Design an A/B test for a recommendation algorithm
Last updated: March 12, 2026
Quick Overview
Design an experiment to test the impact of a redesigned homepage. Include sample size calculation, metrics, and analysis plan.
Databricks
March 12, 202618
4
1,415 solved
Design an experiment to test the impact of a redesigned homepage. Include sample size calculation, metrics, and analysis plan.
Databricks values data-driven decision making. This Onsite question assesses whether you can design experiments, interpret results correctly, and avoid common statistical pitfalls like p-hacking or Simpson's paradox.
What the Interviewer Expects
- Set up the problem formally with proper notation
- Apply the correct statistical test with clear justification
- Interpret results with appropriate caveats and confidence levels
- Discuss practical significance vs statistical significance
- Identify potential confounders and how to address them
Key Topics to Cover
How to Approach This
- Define your hypotheses (H0 and H1) clearly before performing any test.
- Calculate required sample size BEFORE running an experiment, using power analysis.
- Remember the Central Limit Theorem: sample means become approximately normal with large n.
- Watch for Simpson's paradox. Always segment data by key dimensions.
- Distinguish between statistical significance and practical significance.
Possible Follow-up Questions
- How would you design a follow-up experiment based on these results?
- How would you handle multiple comparisons?
- What if the sample size is very small?
Sharpen Your Skills on Codemia
Practice similar problems with our interactive workspace, get AI feedback, and track your progress.
Browse Statistics QuestionsSample Answer
Problem Formulation
To evaluate the impact of a redesigned homepage on user engagement, we will conduct an A/B test. Let:
- A be the control group that sees the old homepage.
- B be the treatment group that sees ...
Solution Approach
To determine the required sample size, we will use the following approach:
- Estimate baseline CTR: Assume from previous data that the CTR for the old homepage, , is 0.10.
- **Decide on...
Submit Your Answer
