What cohort analysis for startups means
Cohort analysis groups customers by when they joined, for example everyone who signed up in January, and tracks each group over time. It answers one question a single average cannot: are newer customers staying as long as older ones? A cohort is simply a batch of customers who started in the same period.
Founders usually watch total customers and a blended churn number. Both can look fine while the product quietly gets worse for new users, because a large base of loyal early customers masks the leavers. Cohorts separate the two stories.
How to build a cohort retention table
Put one row per signup month and one column per month since signup. Each cell shows how many customers from that row are still active, or what percentage of the starting group that is. Month 0 is always 100%, because everyone is active on the day they join.
Define active before you start counting. For a subscription product it can be a paid account; for an app it can be a customer who did something meaningful in the month. Pick one definition and keep it fixed, otherwise you will be comparing different things across cohorts.
Reading the retention curve
Plot one cohort as a line from month 0 onward. Almost every curve falls steeply at first, because trial users and poor-fit customers leave quickly. What matters is where it goes next: a curve that flattens into a plateau means a core group has found lasting value, while a curve that keeps sliding toward zero means the product is not yet sticky.
Take a January cohort of 200 customers. After one month 160 remain (80%), after two months 140 (70%), after three 128 (64%), after four 124 (62%) and after five 122 (61%). The drop slows from 20 points in month one to 1 point in month five. That flattening is the signal investors and founders look for when judging product-market fit.
80%
still active after month 1 (160 customers)
64%
still active after month 3 (128 customers)
61%
still active after month 5 (122 customers)
1 pt
drop between month 4 and month 5: the curve is flattening
The trap: blended retention hides decay
Now add an April cohort of 250 customers, a bigger group thanks to a new ad campaign. At month 3 only 125 of them remain, which is 50%. The January cohort was at 64% at the same age, so newer customers are leaving sooner.
Combine the two and the blended figure is (128 + 125) ÷ (200 + 250) = 253 ÷ 450, about 56%. That number looks acceptable and tells you nothing about the 14-point decline between cohorts. Only the side-by-side view shows that the extra acquisition spend bought lower-quality customers.
| January cohort | April cohort | |
|---|---|---|
| Starting customers | 200 | 250 |
| Active at month 3 | 128 | 125 |
| Month-3 retention | 64% | 50% |
| Revenue at $40/month | $5,120 | $5,000 |
| Blended retention | 56% (253 of 450), hides the gap | Cohorts show a 14-point decline |
What to compare across cohorts
Start with retention at the same age, such as month 1, month 3 and month 6, because comparing a young cohort to an old one is misleading. Then compare revenue per cohort. In the example, the April cohort has 50 more signups, yet at $40 a month it brings in $5,000 at month 3 against $5,120 for January. Growth in signups did not turn into growth in revenue.
Next, slice cohorts by acquisition channel, plan or customer segment. A paid-ads cohort and a referral cohort often have very different curves, and that difference tells you where the next dollar of marketing should and should not go.
How cohort retention connects to runway
Retention changes cash. If the April cohort had matched January's 64% at month 3, you would have kept 0.64 × 250 = 160 customers instead of 125, or 35 more. At $40 each, that is $1,400 of extra monthly revenue from one cohort, which lowers net burn and stretches runway without raising a cent.
It also changes how long it takes to earn back acquisition cost. Customers who leave early never repay what they cost to win, so a falling curve quietly lengthens payback. Read cohort retention next to CAC payback and net revenue retention rather than on its own.
Common cohort analysis mistakes
The first mistake is cohorts too small to read. With 15 customers in a month, one cancellation moves the line by almost 7 points, so group by quarter or combine channels until the numbers are steady. The second is changing the definition of active halfway through, which makes old and new cohorts incomparable.
The third is celebrating the first month. Early retention is easy to inflate with discounts and onboarding calls, and the real test is the plateau months later. Wait until a cohort has had time to flatten before drawing conclusions.
Practice retention decisions in Founder Runway
Retention is easier to understand when your own cash is on the line. Founder Runway runs for 20 turns with four options on every decision, and choices about product, pricing and growth change customers, MRR and runway together. You pick a B2B, B2C or B2G model and a starting stage of Pre-Seed, Seed, Pre-Series A or Series A.
A run ends as a failure, a promising company, EBITDA positive or high-value exit potential. Play one run that chases signups and one that protects the customers you already have, then compare where each ends.
Frequently asked questions
What is cohort analysis for startups?
Cohort analysis groups customers by when they joined and tracks each group over time. It shows whether newer customers stay as long as older ones, which a single blended average cannot. It is mainly used to measure retention and revenue per signup batch.
How do you read a retention curve?
Look at where the curve goes after the initial drop. If it flattens into a plateau, a core group of customers is staying for good. If it keeps sliding toward zero, the product is not yet sticky enough to support steady growth.
Why is blended retention misleading?
A blended number mixes old and new customers, so a large loyal base can hide worsening retention in recent cohorts. In the example, 64% and 50% at month 3 blend into about 56%, which hides a 14-point decline.
How many customers do you need for cohort analysis?
Enough that one cancellation does not swing the result. With about 15 customers in a cohort, a single churn moves retention by roughly 7 points. Group by quarter or combine channels until the lines are steady.
What is the difference between cohort retention and churn rate?
Churn rate is the share of customers lost in a period, usually across the whole base. Cohort retention follows one signup group over its lifetime. Cohorts show when and for which customers churn happens, so they point to the cause.
Test this decision in the game.
Apply the same assumption across one run and see which metric weakened three turns later.