Ask five billing operators when a failed payment should be retried and you will get five confident answers, each certain, each different, and almost none of them tested. Retry timing is usually inherited — from a platform default, a predecessor's spreadsheet, or a blog post (possibly like this one) — and then defended as if it were a finding. It is not a finding. It is a hypothesis about your subscribers' lives, and it deserves the same treatment as any other hypothesis: a test with a result you are willing to be wrong about.
The hypothesis has more dimensions than "which day." There is the interval before the first retry, which trades urgency against the issuer's memory of the refusal. There is spacing between attempts, which for insufficient-funds declines is really a bet on pay cycles — biweekly Fridays, monthly firsts, the weekly rhythm of gig income. There is time of day, which interacts with bank posting order in ways that matter for accounts running close to zero. And there is the total window: how long recovery should keep trying before the honest conclusion is that this subscriber needs a conversation, not another attempt.
Testing any of this is unglamorous and entirely doable. Hold the message copy constant and vary one timing dimension for one decline family at a time — soft declines only, say, so hard declines do not pollute the read. Insufficient funds is the natural first subject because its mechanism is explicitly temporal: if your retries are landing before payday, moving them should move the recovery rate in a way no copy change ever could. Give the test enough failures to mean something; on a small base, a fortnight of declines is anecdote, not evidence, and the discipline is mostly in not peeking early and declaring victory.
Two traps wait for the impatient. The first is testing everything at once — new spacing, new emails, new sender name — and learning nothing except that something changed. The second is reading recovery rate alone. A timing change that recovers more payments while doubling attempts per recovery may be worse, not better: attempts cost fees and issuer goodwill. Record attempts per recovery alongside the rate, and unsubscribes or cancels that arrive inside the dunning window, because timing that bullies people into paying once and leaving is not recovery — it is churn with extra steps.
What you are really building, test by test, is a map of your own subscribers rather than the average subscriber in a benchmark post. Gym members paid on shift cycles, newsletter readers on salaries, course buyers on neither — the right retry calendar is a fact about a specific business, which is why copying a famous schedule so rarely works. A grounded reference on when to retry a failed payment can frame the options; only your own decline file can choose among them.
Start with the payday question. It is the cheapest test in billing, the answer is usually visible within one cycle, and it teaches the habit every other timing decision needs: the schedule works for the subscribers, or it does not work at all.
Built and written by the Swipe Again team, students of the retry calendar.
Top comments (0)