DEV Community

Swipe Again for Swipe Again

Posted on

Stripe Retried at 3 A.M. Your Customer Was Asleep, Then at Work, Then Gone

Stripe's retry schedule is a machine schedule. It fires at intervals that make sense to a system — hours and days counted from the moment of failure — and it is indifferent to the human calendar on the other side of the charge. A renewal that fails late on a Sunday night gets its next attempts in the small hours and in the middle of a working Monday, which is to say at the two moments when a timing failure is least likely to have resolved and when a message asking for a new card is least likely to be read by someone able to act on it.

That indifference is not a defect in Stripe. Stripe Billing was built to be a reliable general-purpose engine, and a general-purpose engine cannot know that your subscribers are paid on the last Friday of the month, or that your typical customer is a parent who deals with household admin on a Sunday evening, or that a particular account has already replaced its card at two other merchants this month. The schedule knows the failure happened. It knows nothing about why, and the why is what decides when the next attempt should happen and what, if anything, should be said before it.

Operators usually discover this in stages. First they turn the native retries up, because more attempts feels like more recovery. Then they watch the recovery rate stay almost exactly where it was while the attempt count climbs, which is the signature of attempts landing in the wrong places: a balance that is empty on Tuesday is usually still empty on Wednesday, and a card that has been replaced will fail identically on the fifth presentation as on the first. More attempts against an unchanged condition produce more declines, not more revenue, and each decline is another line in the customer's statement history that makes the eventual recovery conversation harder.

The useful work sits between the attempts. Read the decline before scheduling anything. A soft timing decline wants a date, not a count — move the attempt to the customer's pay cycle and leave the intervening days quiet. A replaced-card decline wants a lookup, not a retry — the network often already holds the new credential, and asking the customer to type a card they already updated elsewhere is the fastest way to teach them that your billing is not paying attention. An approval-style decline wants a human sentence at a human hour, explaining that their bank is asking them to confirm the charge in their app, because no retry schedule can complete a step that belongs to the cardholder.

None of this requires replacing Stripe. It requires refusing to let the default schedule be the whole strategy. Keep the native engine for the straightforward middle — it handles that well — and put the reading layer where the schedule is blind: before the attempt, not after the account has already exhausted it.

That is the practical content of stripe dunning: live checks before every attempt, timing aimed at paydays instead of intervals, credential repair through the updater networks, and messages timed for when a person can actually answer them. The platform supplies the rails. The recovery lives in the decisions made on top of them.

A subscription business rarely loses customers at the moment of failure. It loses them at the end of a schedule that never once adjusted to anything it learned. The accounts are still there, still willing, still reachable — right up until the last automated attempt fires at another impossible hour and the system closes the file on someone who would have paid on Friday.

That gap between the machine's calendar and the customer's is where Swipe Again works: reading each Stripe failure on its own terms and timing the response to the life on the other side of it, so a renewal that failed at 3 a.m. gets its next honest chance when the customer is actually in a position to say yes.

Top comments (0)