← Amazon Interview Insights

Amazon·Data Scientist·Technical Phone Screen·Senior

Senior
Sep 2025Remote

Summary

Amazon DS technical screen, heavy on experimentation. The whole thing was one big A/B testing case study with real numbers to crunch, and they pushed hard on the statistical reasoning and pitfalls side rather than just asking you to run a t-test and call it a day.

Questions Asked (5)

Q1

Given 7 days of daily exposure and purchase data from a 50/50 checkout A/B test, compute overall conversion rates for both variants, the absolute and relative lift, and run a two-proportion z-test including p-value and 95% confidence interval.

A/B Testing & ExperimentationProduct Analytics & Metrics
Author's notes

This part I felt okay about.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by aggregating the 7 days of data to compute overall conversion rates for each variant, then calculate absolute and relative lift. Next, perform a two-proportion z-test to obtain the p-value and 95% confidence interval, ensuring you check assumptions and interpret results in the context of the experiment.

Pro tip: Always consider the practical significance alongside statistical significance; a small lift with a tiny p-value may not be worth implementing if the confidence interval includes negligible effects.

1. Aggregate Data

Sum exposures and purchases for each variant across all 7 days to get total counts. Compute conversion rates as purchases divided by exposures for control and treatment.

2. Calculate Lift

Compute absolute lift as the difference in conversion rates (treatment - control) and relative lift as (treatment rate / control rate - 1) * 100%.

3. Run Two-Proportion Z-Test

Calculate the pooled proportion, standard error, and z-statistic. Use the z-statistic to find the p-value for a two-tailed test.

4. Compute 95% Confidence Interval

Calculate the standard error for the difference in proportions and construct the 95% CI using the z-critical value (1.96).

5. Interpret Results

Assess statistical significance (p < 0.05) and practical significance (CI width and effect size). Discuss any limitations or next steps.

Key Points to Mention

  • Assumptions of the two-proportion z-test: independent observations, random assignment, and sufficient sample size (np and n(1-p) > 5).
  • Use of pooled proportion for standard error under the null hypothesis for the z-test, but unpooled for confidence interval.
  • Interpretation of p-value: probability of observing such extreme results if null hypothesis is true.
  • Confidence interval meaning: range of plausible values for the true difference in conversion rates.
  • Consideration of multiple testing or peeking if data is analyzed daily, which can inflate Type I error.
  • Practical significance: evaluate if the lift is meaningful for business metrics like revenue or customer experience.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q2

Check for sample ratio mismatch both day-by-day and in aggregate. If you find one, what could cause it and how would you fix it?

A/B Testing & ExperimentationRoot Cause Analysis
Author's notes

September 1st is where it gets ugly.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining what a sample ratio mismatch (SRM) is and why it matters, then describe how you would check for it both day-by-day and in aggregate using statistical tests like chi-square. If an SRM is found, systematically diagnose potential causes (e.g., instrumentation, randomization, filtering) and propose fixes, emphasizing the importance of resolving SRM before analyzing results.

Pro tip: Always check for SRM before looking at any metrics—if you find one, stop the analysis and treat it as a data quality issue. Document the investigation and communicate transparently with stakeholders, as ignoring SRM can lead to false conclusions.

1. Define SRM and its implications

Explain that SRM occurs when the observed ratio of users in experiment groups deviates from the intended ratio, indicating a flaw in the experiment setup or data collection. It can invalidate results, so it must be addressed before analysis.

2. Check for SRM day-by-day and in aggregate

Use a chi-square goodness-of-fit test to compare observed vs. expected group sizes each day and overall. Look for consistent deviations or sudden shifts that might indicate a problem.

3. Diagnose potential causes

Investigate common causes: instrumentation issues (e.g., logging errors), randomization bugs (e.g., biased assignment), filtering (e.g., bot removal, outlier exclusion), and data pipeline issues (e.g., delayed data).

4. Propose and implement fixes

Based on the cause, fix the root issue: correct the randomization algorithm, repair logging, adjust filters, or reprocess data. Then re-run the experiment or re-analyze if the fix is applied retroactively.

5. Validate and prevent recurrence

After fixing, re-check for SRM to ensure resolution. Implement monitoring and automated SRM checks in future experiments to catch issues early.

Key Points to Mention

  • Definition of SRM and why it's critical (invalidates A/B test results)
  • Statistical test: chi-square goodness-of-fit test for observed vs. expected ratios
  • Common causes: instrumentation/logging errors, randomization bugs, filtering (e.g., bot detection), data pipeline delays
  • Day-by-day vs. aggregate analysis: day-by-day can reveal temporal patterns (e.g., SRM only on certain days)
  • Fixes: correct assignment logic, fix logging, adjust filters, reprocess data, and re-run experiment
  • Prevention: automated SRM checks, monitoring dashboards, and pre-experiment validation

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q3

Is it appropriate to pool all 7 days together for analysis? Justify using a fixed-effects versus random-effects framing.

A/B Testing & ExperimentationTechnical Trade-offs
Author's notes

Blanked a little here.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

First, clarify that pooling all 7 days is appropriate only if day-level effects are uncorrelated with the treatment and other covariates; otherwise, it can bias estimates. Then, compare fixed-effects (FE) and random-effects (RE) models: FE controls for unobserved day-specific heterogeneity but may lose power, while RE assumes day effects are random and uncorrelated with predictors, allowing for more efficient estimates if assumptions hold. Use a Hausman test to decide between FE and RE, and consider the experimental design and goals.

