision rule will involve statistical significance on the primary metric, ensuring guardrails are not violated, and considering a minimum detectable effect.

2. Q2

   For Big Billion Days, should randomization happen at customer, session, device, seller, or city tier level? Explain the tradeoffs

   MediumStatistics & Experimentation RoundA/B TestingFlipkart-specific

   Context: Consider cross-device behavior, interference, marketplace effects, and operational feasibility.

   How to answer: Randomization for Big Billion Days (BBD) should ideally happen at the customer level to ensure independent observations and avoid contamination, especially for long-term impact metrics. However, session or device level randomization might be considered for short-term, immediate interaction changes where customer-level identification is challenging or the experiment is very localized. Seller or city-tier randomization is generally unsuitable for BBD customer-facing experiments due to high variance, potential for network effects, and difficulty in isolating treatment effects, but could be relevant for specific seller- or city-focused initiatives. The choice depends on the experiment's goal, metrics, potential for spillover, and data availability.

3. Q3

   Choose primary and guardrail metrics for a Flipkart Plus experiment aimed at improving cart-to-order conversion. What metrics would prevent a harmful launch?

   MediumStatistics & Experimentation RoundA/B TestingFlipkart-specific

   Context: Include user experience, partner health, revenue, reliability, and long-term retention considerations.

   How to answer: For a Flipkart Plus cart-to-order conversion experiment, the primary metric should be 'Order Conversion Rate' (Orders / Carts Initiated) for Flipkart Plus users, measured as a relative lift. Key guardrail metrics include 'Average Order Value' (AOV) to ensure revenue isn't cannibalized by more frequent, smaller orders, and 'Return Rate' to prevent an increase in low-quality purchases. Additionally, 'Customer Lifetime Value' (CLTV) or 'Repeat Purchase Rate' for Plus users should be monitored to ensure long-term customer health isn't negatively impacted by short-term conversion gains. 'Page Load Time' and 'Error Rate' are also crucial technical guardrails.

4. Q4

   During a Fashion Store experiment, the treatment/control split is 52/48 instead of 50/50. How would you diagnose sample ratio mismatch?

   MediumStatistics & Experimentation RoundA/B TestingFlipkart-specific

   Context: Assume assignment logs, exposure logs, and eligibility filters may disagree.

   How to answer: To diagnose a 52/48 split instead of 50/50, first verify the randomization unit (user, session, etc.) and ensure consistent assignment logic. Next, check for implementation bugs in the bucketing mechanism, such as incorrect modulo operations or edge case handling. Analyze the traffic source and distribution over time to identify external factors or specific user segments disproportionately routed. Finally, perform a statistical significance test (e.g., chi-squared) on the observed split to confirm if the deviation is statistically significant or merely random chance.

5. Q5

   The Electronics Store experiment is trending positive after two days. A PM wants to stop early and launch. How do you handle peeking and sequential testing?

   MediumStatistics & Experimentation RoundA/B TestingFlipkart-specific

   Context: Discuss pre-specified stopping rules, alpha spending, business urgency, and risk.

   How to answer: Explain that early stopping due to 'peeking' (repeatedly checking results) invalidates statistical significance and inflates Type I error rates, leading to false positives. Discuss the need to pre-determine sample size and experiment duration based on desired MDE, power, and significance level. Mention sequential testing methods (e.g., AGILE, SPRT, or using O'Brien-Fleming boundaries) as a valid way to potentially stop early while maintaining statistical rigor, but emphasize that simple peeking is not sequential testing. Advise the PM against early stopping without a pre-planned sequential testing framework, recommending to let the experiment run its course or re-evaluate with proper statistical controls.

6. Q6

   A new Seller Hub feature shows a large week-1 lift in cart-to-order conversion, but the effect fades by week 4. What could explain this and how would you design the test duration?

   MediumStatistics & Experimentation RoundA/B TestingFlipkart-specific

   Context: Discuss novelty, learning effects, seasonality, and durable impact.

