Strip the dashboards away from any retry tool and one design decision remains, the one that determines nearly all of its results: in what order, and on what evidence, does it spend a subscriber's limited attempts? Issuers watch retry behaviour. A card retried blindly six times in two days does not merely fail six times; it teaches the issuer's systems that this merchant retries blindly, and the seventh attempt — the one timed to payday, the one that would have worked — arrives carrying the reputation of the first six.
A sound sequence starts by reading before it spends. The decline code on the first failure is evidence: insufficient funds points at the calendar, a replaced credential points at the updater and the customer, a hard decline points at the exit. Each reading implies a different second move, and most of them are not "retry tomorrow." In the strongest implementations the second move is often not an attempt at all — it is a wait, an update request, or a message — because an attempt is the scarcest resource in the system and everything else is comparatively free. That ordering principle is the core of retry software judged by the order of its attempts, not their count, and it is a useful lens for evaluating any tool in this category, including the defaults inside your billing platform.
Platform defaults deserve the teardown too. A fixed cadence — day one, day three, day five, day seven — has one virtue, predictability, and one structural flaw: it applies the same hypothesis to every failure. It retries the empty account on day three because day three arrived, not because anything about the account changed. Software that earns its fee replaces the calendar with conditions: retry when the balance cycle turns, when the credential is refreshed, when the customer has answered — when something is true that was not true at the last attempt.
Consider what the sequence does with the commonest reading of all: insufficient funds. A calendar retries it on day three because day three is next. A sequence retries it when money plausibly exists — after the first and the fifteenth for salaried customers, after the weekend for everyone, never twice inside the same empty window. Multiply that single conditional across a thousand failures a month and the effect stops being subtle: the same number of attempts produces materially more approvals, because attempts stopped being spent into accounts that were provably empty at the moment of spending. Nothing about the customer changed. Nothing about the card changed. Only the order changed, and the order was the product all along — a fact the count-first tools cannot replicate by adding a seventh attempt, because a seventh blind attempt is just the first mistake repeated with more confidence.
The teardown test is simple to run on any vendor demo. Ask what the system does between attempts. If the answer is a schedule, you are looking at a timer with a dashboard. If the answer is a set of reads, waits, updates, and messages — each attempt spent only when the evidence changed — you are looking at recovery software. The market uses one name for both. Your failure file will learn the difference within a quarter, in the only metric that cannot be redefined: how many subscribers are still subscribers.
That is the sequence Swipe Again runs on every failed payment: read the decline, spend nothing until the evidence moves, and let each attempt arrive with a reason the issuer's systems — and the customer — can respect.
Top comments (0)