DEV Community

Swipe Again
Swipe Again

Posted on

Chargebee Already Has Three Retry Systems. The Case Still Needs an Owner.

Billing on Chargebee gives you an unusual problem: abundance. There is the configurable dunning sequence you set yourself, a smart retry option that reacts to gateway errors, and a machine-learning retry engine in the payments layer reading signals about funding cycles and issuer behaviour. Three systems, all retrying. It is entirely possible to run all of that well and still lose subscribers you should have kept — because retrying was never the whole job, and none of those systems owns the case around the retry.

Walk one failure through slowly. A card is declined on a Tuesday. The retry logic, however clever, answers one question: when should this charge be attempted again? It does not answer the others. Has this customer's card been replaced since the last successful charge, making every retry an attempt on a dead credential? Did the customer see any message, in any channel they actually read, before access changed? If they clicked cancel while the payment problem was unresolved, who met them there — and was that conversation connected to the billing case, or a separate product with separate pricing having a separate chat?

Those unanswered questions are not defects in the retry engines. They are the boundary of what a retry engine is. The gap shows up in reporting, too. Native reporting can tell you what the retry schedule recovered in aggregate. It is much weaker at telling you why an individual case ended: which message landed, which attempt worked, whether the recovery was the schedule's doing or a card update the customer made for unrelated reasons. Without case-level ownership, every month ends with a rate and no story, and a rate without a story cannot be improved on purpose.

Configuration depth makes this harder, not easier, to see from inside. Two merchants on the same platform can have completely different involuntary churn purely from settings: how many retries, what spacing, when the subscription is marked unpaid, when the customer is first told. So the first audit is always of your own configuration, before any tool comparison. Read your end-of-sequence rule in particular. What happens after the final attempt — cancellation, a paused state, indefinite limbo — is a policy decision many teams made once, during setup, and never revisited while their prices, plans, and customer mix changed around it.

If you want the full map before deciding anything, this complete guide to Chargebee dunning and where its native stack stops separates the three native pieces, names their edges, and prices the options honestly, including the fact that cancel deflection lives in a different product priced per session.

The practical test for any added layer, ours included, is coordination. Two uncoordinated retry schedules working the same failure is worse than one mediocre schedule: duplicate attempts, contradictory emails, and a customer who feels pursued rather than helped. A recovery layer earns its place only if it checks the live state of the case before every attempt, steps around the native schedule instead of stacking on it, and closes each case with a recorded reason — recovered, cancelled by choice, or lapsed at the end of a stated grace period. Unknown should not be an available ending.

Judge the economics the same unsentimental way. Native dunning is bundled; a dedicated layer is a new line item. Run the fee against your actual monthly failed volume, not a vendor's example account, and ask what small number of additional recoveries — above what your current settings already achieve — would cover it. If that number is modest and your unrecovered queue is not, the case owns itself.

Swipe Again is built to be that owning layer on Chargebee: one flat monthly price, the outreach and card-update path run as a single sequence with the retries, and cancel deflection included rather than sold back to you by the session.

Top comments (0)