Scoring the Accounts That Are Quietly Leaving

By the time a subscriber fills in an exit survey, the useful moment has passed. The survey explains a departure that finished weeks earlier, usually in language too polite to act on. The signals that actually predicted it were emitted earlier, most of them by the billing system rather than the product: a card replaced twice in a quarter, a renewal that needed three attempts, a plan quietly downgraded, a payment method swapped days before the charge. None of these is churn. Together, in the wrong combination, they are a forecast.

A risk scorecard is simply the discipline of writing those signals down, weighting them honestly, and reviewing the top of the list while there is still something to do. The temptation to resist is building a model. Small subscription businesses do not need machine learning to notice that an account with a failed renewal, a support complaint about billing, and a shrinking plan is in trouble. They need the three facts on one list, in one meeting, before the next renewal rather than after the cancellation email.

The practical version in scoring churn risk from billing signals before the exit survey lays out twenty warning signs with rough weights, and its most valuable instruction is about calibration: score your already-lost accounts first. Run last quarter's cancellations through the card and see where they would have ranked while they were still customers. A scorecard that would not have flagged the people you actually lost is a horoscope, however tidy its spreadsheet looks. Adjust the weights until the past makes sense; only then trust it about the future.

Billing signals deserve their weight for a specific reason: they are involuntary in both directions. Product usage can dip because of a holiday, a busy month, a finished project. A deteriorating payment pattern rarely has an innocent explanation that lasts. One failed renewal rescued quickly is an incident. A second failure two cycles later, on a different card, is a household budget quietly reprioritising you — or a customer rehearsing departure without having decided anything consciously yet. Either way, it is the earliest honest signal most businesses ever get.

What to do with a flagged account matters more than the flag. The scorecard fails if its only output is anxiety. Each band needs an action proportionate to it: a quiet payment-method health check at the low end, a human note at the middle, a genuine save conversation — with a real offer and a real ear — at the top. What it must never produce is a discount blast to everyone above a threshold; that teaches healthy customers to perform risk in exchange for a lower price, and corrupts the signal for everyone.

Review cadence is the unglamorous part that decides whether any of this works. Twenty minutes a week, same accounts list, same question: who entered the top band, who left it, and what did we do for the ones who stayed? Over a quarter, that habit builds the thing no template can give you — a felt sense of what departure looks like in your specific business, arriving early enough to be useful. The scorecard is scaffolding. The attention is the building.

Start smaller than twenty signals if you must. Five honest ones, reviewed weekly, will outperform twenty admired ones reviewed never. The accounts you are going to lose next quarter are already telling you, in the only language the billing system speaks. The scorecard is just how you agree, as a team, to listen on schedule.

Swipe Again (swipeagain.com) is flat-fee failed payment recovery for subscription businesses — $179/mo, no percentage of recovered revenue, no sales call required.