Every billing platform has a settings page that somebody meant to finish. In Recurly, that page is often the alerts list: a short set of checkboxes deciding which events the outside world ever hears about. A failed payment fires, the platform retries on its own schedule, and if the alert that announces the failure was never ticked, the recovery tooling downstream learns about the problem days later — or never. The subscription does not die loudly. It dies in a gap between two systems that both assumed the other was watching.
This is worth an hour of any subscription team's time, because the failure mode is invisible in the dashboards people actually check. Recurly's own reporting will show the retry attempts it made. What it cannot show is the work that never started elsewhere: the email that was not sent because no event arrived, the card-update link that was never issued, the case that never opened. Absence leaves no row in a report. Teams discover the gap only when someone reconciles a churn list against a recovery list and finds names on the first that never appeared on the second.
The fix begins with an inventory, not a purchase. List every event your recovery process depends on — payment failed, payment recovered, subscription cancelled, card updated — and then read the platform's alert configuration and mark which of those events are actually switched on. In most accounts I have seen described, the list is shorter than the team believes, because the account was configured once, early, by someone solving a different problem. Refund notifications are a special case: the alert name for them is easy to misread, and guessing wrong means either silence or noise. Verify the exact name in the documentation before ticking anything; a wrong guess here is worse than a pause.
That inventory discipline is the unglamorous core of Recurly dunning that reads the alert before the retry: before arguing about retry timing or email copy, prove that the signal announcing a failure reaches the system responsible for acting on it. A retry schedule built on an event that never fires is a schedule for a world that does not exist. Teams routinely tune the visible parts of dunning — the cadence, the wording, the number of attempts — while the invisible part, event coverage, quietly decides the ceiling on everything else.
Once the alerts are right, a second question follows: what does the receiving system do in the first hour? A failure event that lands in a queue nobody processes until morning has been delivered and wasted at the same time. The practical standard is simple — the event opens a case, the first customer contact goes out while the failure is still news, and the case stays open until the money lands or the subscription genuinely ends. If any of those steps depends on a person remembering, the process is a hope, not a system.
There is also a defensive reason to do this audit now rather than after the next churn review. Platforms change defaults, team members change settings while fixing unrelated problems, and integrations get reconnected in ways that reset choices nobody documented. A quarterly read of the alert list, compared against the list your recovery process depends on, takes twenty minutes and catches the drift while it is still cheap. Write the expected configuration down somewhere a new hire can find it; the next person to touch that settings page should be able to tell the difference between a deliberate change and an accident.
None of this is advanced. That is precisely the point. The subscriptions lost to an unticked alert were not lost to a sophisticated competitor or an unavoidable decline — they were lost to a checkbox. Swipe Again treats event coverage as the first thing to verify in any Recurly recovery setup, because every later improvement is built on the assumption that the failure was heard in the first place.
Top comments (0)