What is CUPED, and why does it increase experimental power?
PICTURE THIS: DATA SPLIT
Fit on train, tune on val, report on test once.
Simple meaning
CUPED is a variance-reduction technique that adjusts the outcome using a pre-experiment covariate, often the same metric before launch.
WHY — A/B Testing instead of guessing?
Why interviewers care about A/B Testing:
people who only read docs from people who shipped.
and tied to Data Science work.
Name the idea, why it exists, then one short example.
End with when you use it and one common pitfall.
STEPS — What happens step by step?
Before you speak the answer, walk the interviewer through these steps:
- 1CUPED is a variance-reduction
technique that adjusts the outcome using a pre-experiment covariate, often the same metric before launch.
- 2By subtracting the part
of noise explained by the covariate, you shrink residual variance and need fewer users for the same MDE.
- 3The covariate must not
be affected by treatment, so it should be measured before assignment.
- 4Give an example
One tiny concrete case you can say aloud.
- 5Common mistake
What juniors usually get wrong.
- 6Close
When you pick this over the alternative.
EXAMPLE — See it in action
Here's a short line you can speak, broken into clear beats:
Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.
Key takeaway
CUPED is a variance-reduction technique that adjusts the outcome using a pre-experiment covariate, often the same metric before launch. By subtracting the part of noise explained by the covariate, you shrink residual variance and need fewer users for the same MDE.