DEV Community

Swipe Again
Swipe Again

Posted on

The Exception Report Is the Meeting

Most recovery reviews start with a rate. The rate is up or down a little, everyone nods, and the meeting ends having changed nothing. I have sat in that meeting. The number was accurate and completely unusable, because a rate cannot tell you which cases to touch, which to leave alone, and which should never have entered the queue at all. The document that can do that work is not the dashboard. It is the exception report: the short list of cases that behaved unlike the rest.

Build the habit around three exception families. First, cases that failed twice with different answers. A soft decline followed by a hard decline is not two failures; it is a story changing mid-sentence, and it usually means the card itself changed state between attempts. Those cases deserve a human look before another automatic try, because a third blind attempt spends issuer goodwill on a question the first two answers already settled.

Second, cases that recovered somewhere else. The customer paid through a support link, a replacement subscription, or a manual invoice while the recovery case was still open. If your system counts those as ordinary expiries, your queue learns the wrong lesson and your reporting understates what actually happened. The exception report should name every case where money arrived outside the case, so the case can close honestly as recovered by replacement and the pattern — which link, which channel, which day — becomes visible.

Third, cases that should have been impossible. A retry fired after a cancellation. A message sent inside quiet hours. An attempt made against a card the customer had already replaced. Each one is small. Together they are the audit trail of whether the system respects its own rules, and they matter far beyond their dollar value, because a merchant who finds one impossible case stops trusting every other number on the page.

Keep the report short enough to be read. One page per family, ten cases per page at most, each case carrying its age, its decline history in plain words, and the last check the system ran. If the report grows past three pages, the problem is no longer the cases — it is the rules upstream generating them, and the meeting should say so out loud rather than admire the thickness of the printout.

Run the review in a fixed order. Read the impossible cases first and fix causes, not cases. Then read the recovered-elsewhere cases and close them properly. Then read the changed-answer cases and decide, one by one, whether the next step is a message, a wait, or a deliberate ending. Only after those three lists are empty does the headline rate deserve a glance — and by then it usually explains itself.

None of this needs new tooling to start. Export a month of failures, sort by what surprised you, and spend one hour a week on the surprises. Teams that do this stop arguing about whose rate definition is right, because the argument moves to named cases with named endings, which is where operational truth lives.

For the fuller operating model behind that weekly habit, an exception-first guide to dunning management software and the queue it should produce lays out the case states, the checks before every attempt, and the closing rules that keep a report honest.

A rate is a summary. An exception report is a workload. Manage the workload and the summary takes care of itself; manage only the summary and the workload quietly manages you.

Swipe Again closes every case with a named ending — recovered, recovered by replacement, paused with its pause, or lost and counted as lost — so the exceptions a merchant reads are the real ones, for one flat monthly fee.

Top comments (0)