Your Best Customers Are Quietly Leaving: A Churn Early-Warning System
The customer who cancels next quarter is already going quiet this quarter. The five signals that predict churn, and how to catch the leak before renewal.
Want this inside your company?
Tell us the outcome you need, and we'll show you what we can build.
Problem: You find out a customer is leaving when they don't renew. By then the decision was made weeks ago, the budget got moved, and your "save" call is a formality. The account was going cold for a whole quarter and nobody on your team saw it.
Quick Win: Build a churn early-warning system, a running check that scores every customer on how likely they are to leave, so you get months of warning instead of a cancellation email. Churn just means customers who cancel or stop buying. The customer who leaves next quarter is already going quiet this quarter: fewer logins, slower replies, a main contact who changed jobs. Those signals are all in data you already hold. The rule we follow: catch the leak before renewal, not at it.
Want this inside your company?
Tell us the outcome you need, and we'll show you what we can build.
Churn Is a Lagging Event With Leading Signals
A cancellation is the last thing that happens, not the first. By the time a customer tells you they're leaving, the interesting part is over. They stopped using the product in March, their main contact left in April, their replies got shorter in May, and the "we've decided to go another direction" email lands in June looking like a bolt from the blue. It wasn't. It was the end of a story you could have read the whole time.
That is the entire premise of early warning: churn is a lagging event, but it has leading signals. A lagging signal tells you what already happened. A leading signal tells you what is about to. Renewal date is lagging. Declining usage is leading. If your retention plan starts at the renewal date, you're reading the last page first.
The good news is that the leading signals are cheap to watch, because you already own them. You don't need to buy anything. The data is sitting in your product logs, your email history, your customer database, and public news about the company. The work is not collecting it. The work is watching it on purpose, before the renewal, instead of scrambling after.
The Retention Math: Why Early Warning Pays for Itself
Before the how, the why. Keeping a customer is worth far more than the effort it takes. Frederick Reichheld of Bain & Company found that increasing customer retention rates by just 5% increases profits by 25% to 95%, depending on the industry (Harvard Business Review). That is not a rounding-error improvement. A small lift in the share of customers who stay swings profit by a quarter to nearly double.
The cost side is just as lopsided. Winning a new customer costs anywhere from five to 25 times more than keeping one you already have (Harvard Business Review). And the odds are better with the customer you have: the probability of selling to an existing customer is 60 to 70%, versus 5 to 20% for a new prospect (Marketing Metrics, via Forbes).
Put those together and the conclusion is blunt. A dollar of churn prevented is worth several dollars of new sales chased. Yet most companies spend almost all their energy winning new customers and almost none watching the back door. Early warning is the highest-return relationship work available because it protects revenue you already paid to earn.
The Five Silent Signals That Come Before a Cancellation
Here is what "going quiet" actually looks like, broken into the five signals that show up before almost every cancellation. Each one is visible in data you already have.
| Signal | What it looks like | Where it lives |
|---|---|---|
| 1. Usage goes quiet | Fewer logins, smaller or less frequent orders, a core feature they used weekly now untouched | Product logs, order history |
| 2. Replies slow down | Emails take days instead of hours, calls get rescheduled, review meetings quietly stop happening | Your inbox, calendar, meeting notes |
| 3. Your main contact leaves | Your main contact changes jobs, gets a new boss, or goes silent and a stranger appears on the thread | Email, LinkedIn, company news |
| 4. Support tickets change shape | Either a spike of frustrated tickets, or the opposite: total silence from an account that used to engage | Support inbox, ticket system |
| 5. Expansion stops | No new users added, no upgrades, and new questions about downgrading, pausing, or the fine print of the contract | Billing, sales notes, renewal terms |
A few of these deserve a closer look.
Usage going quiet is the earliest and most honest signal. People vote with their behavior long before they vote with a cancellation. A customer who logs in every day and then drops to once a fortnight has already half-left. This is the signal that gives you the most lead time, because usage falls off well before anyone writes an email about it.
Your main contact leaving is the most dangerous single signal, and the most measurable. When your main internal advocate (the person inside the company who understands your value and defends the invoice) leaves, there is a 51% chance that account cancels within the next 12 months, and roughly 65% of accounts that go through a senior-leader change do not renew (ChurnZero, citing Sturdy). The one person who defended your bill is gone, and their replacement inherits a line item they never chose. It's also the signal companies miss most, because it happens outside the product: on LinkedIn and in email footers, not in your usage dashboard.
Silence in support is easy to misread as happiness. An account that used to ask questions and now asks nothing has not always gotten comfortable. Sometimes they've stopped investing because they've mentally moved on. Silence is a signal, not the absence of one.
Turn the Signals Into a Customer Health Score
Five signals scattered across five systems is not a system. It's a to-do list nobody does. The point of a customer health score (a single number that rates how likely each account is to stay) is to collapse all of it into one figure per account, so your team acts on a ranked list instead of a hunch.
The mechanics are simple. Give each signal a weight based on how strongly it predicts leaving, add up the points for each account, and sort every customer into red, yellow, or green.
| Signal present | Points | Why the weight |
|---|---|---|
| Main contact left / senior-leader change | 3 | Strongest single predictor; 51% churn within 12 months |
| Usage down 30%+ vs their own baseline | 3 | Earliest honest signal, and hard to fake |
| Replies slowed / review meetings stopped | 2 | Relationship cooling, still recoverable |
| Support pattern flipped (spike or silence) | 2 | Frustration or disengagement, both matter |
| No expansion + downgrade/contract questions | 1 | Confirming signal, rarely the first mover |
Then read the total against a simple cutoff: 0 to 1 is green (leave it alone), 2 to 4 is yellow (watch it, reach out), 5 or more is red (run a save play now). The exact weights matter less than the discipline of scoring every account the same way, every week, against its own past behavior rather than a company-wide average. A law firm that logs in monthly is not unhealthy; a law firm that used to log in weekly and now logs in monthly is.
