Cohort Analysis Is How You Tell Growth From Churn Wearing a Disguise
Total revenue rising tells you nothing about whether customers stay. Grouping them by when they arrived separates a healthy business from one refilling a leaking bucket.
The Problem With Totals
Aggregate revenue answers one question: is the company bigger than last quarter. It cannot answer whether the customers acquired last year are still there, spending more, or quietly leaving.
Those are different businesses with identical headline numbers. A company acquiring aggressively can grow total revenue every single quarter while every individual customer relationship deteriorates.
Blending all customers together lets new arrivals hide the behaviour of the ones already there. Growth becomes a way of not finding out.
What a Cohort Is
A cohort is a group of customers defined by when they started, usually the month they first paid. You then track that specific group forward, never mixing it with later arrivals.
The January group is followed through February, March, April and onward. Whatever happens to them is a property of that group, uncontaminated by whoever signed up afterwards.
Reading the Table
The standard presentation puts start month down the side and months since start across the top. Each cell shows what share of the original group is still active, or what revenue they still generate.
| Cohort | Month 1 | Month 3 | Month 6 | Month 12 |
|---|---|---|---|---|
| January | 100% | 71% | 58% | 49% |
| February | 100% | 73% | 61% | 52% |
| March | 100% | 82% | 74% | - |
Two things are read off this immediately. Reading across a row shows how a single group decays. Reading down a column compares groups at the same age, which is the only fair comparison available.
The Column Comparison Is the Point
Comparing March at month three against January at month three is comparing like with like. If later cohorts hold up better at the same age, the product or the customer targeting improved.
If later cohorts are worse, the company is acquiring lower quality customers, and that is usually the signature of a firm spending harder to sustain a growth rate.
This degradation is invisible in aggregate numbers for a long time, because the total keeps rising while the mix rots.
Revenue Cohorts Versus Logo Cohorts
Counting customers and counting their revenue give different pictures, and both matter.
A logo cohort tracks how many accounts remain. A revenue cohort tracks the money they generate, which can rise above the starting level if the survivors expand their spending.
A business can lose half its customers and still show revenue retention above one hundred percent, because the ones that stayed grew. Whether that is healthy depends entirely on whether the lost customers were ever the target.
Where the Curve Flattens
The most important feature of a cohort curve is not how steeply it falls but whether it flattens.
A curve that declines and then levels off has found a durable core of customers who genuinely need the product. A curve that keeps descending toward zero means there is no retained base at all, and the company must acquire forever just to stay level.
The flattening point, and the height it flattens at, tell you what fraction of acquisition spending is building something permanent.
How It Gets Manipulated
Definitions do the damage. Changing what counts as active, from logged in this month to logged in this quarter, lifts every retention number without anything improving.
Selecting the starting cohort matters too. Presenting only the best performing month, or starting the clock at month one after early departures have already happened, flatters the curve.
And a company can simply stop publishing cohorts when they turn. Disclosure that appears in one investor deck and vanishes from the next is informative on its own.
The Bottom Line
Cohort analysis follows one group of customers through time rather than blending everyone into a total, which is the only way to see whether a business retains what it acquires. Compare cohorts at equal age, watch whether later groups hold up better or worse, and look for the point where the curve flattens. A curve that never flattens describes a company that has to keep buying its own growth.