Amazon Data Analyst Interview Questions (2026)

1. 1Online Assessment (SDE OA)60 minMedium

   Two timed coding problems plus a workplace-simulation and logic section; the main gate for freshers and India volume hiring.

2. 2Phone screen45 minMedium

   One coding problem plus 1-2 Leadership Principle STAR questions with an SDE.

3. 3Coding loop round60 minMedium

   DSA problem to working code, followed by assigned-LP behavioral questions in STAR format.

4. 4System design loop round60 minHard

   Design an Amazon-scale service with capacity math, plus LPs; low-level/OOD design substitutes for junior candidates.

5. 5Hiring Manager round45 minMedium

   Team fit, project deep dives, and Deliver Results/Bias for Action stories with the manager you would report to.

6. 6Bar Raiser60 minHard

   An interviewer from outside the team stress-tests LP stories and overall bar with the hardest cross-examination of the loop; holds veto.

1. Q1

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

   MediumStatistics & Experimentation RoundA/B TestingAmazon-specific

   Context: Context: Amazon wants to grow retail conversion while keeping delivery promises and inventory healthy.

   How to answer: A strong answer will define a clear, measurable hypothesis for the ranking/recommendation change, such as 'The new algorithm will increase customer conversion rate.' The primary metric should directly test this hypothesis, like 'Add-to-Cart Rate' or 'Purchase Rate per session.' Guardrail metrics are crucial to ensure no negative unintended consequences, such as 'Page Load Time' or 'Customer Service Contact Rate.' The randomization unit should be carefully chosen, typically 'User ID' to maintain consistency, and the launch decision rule should specify statistical significance thresholds and guardrail performance criteria for a go/no-go decision.

2. Q2

   For Prime, should randomization happen at customer, session, device, seller, or region level? Explain the tradeoffs

   MediumStatistics & Experimentation RoundA/B TestingAmazon-specific

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

   How to answer: Randomization for Prime should ideally happen at the customer level to ensure independent observations and avoid contamination, as Prime is a customer-centric subscription. Randomizing at a higher level like region might introduce too much variance and reduce sensitivity, while lower levels like session or device could lead to a single customer experiencing both control and treatment, invalidating results. Trade-offs involve statistical power, potential for spillover effects, and operational complexity. Customer-level randomization balances these factors best for a subscription service.

3. Q3

   Choose primary and guardrail metrics for a Buy Box experiment aimed at improving purchase conversion rate. What metrics would prevent a harmful launch?

   MediumStatistics & Experimentation RoundA/B TestingAmazon-specific

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

   How to answer: The primary metric for a Buy Box experiment focused on purchase conversion is 'Purchase Conversion Rate' (e.g., units ordered / unique visitors). Guardrail metrics should include 'Revenue per User' to ensure overall monetary value isn't negatively impacted, 'Add-to-Cart Rate' to monitor earlier funnel engagement, and 'Page Load Time' or 'Error Rate' to catch technical regressions. A harmful launch would be prevented by monitoring for significant negative impacts on these guardrail metrics, especially revenue, even if the primary metric shows a positive lift.

4. Q4

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

   MediumStatistics & Experimentation RoundA/B TestingAmazon-specific

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

   How to answer: To diagnose sample ratio mismatch (SRM), I would first check for obvious implementation errors in the random assignment logic, such as incorrect bucket sizes or a faulty random number generator. Next, I would perform a chi-squared goodness-of-fit test on the observed counts of users in treatment and control to statistically determine if the 52/48 split is significantly different from the expected 50/50. If the p-value is below a chosen significance level (e.g., 0.05), it indicates a statistically significant mismatch. Finally, I would investigate potential causes like user ID collisions, caching issues, or specific user segments being disproportionately assigned.

