Before you rewrite a dunning email, add a retry, or shop for a recovery tool, run one query: last month's failed Stripe charges, grouped by decline_code. It takes an afternoon at most, and it replaces almost every opinion in the room with a distribution. Most teams who run it for the first time discover the same surprise — their failures are not a hundred different problems. They are three or four codes wearing most of the volume, and each of those codes wants a different, specific response.
Start with the card family. expired_card means the date passed; no retry on any calendar can succeed, and the only move is a prompt, friendly request for a new card through a secure update link. incorrect_cvc and its cousins mean the stored details are wrong; again, the customer is the fix, and attempts should stop until fresh details arrive. lost_card and stolen_card are the bank's own verdict — never retry, ask for a new payment method, and keep the tone routine, because from the customer's side this is paperwork they have already done with their bank.
Then the bank family, where timing is the tool. insufficient_funds is usually the largest code in a subscription account, and it is a calendar problem: the card is valid, the customer intends to pay, and the account was empty that day. Spread attempts across the pay cycle — around the 1st and 15th patterns — instead of burning them in the first week. do_not_honor and generic_decline carry no explanation at all; cool off at least a day, retry sparingly, and pair the attempt with a message asking the customer to call their bank, because the bank is waiting to hear from them, not from you. card_velocity_exceeded is the confession code: it usually means your own schedule retried too fast, and the fix is to back off and audit the cadence that caused it.
Finally the action family. authentication_required can only be cleared by the customer completing a confirmation step — send the link, say plainly what the bank is asking, and do not waste retries in the meantime. processing_error is the opposite: nothing is wrong with the card or the customer, so a short-delay retry is right, and only repetition should promote it to the slow, outreach-led path.
For the full routing — every common code with its meaning, whether to retry, and the exact recovery move — stripe dunning is the working reference. Print the table, mark your top three codes from the export beside it, and you have next month's recovery plan in one page.
The export also settles an argument teams have for quarters: whether to fix outreach or timing first. If your top code is expired_card, outreach is the business — message quality, send timing, and the update page decide your recovery rate. If it is insufficient_funds, the calendar decides it, and prettier emails change little. If do_not_honor leads, your recovery depends on customers phoning banks, which means clarity beats charm. The distribution tells you which craft to practice.
Log the code with every failure from today forward, and stamp what recovered each payment — a retry, a card update, a customer's bank call. A quarter of that logging turns the next planning meeting into arithmetic.
Swipe Again runs this routing automatically across Stripe and eight other platforms: decline-aware retries, hard declines never retried, branded outreach with secure card-update links, all at one flat monthly price. The export exercise above is free, though — run it first, and let your own codes tell you what you need.

