Cohort
A group of customers who started in the same period, tracked over time so you can see whether the product is getting better.
A cohort is a group of customers grouped by when they started — everyone who signed up in March, everyone who bought the second edition. You then follow each group forward separately instead of averaging them together.
This is the only way to answer the question that matters: is the thing improving? A blended average cannot tell you. If March's cohort keeps 40% at six months and September's keeps 55%, something you changed worked, and you can go and find out what. Blended into one number, both facts disappear into a line that drifts for reasons nobody can name.
Cohorts also expose the growth illusion. A business adding customers fast will show rising totals and improving averages even while every individual cohort decays faster than the last, because the newest, healthiest-looking members dominate the mix. When growth slows, the underlying decay arrives all at once and looks like a sudden collapse. It was never sudden.
Worked: the January cohort of 100 has 62 active at month three. April's cohort of 140 has 91 — 65%. May's has 58%. Total actives rose every month, so the dashboard looks fine, but the trend across cohorts turned in May, four months before the blended number will show it.
Also known as
- cohort analysis
- signup cohort
Relevant for
- Founders
- Cohorts are how you find out whether last quarter's changes worked; a blended average will tell you the business is fine for four months longer than it is.
- Creators
- Track subscribers by the month they arrived — the cohort that came from one viral post almost never behaves like the ones who found you slowly.