5. Q5

   The Alexa Shopping 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 TestingAmazon-specific

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

   How to answer: A strong candidate would first explain that peeking early can lead to false positives due to increased Type I error rates, especially when the experiment hasn't reached its predetermined sample size or duration. They would then discuss the importance of pre-setting experiment parameters like sample size, duration, and MDE based on power analysis to ensure statistical validity. The candidate should propose methods to address peeking, such as using sequential testing methodologies (e.g., Always Valid p-values, O'Brien-Fleming boundaries) if early stopping is truly desired and planned for. Finally, they would recommend continuing the experiment to its planned duration unless there's an extremely compelling reason (e.g., severe negative impact, critical bug) to stop early, emphasizing the risk of launching a change based on potentially spurious early results.

6. Q6

   A new Fulfillment by Amazon feature shows a large week-1 lift in purchase conversion rate, but the effect fades by week 4. What could explain this and how would you design the test duration?

   MediumStatistics & Experimentation RoundA/B TestingAmazon-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 a new feature, or selection bias if early adopters were more engaged. It could also be attributed to a 'honeymoon period' where the feature's benefits are short-lived or not sustainable. To design the test duration, one should consider the user's natural behavioral cycle (e.g., purchase frequency), the time needed for the novelty effect to wear off, and sufficient time to observe long-term impact and potential habit formation. A typical duration might be 4-8 weeks, but could extend to a full quarter for features with longer cycles or significant habit formation components.

7. Q7

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

   MediumStatistics & Experimentation RoundA/B TestingAmazon-specific

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

   How to answer: Network effects, or 'interference,' bias an A/B test when the treatment group's actions influence the control group's behavior, violating the stable unit treatment value assumption (SUTVA). In a marketplace, this could manifest as 'spillover effects' where increased demand from treatment users (e.g., faster delivery) strains supply for control users, or 'deflection' where treatment users leave, benefiting control users. This bias typically leads to an underestimation or overestimation of the true treatment effect, making the experiment results unreliable. To mitigate, one must choose an appropriate unit of randomization that isolates treatment and control, such as geographic clusters or distinct marketplace segments.

8. Q8

   purchase conversion rate is a low-frequency event for Prime. How would you set up an experiment with enough power without waiting too long?

   MediumStatistics & Experimentation RoundA/B TestingAmazon-specific

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

   How to answer: To address low-frequency events like Prime purchase conversion, one should consider using a proxy metric that is a higher-frequency leading indicator of the ultimate conversion. Examples include 'add to cart' or 'view product page' for Prime-eligible items. Alternatively, increasing the sample size or extending the experiment duration are direct but potentially costly methods. For Prime, focusing on a specific segment of users (e.g., new users, users with high intent signals) could also accelerate observation of the effect.

9. Q9

   Design a geo or region-level experiment for Buy Box. When is this better than user-level randomization, and what are the analytical downsides?

   MediumStatistics & Experimentation RoundA/B TestingAmazon-specific

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

   How to answer: A geo-level experiment for Buy Box involves randomizing entire geographic regions (e.g., cities, states, or zip codes) to either the treatment or control group. This approach is superior to user-level randomization when there are strong network effects, spillover effects, or when the feature inherently impacts a local marketplace, such as Buy Box pricing or availability. Analytical downsides include lower statistical power due to fewer experimental units, increased risk of confounding variables if regions are not well-matched, and challenges in detecting smaller effect sizes.

10. Q10

    The Sponsored Products experiment lifts purchase conversion rate overall, but only for new users and only in one category. How would you evaluate heterogeneous treatment effects?

    HardStatistics & Experimentation RoundA/B TestingAmazon-specific

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

    How to answer: To evaluate heterogeneous treatment effects, I would first segment the data by user type (new vs. existing) and product category to isolate the observed lift. Then, I would perform separate A/B test analyses within each segment, focusing on statistical significance and practical impact (e.g., confidence intervals for conversion rate lift). For segments showing no lift, I'd investigate potential suppressors or interactions, perhaps using regression with interaction terms. Finally, I would recommend a targeted rollout strategy based on the identified segments of positive impact.

