Cohort Analysis
Also known as: Cohort Report
Cohort analysis groups users by a shared characteristic or start date, such as the week they first visited, and tracks how each group behaves over time. It reveals retention and value patterns that aggregate numbers hide.
Key Takeaways
- Cohort analysis groups users by a shared trait or start date, such as signup week, then tracks each group's behavior over time.
- It reveals retention and value patterns that blended, aggregate numbers hide by mixing old and new users together.
- Comparing cohorts shows whether newer users are more or less valuable than earlier ones after product or channel changes.
- A rising overall average can mask declining recent cohorts when strong older groups prop up the blended number.
- Retention curves and revenue-per-cohort are the clearest way to see whether changes actually improve long-term outcomes.
How It Works
Cohort analysis sorts users into groups defined by when they started or a trait they share, then follows each group across later time periods. Instead of one blended average, you get a grid showing how the January cohort, the February cohort, and so on behave in week one, week two, and beyond. This isolates the effect of changes because each cohort experienced the product at a different moment.
The method is central to measuring Customer Lifetime Value, since it shows how revenue accumulates per cohort over time rather than as a single mixed figure. It also sharpens tracking of a North Star Metric by revealing whether recent cohorts are healthier or weaker than past ones.
Because cohorts separate new from established users, they expose problems that aggregate Engagement Rate or blended Conversion Rate numbers can hide. If recent cohorts retain worse but the overall average still rises, the analysis flags the decline early instead of letting strong legacy cohorts disguise it.
Why It Matters
It shows whether newer users are more or less valuable than earlier ones, exposing the real effect of changes to product, targeting, or channels. It is the clearest way to see retention and lifetime value trends.
Example
A subscription app compares monthly signup cohorts. Overall active users keep climbing, which looks healthy. The cohort grid tells a different story: users who joined in the last three months drop off faster than earlier groups. A recent onboarding change is the suspect. The team fixes onboarding and watches whether newer cohorts start retaining like the older ones did.
Common Mistake
Reading blended averages instead of cohorts. A rising overall average can mask that recent cohorts are performing worse, because strong older cohorts prop up the number.
Frequently Asked Questions
What is a cohort in analytics?
A cohort is a group of users who share a starting point or characteristic, most often the time period when they first signed up or converted. Tracking that group over time shows how its behavior evolves.
How is cohort analysis different from a blended average?
A blended average mixes all users regardless of when they arrived, so trends get muddied. Cohort analysis keeps each group separate, so you can see whether recent users behave better or worse than earlier ones.
What can cohort analysis measure?
Common uses include retention curves, repeat purchase rates, and revenue or lifetime value per cohort. Any behavior you can track over time can be split by cohort to reveal patterns aggregates hide.