Last Updated on September 13, 2026 by Caesar Fikson
Quick answer
Customer churn analysis is working out who is leaving, when, and why — in that order. Most businesses stop at a monthly churn percentage, which is the least useful number in the whole exercise because it tells you nothing you can act on. The analysis that changes anything is cohort-based: group customers by when they joined, track each group separately, and find the point in the lifecycle where people leave. That point is almost always earlier than anyone expects, and it is where the fix belongs.
Churn guides are usually a formula, a list of reasons customers leave, and some advice about improving onboarding. All true, none of it connected to your actual data.
The gap is that a single churn rate hides everything interesting. It averages a customer who left in week two with one who left in year three, and those are entirely different problems with entirely different fixes.
Measure it properly first
Two numbers, and businesses routinely quote one while meaning the other.
Customer churn rate
customers lost in period ÷ customers at start of period
Revenue churn rate
revenue lost in period ÷ revenue at start of period
These diverge, and the gap is informative:
Customer churn HIGH, revenue churn LOW
→ you are losing small accounts. Often fine
Customer churn LOW, revenue churn HIGH
→ you are losing big accounts. Urgent
Net revenue retention above 100%
→ expansion from existing customers exceeds
what you lose. The strongest signal there isIf you track one number, track net revenue retention. It captures churn, downgrades and expansion in a single figure, and it is the number that tells you whether the business compounds or leaks.
Cohort analysis, which is the actual technique
Group customers by the month they joined. Track each group’s retention separately over time. Everything useful comes out of this and it can be done in a spreadsheet.
| Joined | Month 1 | Month 2 | Month 3 | Month 6 | Month 12 |
|---|---|---|---|---|---|
| January | 100% | 72% | 64% | 58% | 54% |
| February | 100% | 70% | 63% | 57% | — |
| March | 100% | 81% | 76% | 71% | — |
| April | 100% | 83% | 78% | — | — |
First: the big drop is between month one and month two, in every cohort. That is an onboarding and activation problem, not a product problem — people left before they could have formed a view of the product.
Second: March and April retain noticeably better. Something changed. Find out what — a pricing change, a different acquisition channel, an onboarding tweak — because whatever it was, it worked, and that is more valuable than any churn survey.
My take
Build the cohort table before you do anything else, and build it in a spreadsheet rather than buying an analytics tool. It takes an afternoon and it reframes the problem for most businesses I have seen — people arrive expecting a retention problem and discover an activation problem, or expecting a product problem and discover a channel problem. Buying software to answer a question you have not framed yet just gives you a prettier version of the same confusion.
Segment before you conclude
An overall churn rate is an average of populations that behave nothing alike. Cut it by:
- Acquisition channel. The most consistently revealing cut. Customers from a discount campaign churn very differently from referrals, and if your paid channel produces customers who leave in month two, your cost per acquisition is worse than you think.
- Plan or price point. Cheapest tiers usually churn most. Whether that matters depends on whether they were ever going to be profitable.
- Whether they reached the activation moment. Defined below, and the single strongest predictor you will find.
- Company size or customer type, if you serve more than one.
- Whether they ever contacted support. Counter-intuitively, customers who contact support and get a good answer frequently retain better than silent ones.
Finding your activation moment
The action that separates customers who stay from customers who leave. Not a vanity metric — a specific behaviour, usually early, that correlates strongly with retention.
- List candidate early actions. Completed setup, invited a colleague, imported data, used the core feature three times, connected an integration.
- For each, compare retention between customers who did it in their first fortnight and those who did not.
- Find the biggest gap. If people who invited a colleague retain at 80% and those who did not at 35%, you have found something worth acting on.
- Restructure onboarding around getting people there, and measure the share of new customers who reach it.
Watch out
Correlation is not causation, and this is where churn analysis goes wrong most expensively. Customers who invite a colleague may retain better because they were always going to be more committed, not because inviting caused it. Pushing everyone to invite may achieve nothing. The only way to know is to change onboarding for a random half of new customers and compare — otherwise you are optimising a symptom.
Why the reasons customers give are unreliable
Exit surveys are worth running and worth distrusting. People rationalise, they avoid awkwardness, and they pick whichever option is easiest.
