← Google Interview Insights

Google·Product Manager·Onsite - Product Sense / Strategy·Senior

Senior
Jun 2026

Summary

Google PM interview, one question on experimentation that felt deceptively simple but had a lot of layers once I started talking through it.

Questions Asked (1)

Q1

How would you A/B test a new feature?

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

I started with the basics, hypothesis, control vs treatment, sample size, and it felt fine until they pushed on how I'd handle network effects or situations where users influence each other.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the feature's goal and the hypothesis you want to test, then outline a structured A/B testing plan covering design, execution, and analysis. Emphasize statistical rigor, guardrail metrics, and how you would translate results into a product decision.

Pro tip: Always define success metrics and guardrail metrics upfront, and consider the minimum detectable effect to ensure your test is adequately powered. Mention that you would run a pre-test sanity check (A/A test) to validate the experimentation setup.

1. Define Hypothesis and Metrics

Articulate a clear, testable hypothesis about how the feature will impact user behavior. Identify primary success metrics (e.g., conversion rate) and guardrail metrics (e.g., latency, error rates) to monitor for unintended consequences.

2. Design the Experiment

Determine the target population, randomization unit (e.g., user, session), and sample size based on desired statistical power. Decide on the control and treatment groups, and ensure the test duration accounts for novelty effects and weekly seasonality.

3. Execute and Monitor

Launch the experiment and monitor data quality, sample ratio mismatch, and guardrail metrics. Ensure that the feature is implemented correctly and that there are no technical issues that could invalidate results.

4. Analyze Results

Perform statistical analysis to determine if the observed differences are significant. Look beyond averages to understand heterogeneous treatment effects and segment-level impacts. Consider practical significance, not just statistical significance.

5. Decide and Iterate

Based on the results, make a data-driven decision: ship, iterate, or abandon the feature. Document learnings and consider follow-up experiments to refine the feature or test related hypotheses.

Key Points to Mention

  • Hypothesis-driven approach: Start with a clear hypothesis and tie it to business objectives.
  • Statistical power and sample size calculation to avoid false negatives.
  • Guardrail metrics to ensure the feature doesn't harm key user experiences.
  • Randomization and avoiding common pitfalls like sample ratio mismatch (SRM).
  • Segmentation analysis to uncover heterogeneous effects and actionable insights.
  • Practical significance vs. statistical significance in decision-making.

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