Pro tip: Mention that in A/B testing, days are often not exchangeable due to day-of-week effects, so ignoring them can inflate false positives; but if you include day fixed effects, ensure you have enough within-day variation to estimate treatment effects precisely.

1. Clarify the goal and data structure

Determine if the analysis aims to estimate an overall treatment effect or day-specific effects, and note that data are panel-like with repeated observations per day.

2. Define fixed-effects and random-effects models

Explain that FE treats day effects as parameters to estimate (controlling for all time-invariant day characteristics), while RE treats day effects as random draws from a distribution, assuming they are uncorrelated with treatment.

3. Assess assumptions and trade-offs

Discuss whether day effects are likely correlated with treatment (e.g., if treatment assignment varies by day) and the bias-variance trade-off: FE is consistent but less efficient; RE is efficient but biased if assumptions fail.

4. Apply a test or diagnostic

Recommend the Hausman test to compare FE and RE estimates; if the null hypothesis of no correlation is rejected, FE is preferred.

5. Conclude and recommend

Based on the context (e.g., Amazon A/B test with random assignment), argue that pooling may be appropriate if days are exchangeable, but often include day fixed effects to control for day-of-week patterns, and check robustness.

Key Points to Mention

  • Day-of-week effects can confound treatment if not balanced across days.
  • Fixed-effects model controls for unobserved day-specific heterogeneity but uses up degrees of freedom.
  • Random-effects model assumes day effects are uncorrelated with treatment, allowing for more efficient estimates.
  • Hausman test can guide model choice by testing for correlation between day effects and regressors.
  • In A/B testing, randomization often ensures day effects are uncorrelated with treatment, but checking balance is crucial.
  • Pooling all days may be appropriate if no day-treatment interaction exists and day effects are not confounders.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q4

Identify at least three pitfalls in this experiment window and propose guardrails you'd put in place before launching.

A/B Testing & ExperimentationAdaptability & Ambiguity
Author's notes

The September 1st spike is screaming seasonality or a holiday effect, so that was the easy one.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the experiment's design and metrics, then systematically identify pitfalls across statistical, operational, and business dimensions. For each pitfall, propose a concrete guardrail that prevents or mitigates it, emphasizing Amazon's bias for action and customer obsession.

Pro tip: Frame guardrails as pre-launch checks that balance speed and rigor, and mention Amazon's 'two-way door' decisions to show you can move fast without compromising on learnings.

1. Clarify experiment context

Ask questions to understand the experiment's goal, metrics, duration, and population to tailor your pitfalls and guardrails.

2. Identify statistical pitfalls

Discuss issues like low power, peeking, multiple comparisons, and novelty effects that can invalidate results.

3. Identify operational and business pitfalls

Cover problems such as sample ratio mismatch, instrumentation errors, and cannibalization that affect execution and business impact.

4. Propose guardrails

For each pitfall, suggest specific guardrails like power analysis, sequential testing, A/A tests, and pre-registration.

5. Summarize and prioritize

Highlight the most critical pitfalls and guardrails, and explain how they enable faster, safer experimentation.

Key Points to Mention

  • Sample size and power analysis to avoid underpowered tests
  • Sequential testing or alpha spending to control false positives from peeking
  • A/A tests and sample ratio mismatch checks for instrumentation validity
  • Pre-registration of metrics and hypotheses to prevent p-hacking
  • Guardrail metrics to monitor for unintended consequences
  • Novelty and primacy effects and how to mitigate with longer run times or holdouts

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q5

Your MDE was a +5% relative lift. Given the data, estimate whether the test achieved sufficient power, and decide: roll out, iterate, or extend?

A/B Testing & ExperimentationProduct Strategy
Author's notes

This is where I got a bit hand-wavy.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

First, clarify the available data (sample size, variance, confidence interval, p-value) and compute the achieved power or minimum detectable effect (MDE) to assess sufficiency. Then, weigh statistical significance against practical significance and business impact to recommend rollout, iterate, or extend.

Pro tip: Always connect the statistical decision to business metrics (e.g., revenue impact) and consider the cost of false positives/negatives; this shows you think like a product owner, not just a statistician.

1. Clarify the Data and Metrics

Ask for or state the sample size, variance, confidence level, and observed effect size. Confirm that the +5% relative lift is statistically significant (e.g., p < 0.05) and note the confidence interval.

2. Assess Power and MDE

Calculate the achieved power given the sample size and observed effect, or compare the observed effect to the pre-computed MDE. If power is low (<80%) or the CI is wide, the test may be underpowered.

3. Evaluate Practical Significance

Determine if the lift is meaningful for the business (e.g., revenue, user engagement). Consider implementation costs, risks, and whether the effect is likely to persist.

4. Decide: Roll Out, Iterate, or Extend

If powered and practically significant, roll out. If underpowered but promising, extend the test. If powered but lift is small or negative, iterate on the feature or hypothesis.

Key Points to Mention

  • Statistical power (typically 80%) and its relationship to sample size and effect size
  • Minimum Detectable Effect (MDE) and how it's determined during test design
  • Confidence intervals and p-values to assess significance and precision
  • Practical significance vs. statistical significance (business impact)
  • Cost-benefit analysis of extending the test (time, resources, opportunity cost)
  • Potential novelty effects or seasonality that could affect results

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.