A smart retry is a reattempt of a failed recurring or card-on-file payment whose timing is chosen by a model rather than a fixed schedule. Billing platforms and payment processors train machine learning models on large numbers of past transactions, learning how the chance of success varies with the decline code, the issuing bank, the card type, the timing and the customer's own history. Each failed charge is retried when the model predicts success is most likely, and abandoned when further attempts look futile.
The aim is to recover revenue that would otherwise be lost to involuntary churn while avoiding wasted attempts. Card schemes limit repeated retries of declined transactions and can charge merchants for excessive ones, so retry logic respects decline categories: a temporary soft decline may be retried, a hard decline such as a lost or stolen card should not be. Smart retries are usually one part of a wider smart dunning process.