A merchant comparing recovery tools faces a strange market: every vendor publishes a recovery rate, almost none publish what that rate counts, and pricing models differ enough that two quotes for the same business can differ by an order of magnitude. Comparing honestly takes an afternoon of reading, not a demo call. That afternoon is not wasted time. It is the difference between buying a reporting definition and buying an operation. If you build or run billing code, you already know this pattern from observability tools: the dashboard number is easy to quote, and the event definition underneath it is the whole argument.
1. Counting: one dollar, one door, one stamp
A recovery rate is only meaningful if a recovery means a dollar that settled through one door, stamped once. Tools that count the processor's native retries as their own work will always look better on paper, because the platform was already going to make some of those attempts before any second tool arrived.
Ask where attribution is stamped. A disciplined answer describes a single choke point, in a fixed order: retry first, then card update, then a save offer, then honest external for everything else. That last bucket matters. Some payments recover because a customer fixed a card somewhere else, or because time passed. Calling that vendor recovery is not optimism; it is double counting, and it will survive every renewal conversation until someone rebuilds the ledger from settled invoices.
Try a small test in a spreadsheet. Take twenty recent failed invoices. For each one, write down what finally settled, which attempt settled it, and whether that attempt would have happened anyway under the native schedule. The exercise takes an hour, and it changes the questions you ask. You stop asking, "What is your rate?" and start asking, "Which rows in this list would you have changed, and how would you prove it?"
2. The ceiling: two engines should not race one card
Next, ask how a tool's attempts interact with the platform's native schedule. Two engines racing the same card is how merchants get flagged by issuers, annoy customers with duplicate notices, and lose track of who caused what.
A serious answer sounds like a shared cap and attempts woven into the gaps. In our own case, retries weave into the gaps of the platform's native schedule under a combined cap of 15 attempts, and our own ceiling is 8 attempts per failed payment. Before every attempt, the invoice is re-read fresh, so a charge that already settled is never chased again. Hard declines stop and route to a card-update request. Soft declines are re-read and re-spaced. Insufficient-funds retries ride a payday ladder out toward day 25, instead of clustering in the first anxious week.
Notice what that description contains: limits, ordering, and stopping rules. Dunning emails on day 1, 3, and 7, and SMS where a merchant's program allows, are communication around the attempts — they are not a substitute for deciding when not to attempt. A vendor who can only describe effort, not restraint, has told you something important.
3. Price shape: what happens to the bill when it works?
Third, compare price shape, using prices checked in October 2026. A revenue share, the model Butter uses, costs nothing until it works and then scales with your own recovered money. A per-seat or platform add-on bundles recovery into another bill: Baremetrics Recover is a $129/mo add-on. Tiered flat fees are predictable but step up: Churn Buster is $269/mo up to $50k MRR and $430 above that; Churnkey is $250/mo billed yearly or $300 monthly. A single flat fee never grows with recovery: Swipe Again is $179/mo flat.
None of these shapes is automatically right. A share can suit a business that wants no fixed cost. An add-on can suit a team already living in that dashboard. What you cannot do honestly is compare a share, an add-on, a tier, and a flat fee as if they were four prices for the same month. Run each model at your current recovered revenue, then at twice that number, because success is exactly the case you are buying for.
The honest comparison of dunning solutions is mostly these three questions — counting, ceiling, price shape — asked of every vendor in the same order. Write the answers down in the vendor's own words. If a definition shifts between the first question and the third, keep the earlier definition and note the shift.
None of this requires a sales call, and a vendor who insists the answers need one has answered the first question already. The facts that decide this purchase live in invoices, decline codes, schedules, and published prices. An afternoon with those will tell you more than a polished demo, and it will leave you with a ledger you can check again after the first month, whichever tool you choose.
Published by Swipe Again - failed-payment recovery for subscription businesses.
Top comments (0)