Data Analyst interview question
How do you run cohort analysis?
Define cohorts by a stable entry event and period, choose an outcome such as retention, align elapsed time since entry, calculate comparable rates, and segment only when sample sizes remain meaningful.
Direct answer
What a strong answer covers
How it works
A cohort groups users by when they did a defining thing, usually sign-up or first purchase, so that behaviour is compared at the same age rather than the same calendar date. The analysis fixes the entry event and period, chooses an outcome such as being active or making a repeat purchase, aligns each user’s timeline to weeks or months since entry, and computes the rate per cohort per period. The resulting triangle shows whether newer cohorts retain better than older ones, which a blended metric hides. Segmentation by channel or plan is valid only while each cell keeps enough users to be stable.
Example
Weekly sign-up cohorts from January to March were plotted for week-4 retention; blended retention looked flat at 30 percent, but the cohort view showed February cohorts at 38 percent after an onboarding change and March back at 28 percent after a pricing change, which the average had completely masked.
Common mistakes
- Defining the cohort by an event users can repeat, so the same user enters several cohorts.
- Comparing a cohort’s week 8 with another cohort’s week 2.
- Slicing into segments of a few dozen users and reading noise as trend.
A useful answer structure
Answer from your own work.
Do not memorise a script: use only examples you can explain honestly in follow-up questions.
Open the desktop workspace and rehearse a clear answer in your own words.