← Google Interview Insights

Google·Software Engineer·Technical Phone Screen·Senior

Senior
Apr 2026

Summary

Google data science or analytics interview, single question about experimental design. Pretty conceptual but it has real teeth if you haven't thought carefully about when A/B testing breaks down.

Questions Asked (1)

Q1

Describe a situation where an A/B test would not be appropriate. What type of test would you run instead, and why?

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

I started rattling off the obvious cases like network effects and small sample sizes, but then blanked a bit on what to actually propose as alternatives.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining when A/B tests are inappropriate, such as when randomization is impossible, ethical concerns exist, or the user base is too small. Then propose an alternative method like observational study, quasi-experiment, or qualitative research, and justify why it fits the scenario. Emphasize the trade-offs and how you would validate findings.

Pro tip: Mention that A/B tests require stable, independent user assignment and sufficient sample size; if these are violated, consider switchback tests or holdout groups. Also, highlight that sometimes the best 'test' is to not experiment but to use causal inference methods like difference-in-differences.

1. Identify when A/B testing is not suitable

Discuss scenarios like network effects, ethical concerns, low traffic, or when the change is irreversible. Explain why randomization or control groups are problematic.

2. Choose an alternative testing method

Select a method such as observational study, quasi-experiment (e.g., difference-in-differences), qualitative user research, or switchback testing. Justify based on the scenario.

3. Explain why the alternative is appropriate

Highlight how the alternative addresses the limitations of A/B testing, such as handling interference, ethical constraints, or small sample sizes.

4. Discuss validation and limitations

Acknowledge the trade-offs of the alternative method, such as potential confounding variables, and how you would mitigate them (e.g., using propensity score matching).

5. Conclude with a practical example

Provide a concrete example from your experience or a hypothetical scenario to illustrate your reasoning and demonstrate applied knowledge.

Key Points to Mention

  • Network effects and interference between users (e.g., social features)
  • Ethical or regulatory constraints (e.g., testing on vulnerable populations)
  • Small sample size or low traffic leading to insufficient power
  • Irreversible changes or high-risk modifications (e.g., pricing changes)
  • Alternative methods: switchback tests, difference-in-differences, observational studies, qualitative research
  • Trade-offs: confounding, causality vs. correlation, and need for robustness checks

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