   How to answer: The fading effect could be due to novelty effect, where users initially engage more with the new feature due to its newness, but revert to old habits or find less utility over time. Another explanation is selection bias, if early adopters or more engaged sellers were disproportionately exposed or opted into the feature first. The test duration should be designed to capture the full user lifecycle and allow the novelty effect to subside, typically 4-8 weeks for a feature like this, or even longer if it impacts seasonal behavior or long-term seller habits. It's crucial to analyze metrics over time, not just week-1, and segment users to understand different adoption patterns.

7. Q7

   In a marketplace-like Flipkart Marketplace feature, treatment users may affect control users. How would network effects or interference bias the experiment?

   MediumStatistics & Experimentation RoundA/B TestingFlipkart-specific

   Context: Examples include seller supply, content inventory, delivery capacity, or pricing pressure.

   How to answer: Network effects in a marketplace like Flipkart mean treatment users' actions can influence control users (e.g., through inventory changes, pricing, or reviews). This interference can lead to an underestimation or overestimation of the true treatment effect, biasing the experiment results. Specifically, if treatment users reduce inventory, control users might see less availability, negatively impacting their experience and making the treatment look worse than it is. Conversely, if treatment users generate more reviews, control users might benefit from increased trust, making the treatment look better.

8. Q8

   cart-to-order conversion is a low-frequency event for Big Billion Days. How would you set up an experiment with enough power without waiting too long?

   MediumStatistics & Experimentation RoundA/B TestingFlipkart-specific

   Context: Discuss proxy metrics, variance reduction, larger samples, longer windows, and risk of metric gaming.

   How to answer: To set up an experiment with enough power for a low-frequency event like cart-to-order conversion during Big Billion Days without waiting too long, one should first consider increasing the sample size significantly by running the experiment across a larger user base or for a longer duration, if feasible. Second, identify and use a suitable proxy metric that is highly correlated with cart-to-order conversion but occurs with higher frequency earlier in the funnel (e.g., 'add-to-cart' or 'proceed-to-checkout' clicks). Third, utilize a sequential testing approach or Bayesian methods to allow for earlier stopping if a significant effect is observed, or if the probability of reaching significance becomes too low. Finally, ensure proper power analysis is conducted upfront, adjusting for the expected uplift and baseline conversion rate, and consider using a difference-in-differences approach if there's significant day-to-day variability.

9. Q9

   Design a geo or city tier-level experiment for Flipkart Plus. When is this better than user-level randomization, and what are the analytical downsides?

   MediumStatistics & Experimentation RoundA/B TestingFlipkart-specific

   Context: Use matched markets, pre-period balancing, spillover checks, and fewer experimental units.

   How to answer: A geo/city tier-level experiment for Flipkart Plus would involve randomizing entire cities or city tiers (e.g., Tier 1, Tier 2) into control and treatment groups, rather than individual users. This is superior to user-level randomization when there are significant network effects, spillover effects, or policy changes that are difficult to implement at a user level. Analytical downsides include lower statistical power due to fewer experimental units, increased variance, and potential for selection bias if city assignment isn't truly random or if cities are not comparable.

10. Q10

    The Fashion Store experiment lifts cart-to-order conversion overall, but only for new users and only in one category. How would you evaluate heterogeneous treatment effects?

    HardStatistics & Experimentation RoundA/B TestingFlipkart-specific

    Context: Balance pre-planned segments with exploratory slicing and multiple testing risk.

    How to answer: To evaluate heterogeneous treatment effects (HTE), first define relevant subgroups based on user tenure (new vs. existing) and product categories. Then, perform subgroup analysis by running separate t-tests or chi-squared tests for cart-to-order conversion within each defined segment. For more robust analysis, especially with multiple subgroups, consider using regression-based methods like interaction terms in a linear or logistic regression model, or Causal Forests for non-linear relationships. Finally, validate findings by checking for statistical significance, practical significance, and potential confounding factors, ensuring the observed effects are not due to chance or other variables.

11. Q11

    Treatment improves cart-to-order conversion but worsens promise breach rate for Electronics Store. Walk through a launch recommendation

    HardStatistics & Experimentation RoundA/B TestingFlipkart-specific

    Context: Make a decision under conflicting metrics and quantify tradeoffs for stakeholders.

