DEV Community

Swipe Again for Swipe Again

Posted on

Four Jobs Wearing One Label

"Dunning solution" is one of those category labels that sounds precise and means almost nothing until you unpack it. Under the label sit at least four different jobs, and vendors differ mainly in which of the four they actually do. Buying the label without separating the jobs is how a team ends up with excellent email templates attached to a retry engine that guesses, or a clever retry engine whose customer messages read like legal threats. The fix is a short taxonomy, applied before any demo.

Job one: asking the bank again, intelligently. Retries, scheduling, decline-code handling. This is the mechanical core — when to re-present a charge, how the response to one code differs from another, when to stop. It is also the job most often oversold: a platform's built-in retry already does a crude version, so the question is what the specialist adds — per-code logic, payday-aware timing, card-updater integration — stated concretely, not as "AI-powered optimization."

Job two: asking the customer, humanely. Emails, texts, in-app notices, and the card-update page they land on. Tone, timing, and friction live here. A solution can win job one and lose job two completely: perfectly timed retries paired with a message the customer never sees because it landed in spam, or a landing page that demands a full login before it accepts a new card. Test this job as a customer, on a phone, before believing any of it.

Job three: watching the credentials. Account updater services, expiry warnings, pre-dunning notices before a known-bad card is charged at all. This is the quietest job and often the highest-yield, because a failure prevented never enters the queue. Not every vendor does it, and some platforms do a version natively — mapping who watches credentials, yours or theirs, prevents both gaps and double-paying.

Job four: knowing what happened. Reporting that separates recovered-by-retry from recovered-by-customer-action from simply written off, and that prices the tool against the recovery it caused. Without job four, the other three are vibes: nobody can say which job is failing when the recovery rate sags, so every renewal of the contract is an argument instead of a review.

Comparing stripe dunning by the four jobs it actually has to perform turns a vague category purchase into four answerable questions. Which jobs do we need done that our billing platform does not already do? Which does this vendor demonstrably do, with what evidence? What does the bundle cost against the job it replaces — remembering that a percentage-of-recovery fee taxes job one hardest exactly when it works, while a flat fee prices all four jobs the same whether the month was heavy or light?

The taxonomy also explains most disappointing purchases in this category. The team needed job two and bought job one. The team had job three handled by their processor and paid for it twice. The team never staffed job four, so a tool that was working looked identical to one that was not, and got cancelled in a budget pass. None of those are vendor scandals; they are labeling failures, and they are avoidable with an hour of separation work before the contracts come out.

One more discipline follows from the taxonomy: re-run it annually. Platforms absorb jobs over time — a billing platform that lacked updater coverage this year may ship it next year, quietly converting a paid vendor job into a duplicate. Vendors drift the other way, sunsetting the reporting you bought them for. The one-page map, refreshed each renewal season, is what keeps the stack honest between purchases.

Write the four jobs on one page, mark who owns each today, and let vendors bid against the gaps. The label will still be vague. Your purchase will not be.


Published by Swipe Again — failed payment recovery at a flat $179/mo.

Top comments (0)