The output you want is not a dashboard nobody opens. It's a short weekly list: here are the seven accounts that turned red or yellow this week, here's why, here's who owns the save. That list is the whole product.
The Save Play: The 60 Days Before Renewal
A red account is not a lost account. It's an account you found in time. What you do next decides whether the early warning was worth building.
Speed is the multiplier. When customer success teams act on a senior-leader change within the first 48 hours, that customer is 33% more likely to renew (ChurnZero, citing Sturdy). The window closes fast because the replacement is forming their opinion right now, with or without you in the room.
A save play that works has four moves, and none of them is "offer a discount":
- Name the real reason. Match the red flag to a cause. Usage dropped because the person who used it left? Because a workflow changed? Because a competitor got introduced? Guessing wastes the window.
- Get to the new decision-maker fast. If your main contact left, your relationship is now with a stranger. Introduce yourself, learn what they inherited, and re-earn the value story from scratch. Do not assume the old goodwill transferred. It didn't.
- Show value they've forgotten they got. A quiet account has usually lost track of what your product did for them. A one-page recap of results delivered, in their own numbers, does more than any discount.
- Fix the specific thing, then confirm it's fixed. If the signal was a support spike, the save is resolving the underlying issue and closing the loop, not a friendly check-in that ignores it.
The discount is the lazy save, and it's often the wrong one. If the customer is leaving because your main contact is gone and the new one sees no value, a lower price just means they cancel later at a discount.
Where the Early-Warning System Breaks
This is not magic, and pretending it is gets you ignored. It fails in specific, predictable ways.
A red flag is sometimes just noise. Usage drops in August because half the company is on holiday, not because they're leaving. A single signal in isolation is a question, not a verdict. This is exactly why you score on multiple signals and against each account's own baseline. One flag is a maybe. Three flags pointing the same way is a pattern.
Some accounts are already gone. By the time three signals stack up and the renewal is two weeks out, the decision may be final. Early warning only works if it's actually early. A system that flags red accounts 14 days before renewal is a post-mortem, not a warning. The value is in the 60-to-90-day lead time, and a save play launched inside the last two weeks is mostly theater.
Watching everything recreates the problem. If every account throws off twenty metrics and every dip triggers an alert, you get a wall of noise nobody reads, and the real red accounts drown in it. The discipline is a small number of high-signal indicators scored consistently, not a firehose. More alerts is not more warning.
A score with no owner does nothing. The most common failure is building the health score and then leaving it in a dashboard. A flag that isn't assigned to a person with a save play is just a more organized way to watch customers leave.
Detection as a Running System, Not a Quarterly Scramble
Most companies "do retention" as a fire drill: a churn number spikes, a task force forms, everyone stares at spreadsheets for two weeks, the panic fades, and the same leak reopens next quarter. That's not a system. That's a reaction.
The version that works runs quietly in the background. Every week, the signals get pulled from the systems you already have, every account gets scored, and a short list of accounts that turned red or yellow lands in front of the person who owns them, with the reason attached. No scramble, no task force, no surprise at renewal. Just a standing queue of at-risk accounts, caught while there's still time to act.
This is the same discipline behind catching a slipped lead before it dies. If a slow follow-up quietly kills a new deal (the real cost of slow follow-up), a slow read on a cooling customer quietly kills renewal revenue, which is worth far more. The fix is the same shape: one source of truth for every relationship, watched automatically for the silence that means trouble. And the save play itself is exactly the kind of recurring, high-value work worth taking off your team's plate (your salespeople only sell 40% of the time).
Related Reading
- The real cost of slow follow-up, the same leak on the new-deal side of the business
- Your salespeople only sell 40% of the time, why the save work should run automatically
- Recover the follow-ups that slip through the cracks, one source of truth for every relationship
Frequently Asked Questions
How early can you actually catch churn?
For most companies, 60 to 90 days before the renewal, if you watch the leading signals instead of the renewal date. Usage decline is the earliest, showing up well before anyone writes an email. Your main contact leaving is the strongest, and it's public. The mistake is starting at the renewal date, which is the one moment when it's already too late to change the outcome.
Do we need a new tool to build a health score?
No. The five signals that predict leaving all live in systems you already have: product logs, your inbox and calendar, your support queue, your billing records, and public news about the company. The work isn't buying access. It's pulling those signals together, scoring every account the same way each week, and putting a short list of at-risk accounts in front of the person who owns them.
What's the single strongest churn signal?
The departure of your main contact. When that internal advocate leaves an account, there's a 51% chance it cancels within 12 months, and about 65% of accounts with a senior-leader change don't renew (ChurnZero, citing Sturdy). It's also the signal companies miss most, because it happens outside your product and shows up on LinkedIn, not your dashboard.
If your team finds out a customer is leaving at renewal, you're reading the last page first, and losing revenue that costs five to 25 times more to replace than to keep. We install the other version: a running early-warning system built from data you already hold, that scores every account, catches the silent ones months ahead, and hands your team a ranked queue of who to save and why, while the window is still open. See what we build for companies →
Want this inside your company?
Tell us the outcome you need, and we'll show you what we can build.
How to Reopen Dead Deals: A Dormant Pipeline Revival System
Most closed-lost B2B deals were never a no, just a not-yet. 40-60% die to no decision, not a competitor. How to find the ones worth reopening.
Stop Building Internal AI Tools, Buy the Outcome
Should you build internal AI tools yourself or buy the finished result? A minimal in-house team costs $400–600K a year to run, and about 70% of these projects never ship. When to build, when to buy the outcome, and the real hidden cost.