How to spot a churning customer before they leave

By the time a customer has "churned," they didn't decide today. They decided weeks ago — quietly, without telling you. The uncomfortable truth is that the signals were almost always there. The problem is that nobody, and nothing, was watching for them.

What churn actually looks like

Churn is rarely a dramatic exit. For most local and repeat-purchase businesses it looks like a slow fade: visits stretch out, orders shrink, replies stop. No angry email, no cancellation — just a customer who was here every month and now isn't. Because there's no single moment, there's nothing to react to. That's why it goes unnoticed until the revenue is already gone.

The three signals that come before the exit

1. Cadence is slowing

The earliest and most reliable signal is timing. A customer who bought every 3 weeks is now at 6. Nothing looks wrong on any single day — the gap only means something against their own history. Miss the baseline and you miss the warning.

2. Engagement is dropping

Opens, clicks, redemptions, and replies taper off before purchases do. A customer disengaging from your messages is often weeks ahead of a customer disengaging from your business. It's a leading indicator hiding in data you already collect.

3. Silence after a bad experience

A late order, a mediocre visit, a support issue that dragged — followed by quiet — is one of the strongest churn predictors there is. The event plus the silence together mean far more than either alone.

Why people miss all three

None of this is a knowledge problem — any owner, shown one customer's history, could spot the drift. It's a scale problem. You cannot hold the individual cadence of a thousand customers in your head, and a dashboard of aggregate metrics averages the danger away. The customer about to leave is a rounding error until they're gone.

How a memory engine catches it in time

This is exactly what a customer memory engine is for. It learns each customer's baseline cadence, watches engagement trend against it, notices the silence after a bad moment — and surfaces the specific people at risk while there's still a two-week window to act. Then it can fire the win-back automatically: the right offer, on the channel they actually use, before they've decided for good.

You don't need to watch a thousand customers. You need something that never stops watching, and tells you the handful that need you today.