- “Too expensive” usually means not valuable enough. Price complaints scale inversely with perceived value. Nobody cancels something they cannot work without because it costs a little.
- “Missing feature X” is sometimes literal and frequently a proxy for never having got the product working at all.
- “No longer needed” is the honest one and the one you can do least about.
- Silence is the most common answer, and silent churners are the majority. Whatever your survey says represents a self-selected minority.
Behavioural data beats stated reasons. What someone did in their last thirty days tells you more than what they typed into a dropdown on the way out.
Involuntary churn, which is free money
A meaningful share of cancellations in any subscription business are not decisions. They are expired cards, insufficient funds and failed renewals — customers who wanted to stay and were removed by a payment failure.
- Measure it separately. If you are not splitting voluntary from involuntary churn, part of your churn rate is a billing problem being treated as a product problem.
- Turn on smart retries in your payment processor. Retrying a failed charge on a better schedule recovers a real proportion at zero effort.
- Email before the card expires, not after it fails.
- Use card account updater services where your processor offers them.
This is the highest-return work in churn reduction and the least discussed, because it is plumbing rather than strategy.
An afternoon’s work that produces answers
- Export customers with signup date, cancellation date, plan, channel and revenue.
- Build the cohort retention table. Months since signup across, signup month down.
- Find where the biggest drop happens. That is your problem, and its position tells you which team owns it.
- Split involuntary from voluntary churn. Fix the billing half immediately; it needs no strategy.
- Segment by acquisition channel and check whether any channel is buying customers who leave.
- Test one activation hypothesis properly, with a control group.
My verdict
Churn analysis fails when it produces a number instead of a decision. A monthly percentage on a dashboard makes people anxious and tells them nothing about what to change.
Cohorts tell you when people leave, which tells you which part of the business owns the problem. Segmentation tells you who. Splitting involuntary churn tells you how much of it is not a churn problem at all. Together those take an afternoon in a spreadsheet and are worth more than any dashboard.
And do the payment retries first. It is unglamorous, it needs no strategy meeting, and it is the closest thing to free revenue in this entire exercise.
For keeping customer data in one place so this analysis is possible at all, see CRM software examples and all-in-one marketing software.
Frequently asked questions
What is customer churn analysis?
Working out who is leaving, when in their lifecycle, and why. Most businesses stop at a monthly churn percentage, which is the least actionable number available because it averages someone who left in week two with someone who left in year three. Useful analysis is cohort-based, grouping customers by when they joined.
How do I calculate churn rate?
Customer churn is customers lost in a period divided by customers at the start of that period. Revenue churn is revenue lost divided by revenue at the start. Track both, because the gap between them tells you whether you are losing small accounts or large ones — very different problems requiring different responses.
What is cohort retention analysis?
Grouping customers by the month they joined and tracking each group’s retention separately over subsequent months. It reveals two things a single churn rate hides: the point in the lifecycle where people actually leave, and whether recent cohorts retain better or worse than older ones, which shows whether something you changed worked.
What is an activation moment and how do I find it?
The specific early action that separates customers who stay from those who leave. List candidate early behaviours such as completing setup, inviting a colleague or importing data, then compare retention between customers who did each within their first fortnight and those who did not. The biggest gap is your candidate.
Why should I be careful with activation metrics?
Because correlation is not causation. Customers who invite a colleague may retain better because they were always more committed, not because the invitation caused retention. Pushing everyone to invite may then achieve nothing. Test it by changing onboarding for a random half of new customers and comparing, rather than optimising a symptom.
Are exit surveys worth running?
Worth running and worth distrusting. Too expensive usually means not valuable enough, since nobody cancels something they cannot work without over a small price. Missing feature X is often a proxy for never getting the product working at all. Most churners say nothing, so any survey reflects a self-selected minority.
What is involuntary churn?
Cancellations that were not decisions — expired cards, insufficient funds, failed renewals. Customers who wanted to stay and were removed by a payment failure. If you are not measuring it separately, part of your churn rate is a billing problem being addressed as a product problem, which wastes effort on both.
What is the quickest way to reduce churn?
Fix involuntary churn first. Turn on smart payment retries in your processor, email customers before their card expires rather than after it fails, and use card account updater services where available. It requires no strategy, recovers a real proportion of lost revenue, and is the closest thing to free money in churn work.
