How churn actually happens
Churn almost never arrives as a decision. It arrives as a slow narrowing that nobody inside the customer notices either.
The rollout stalls after the first team. One person becomes the only person who uses it. A process changes and the tool stops being part of it. The champion moves department. Then a renewal appears in front of a finance manager who sees a cost line and one active user, and asks a reasonable question.
The reframe
You are not looking for unhappy customers. Unhappy customers tell you. You are looking for customers who have quietly stopped depending on you, which looks like nothing at all until the renewal.
The signals, in the order they appear
They appear in a fairly consistent sequence, which is what makes them useful.
| Stage | Signal | What it means |
|---|---|---|
| Early | Use narrows to one person | The rollout stalled and nobody said so |
| Early | Questions stop entirely | They stopped trying to get more out of it |
| Early | Use narrows to one task | It has become a niche tool, not infrastructure |
| Middle | Champion changes role or leaves | Your relationship is now with nobody |
| Middle | A new senior person appears | Reviews are coming, and you are a cost line |
| Middle | They ask about exporting data | Someone is evaluating alternatives |
| Late | Invoice queries and payment delays | Cost is being examined |
| Late | Renewal date approaches with no conversation | The decision may already be made |
The most useful early signal is the shift from multiple users to one. It is a change in shape rather than a change in volume, so it does not show up in a usage total, and it happens months before anything looks wrong.
Questions stopping is under-rated too. A customer asking how to do something better is a customer still investing. When the questions stop and are not replaced by anything, the account has usually settled into a fixed, minimal use that will not survive a budget review.
Checking without a data team
You do not need scoring models. For a small book of accounts this is a twenty minute monthly review.
List accounts by renewal date, not alphabetically
Anything renewing in the next ninety days gets attention first. Everything else gets a glance.
Sorting by renewal is what turns this from a general review into a prioritised one.
Monthly20 minutesFor each, answer three questions from memory
How many people are actually using it? When did we last hear from them, and what about? Is the person who bought it still in the same role?
If you cannot answer any of the three for an account, that itself is the finding.
Three questionsCheck the two facts you can look up
Number of active users this month against three months ago, and the date of the last inbound message from them.
Two numbers. Resist building a dashboard before you have found out whether two numbers are enough, because for most small teams they are.
Two numbers onlyFlag anything with two or more signals
One signal is noise. Two together is a pattern worth a conversation.
Record the flag on the account so the next review can see whether it improved, which is the same discipline as recording loss reasons in [win/loss analysis](/blog/win-loss-analysis).
Two signals, not one
What to do when you see one
Reach out about the specific thing, not with a general check-in. "I noticed only Priya has been using it since your reorganisation, has something changed in how the team works?" gets an honest answer. "Just checking in to see how things are going" gets "all good", which is the answer that lets an account drift to renewal.
Ask what has changed on their side, not what is wrong with the product. Most narrowing comes from a change in their business: a process moved, a team restructured, a priority shifted. The product is often not the problem, and asking about it invites a polite non-answer.
When the champion has left, rebuild immediately. This is the highest-risk event in any account. Ask for an introduction to whoever picked up their responsibilities, and treat it as a new relationship: what do they need, what do they think this is for, what would make it useful to them. The same principles that keep a deal from being single-threaded apply to accounts, and for the same reason.
Fix the narrowing rather than the number. If use has collapsed to one task, the useful conversation is about the second task, not about their usage statistics. Customers do not care about their adoption metrics; they care about whether the next thing is worth their effort.
Be honest when it is not working. An account that genuinely does not need what you sell is better released cleanly than defended with a discount. They may come back, they will speak well of you, and you stop spending time on an account that will churn anyway.
Two structural points worth making. Almost all effective churn work happens before notice is given, which is why the monthly twenty minutes is worth more than any save playbook. And the best predictor of whether an account reaches this state at all is what happened in its first month, which our guide to onboarding a new customer in their first 30 days covers. A customer who reached real value in week one and spread to a second person is a different kind of account from one that never quite got started.
Frequently asked questions
- What is the earliest sign a customer will churn?
- Usually a narrowing of use rather than a drop in it. The account keeps working but only one person is doing it, and only for one task. That happens months before usage falls far enough to be obvious.
- Does losing a champion always mean churn?
- No, but it is the single highest-risk event in an account. If the person who bought and believed in it leaves and nobody else has a relationship with you, the next renewal is decided by someone with no history and a cost line to review.
- How often should you check for churn signals?
- Monthly for a small book of accounts, which for most small teams is a twenty minute review. Quarterly is too slow: a signal spotted in month one is solvable, and the same signal at renewal is not.
- Is a customer who complains more likely to churn?
- Less likely than a silent one, usually. Complaints mean they still expect it to work and are willing to spend effort. Silence more often means they have stopped expecting anything, which is much harder to reverse.
- Can you win back a customer who has already given notice?
- Sometimes, but the odds are poor and the discount required is usually high. Nearly all effective churn work happens in the months before notice, which is why the signals matter more than the save conversation.