DEV Community

Swipe Again for Swipe Again

Posted on

Smart Retries Are a Bet About Strangers: Reading Stripe's Timing Like Your Own

Stripe's Smart Retries make a genuinely reasonable bet: across millions of subscriptions, machine-learned timing recovers more than any fixed schedule a small team would design by hand. The bet is about strangers — the average subscriber, in the average week, at the average bank. Your renewal queue is not made of strangers. It is made of your subscribers, whose paydays cluster around your pricing, whose banks cluster around your market, and whose failures cluster around the specific week your product renews them. The gap between the average and your queue is where a billing team earns its keep.

Start by separating two questions that dashboards habitually merge. The first is whether the platform's retries recovered a payment. The second is whether they recovered it because of the timing, or simply before your own process would have. A recovery on the platform's eighth day looks identical in the summary to a recovery your card-update email caused on day two — but they imply opposite decisions about where to invest next. Until a team stamps each recovery with what actually preceded it, every argument about retry strategy is an argument about fog.

The decline mix matters more than the retry count. Smart timing helps most against the soft, situational failures — insufficient funds early in a pay cycle, a bank's temporary risk posture, a credential mid-replacement. Against a card that was cancelled in a fraud sweep, no timing model on earth recovers anything; the only route is a human updating a card, which means the email, the hosted update page, and the tone of both matter more than the schedule. Teams discover, once they sort by decline family, that their queue is usually two different problems wearing one label.

stripe dunning software that sorts declines before it schedules retries lays out that sorting in detail: which decline families respond to patience, which respond only to a card update, and how to read the platform's recovery numbers without crediting the schedule for the subscriber's own initiative.

A small experiment makes the point concrete. Take one month of failures and hand-sort them, on paper, into three piles: paid after a retry and nothing else, paid after a subscriber touched the card, and never paid at all. Most teams who do this expect the first pile to dominate. What they usually find is a humbling second pile — subscribers who fixed the problem themselves after a message, while the retry schedule was still warming up — and a third pile dominated by one or two decline codes they had never isolated. That single afternoon of sorting routinely changes a team's roadmap more than a quarter of dashboard reviews, because it replaces a platform's average story with the queue's actual one.

The practical posture that follows is neither worship nor replacement. Leave the platform's retries running — they catch the easy timing wins at zero marginal effort — and build your own work in the spaces the model cannot occupy: the subscriber-facing messages, the card-update path, the watch on credentials that changed upstream, and the honest ledger that says which recoveries were yours. Where the two schedules meet, weave rather than collide: your outreach lands in the platform's quiet stretches, never an hour after its loudest attempt.

Judged that way, Smart Retries stop being a feature you either trust or resent and become what they actually are — a competent default for the average case, running underneath a queue that is not average at all. The businesses that recover best on Stripe are rarely the ones with the cleverest retry settings. They are the ones that stopped asking the platform's model to know their subscribers, and started keeping the records that let them know the subscribers themselves.


Published by Swipe Again — failed payment recovery at a flat $179/mo.

Top comments (0)