    How to answer: A strong candidate would first identify the conflicting metrics and the need for a holistic view, likely through a weighted scoring model or a financial impact assessment. They would propose quantifying the monetary value of improved conversion versus the cost of increased breaches, considering factors like customer lifetime value and potential churn. The recommendation would then hinge on which impact is greater, potentially suggesting a phased rollout, a targeted launch (e.g., specific customer segments or product categories), or further experimentation to mitigate the breach rate while retaining conversion gains.

12. Q12

    How would you design ramp-up, holdback, and post-launch monitoring for a successful Seller Hub A/B test?

    HardStatistics & Experimentation RoundA/B TestingFlipkart-specific

    Context: Include ramp stages, persistent holdback, alert thresholds, rollback criteria, and owner accountability.

    How to answer: For ramp-up, I would start with a small percentage (e.g., 5-10%) of low-impact sellers, gradually increasing exposure while closely monitoring key operational metrics and system health. Holdback involves reserving a small, representative control group (e.g., 1-2%) from the experiment for a longer duration to assess long-term novelty effects and ensure no unforeseen negative impacts. Post-launch monitoring requires establishing a dashboard with key business metrics (e.g., seller NPS, GMV, listing creation rate, support tickets), system performance metrics (latency, error rates), and A/B test specific metrics, with automated alerts for significant deviations. This continuous monitoring helps detect regressions, seasonality impacts, and long-term treatment effects.

13. Q13

    Midway through the Flipkart Marketplace test, tracking for Flipkart Plus changed. How would you decide whether the experiment results are still usable?

    HardStatistics & Experimentation RoundA/B TestingFlipkart-specific

    Context: Compare instrumentation versions, affected traffic share, raw logs, and sensitivity analyses.

    How to answer: The primary concern is whether the change in Flipkart Plus tracking introduced a systematic bias or increased variance that disproportionately affected either the control or treatment group, or both. I would first identify the exact nature and timing of the tracking change relative to the experiment's start and the point of change. Then, I would analyze key metrics (e.g., Flipkart Plus enrollment rate, engagement) for both groups pre- and post-change, looking for significant shifts or differential impacts. If the change was minor, applied uniformly, and didn't alter core experiment metrics, the data might still be usable with caveats; otherwise, the experiment may need to be restarted or the affected period excluded.

14. Q14

    Two overlapping experiments on Big Billion Days both affect GMV after returns. How would you detect and manage interaction effects?

    HardStatistics & Experimentation RoundA/B TestingFlipkart-specific

    Context: Discuss experiment registry, factorial design, exclusion rules, and interaction terms.

    How to answer: A strong candidate would first identify the need for pre-analysis (e.g., historical data, business logic) to hypothesize potential interactions. They would then propose statistical methods like ANCOVA or regression analysis with interaction terms to detect significant interaction effects on GMV after returns. Mitigation strategies would include sequential testing, re-randomization, or segmenting the user base to isolate experiment groups. Finally, they would emphasize the importance of monitoring key metrics and having a rollback plan.

15. Q15

    Flipkart's Flipkart Marketplace revenue suddenly drops 10% week over week. Structure a business case to diagnose the issue and identify the most likely drivers

    MediumProduct Analytics & Business CaseBusiness CasesFlipkart-specific

    Context: Consider traffic, conversion, pricing, mix, supply/inventory, outages, marketing, and seasonality.

    How to answer: A strong candidate would structure their diagnosis by first clarifying the scope (Marketplace revenue vs. other segments) and the exact time period. They would then propose a top-down approach, starting with the core revenue formula (GMV x Commission Rate) and breaking down GMV into its components: (Number of Orders x Average Order Value) or (Number of Sellers x Number of Listings x Conversion Rate x AOV). The diagnosis would involve checking data integrity, then systematically analyzing each component for a 10% drop, segmenting by key dimensions like category, seller tier, customer segment, and region. Finally, they would prioritize potential root causes (e.g., policy changes, tech issues, competitor actions, seasonality) and suggest data-driven next steps for validation.

Practice these with instant AI feedback in a live mock interview → Start a Flipkart Analytics Engineer mock

Practice DSA + machine-coding rounds; prepare system design; know your projects

Indicative Analytics Engineer pay in India: ~₹9–40 LPA (role-level range, not a Flipkart-specific figure).

Other roles at Flipkart

Analytics Engineer interviews at other companies