11. Q11

    Treatment improves purchase conversion rate but worsens on-time delivery rate for Alexa Shopping. Walk through a launch recommendation

    HardStatistics & Experimentation RoundA/B TestingAmazon-specific

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

    How to answer: A strong recommendation balances the trade-offs between purchase conversion and on-time delivery. The candidate should propose a phased rollout, starting with a small segment to gather more data on the long-term impact of delayed deliveries on customer loyalty and returns. They should also suggest investigating the root cause of the delivery degradation and exploring mitigation strategies, such as improving logistics or setting more realistic delivery expectations. The final recommendation should be data-driven, considering customer lifetime value and potential brand damage from poor delivery experiences, not just immediate conversion gains.

12. Q12

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

    HardStatistics & Experimentation RoundA/B TestingAmazon-specific

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

    How to answer: For ramp-up, I'd start with a small percentage (e.g., 1-5%) of traffic, closely monitoring key operational metrics like latency, error rates, and order fulfillment success, not just A/B test metrics, to ensure system stability and detect any unforeseen bugs or performance regressions. Holdback would involve reserving a small, representative portion of the original control group (e.g., 1-2%) even after a full launch, allowing for long-term impact assessment, detecting novelty effects, or providing a baseline for future comparisons. Post-launch monitoring requires ongoing vigilance over a dashboard tracking primary A/B test metrics, guardrail metrics (e.g., customer service contacts, delivery speed), and operational health, with automated alerts for significant deviations to ensure sustained positive impact and quick issue resolution.

13. Q13

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

    HardStatistics & Experimentation RoundA/B TestingAmazon-specific

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

    How to answer: A strong candidate would first assess the nature and timing of the Buy Box tracking change, specifically if it was a bug fix or a definition change, and if it affected both control and treatment groups equally. They would analyze the pre-change data for both groups to establish a baseline and compare it with the post-change data, looking for a significant shift in trends or absolute values. If the change was symmetrical and a bug fix, results might be usable with a clear annotation. If asymmetrical or a definition change, the experiment would likely need to be restarted, or at minimum, the pre- and post-change data analyzed separately with strong caveats.

14. Q14

    Two overlapping experiments on Prime both affect gross merchandise sales. How would you detect and manage interaction effects?

    HardStatistics & Experimentation RoundA/B TestingAmazon-specific

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

    How to answer: Detecting interaction effects between two overlapping A/B tests on Prime, both impacting GMS, requires careful experimental design and statistical analysis. First, ensure proper randomization units to minimize contamination, ideally using a factorial design if feasible to explicitly test interactions. If not, segment users into groups exposed to A only, B only, A and B, and control, then compare GMS differences across these segments. Statistical methods like ANCOVA or regression analysis with interaction terms can quantify the effect, and if a significant interaction is found, subsequent decisions should account for the combined impact rather than individual test results.

15. Q15

    Amazon's Retail 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 CasesAmazon-specific

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

    How to answer: A strong business case would start by clarifying the problem (magnitude, scope, timing) and then structure the investigation into internal vs. external factors. Internally, focus on changes to Amazon's platform (website issues, pricing errors, policy changes, fulfillment disruptions) and seller performance (new seller policies, payment issues). Externally, investigate macroeconomic trends, competitor actions, and shifts in consumer behavior. Prioritize data analysis by looking at key metrics like traffic, conversion rate, average order value, and number of active sellers, segmenting by category, geography, and device to pinpoint the root cause.

Practice these with instant AI feedback in a live mock interview → Start a Amazon Data Analyst mock

Prepare 8-12 STAR stories mapped to Leadership Principles; expect a Bar Raiser; quantify impact

Indicative Data Analyst pay in India: ~₹6–22 LPA (role-level range, not a Amazon-specific figure).

Other roles at Amazon

Data Analyst interviews at other companies