A Braintree vault record looks static when you open it: a customer, a payment method, a token, a set of dates. It reads like a filing cabinet. In practice it is closer to a medical chart. Cards get replaced and the vault learns about it through the account updater. Customers update a payment method on a Tuesday and the old token sits there next to the new one. A subscription created eighteen months ago still points at whatever the vault held on the day it was sold. Every retry a business runs is only as good as its reading of that chart at the moment of the attempt.
That is where Braintree renewals quietly go wrong. The platform's own recurring billing will try again on its schedule, and those retries are not stupid — but they work from the subscription's stored payment method, and they do not pause to ask whether the vault has changed since the failure. A card can be updated in the vault on Monday, by the customer or by the updater service, while the retry logic keeps presenting the instrument that already failed. The renewal dies holding a stale token, in a vault that contained the answer the whole time.
Working this properly is mostly a discipline of checking before acting. Before an attempt, read the current state: has a newer payment method appeared, has the updater refreshed the card's number or expiry, has the customer already fixed the problem through a hosted page and simply not been retried yet? Each yes changes the action. A refreshed card wants an immediate, quiet attempt. A brand-new payment method wants the same. A vault that has not changed at all wants patience and a message, not another identical charge presented to the same bank with the same details and the same predictable answer.
The decline itself also has to be read against the vault. An expired-card decline with a refreshed expiry already sitting in the record is a solved problem waiting for an attempt. The same decline with no refresh is a request to send the customer a secure update link and wait. Insufficient funds has nothing to do with the vault at all — it is a calendar problem, and the vault will look identical on the day the retry succeeds. Sorting those three situations apart is the entire job; tools and sequences are just ways of doing the sorting reliably at volume.
The full working method is laid out in the Braintree dunning guide that treats the vault as a living record, including how the account updater, the platform's native retries, and an outside recovery layer can be arranged so they check the same current record instead of three different snapshots of it.
There is a reporting consequence too. When a renewal recovers because the account updater refreshed a card in the background, that is a different kind of win from a customer clicking an update link, and both differ from a well-timed retry after payday. Lumping them into one recovery number hides which part of the system is earning its keep. Read the vault's history and the recovery report becomes a diagnosis: updater coverage doing the work, customer outreach doing the work, or timing doing the work. Each answer points at a different next improvement, and none of them is visible in a single blended percentage.
None of this requires exotic technology. It requires treating a stored payment method as a claim about the present that expires daily, and building the retry habit around re-reading the record before every attempt. Businesses that do this stop losing the strangest losses in subscription billing: the renewals that failed while holding the customer's new card in their own vault, unconsulted.
That check-before-every-attempt discipline is how Swipe Again handles Braintree recovery: each case re-reads the current payment state before acting, weaves its attempts into the gaps of the platform's own schedule, and closes the moment a card update or a successful charge settles it — for one flat monthly fee.
Top comments (0)