Failed payment recovery rates: what is actually recoverable
Last reviewed 2026-10-07
Our audit estimates what share of your failed payments was recoverable. That estimate rests on three numbers, and this page exists to tell you exactly where they come from. The short version: they are conservative assumptions, not measurements, and we would rather you knew that than trusted a figure we cannot defend.
The three numbers we use
| Family | Rate we apply | Reasoning |
|---|---|---|
| Card needs updating | 60% | The money usually exists; the card does not. Recovery depends entirely on the customer entering new details, which some will and many will not. |
| Insufficient funds | 25% | The card is valid and the customer intends to pay. A later attempt can succeed, but a balance that was empty once is often empty again. |
| Hard declines | 0% | Lost, stolen, fraudulent, issuer refusal. We count none of it, because retrying cannot work and Stripe itself refuses to retry several of these codes. |
| Other | 0% | Codes we cannot classify. We count none of it rather than assume the best, because an unexplained decline is not evidence of an opportunity. |
These percentages are read directly from the configuration the audit runs on. If we ever change them, this page changes with them — it cannot quietly drift from the product.
They are assumptions, not measurements
We want to be unambiguous about this, because the whole value of the audit rests on you being able to trust what it tells you. The three rates above are judgement calls. They are not derived from a study of your account, they are not derived from a study of anyone's account that we conducted, and we cannot point you to a dataset that produces exactly sixty and twenty-five.
They are deliberately set low. If we are wrong, we would rather be wrong in the direction that understates what you could get back, because you are going to check this number against reality and an inflated estimate is the fastest way to lose your trust permanently.
This is also why the recoverable figure is presented with a tilde and never as a forecast. It is an order of magnitude intended to answer one question — is this worth doing something about — and not a target to plan against.
The one measured number we trust
There is one public figure in this category with a stated methodology, and it is worth more than everything else we found combined.
Baremetrics publishes a median attempted recovery rate of 12.7%, drawn from 119 US B2B SaaS companies actively using its recovery product during May 2026, 94% of them on Stripe. The sample is their own paying customers, so it is biased toward businesses with a real problem — but the methodology is stated, the period is stated, and the sample size is stated. That is three more things than most figures in this field come with.
Note what that number implies. With a paid recovery tool running, the median company recovered roughly one in eight of its attempted recoveries that month. Any claim that failed payments are mostly recoverable has to be reconciled with that, including ours.
Why we did not use the figures everyone quotes
Search for recovery rates by decline reason and you will find consistent-looking ranges: insufficient funds recovering at 70–80%, expired cards at 40–60%, an industry median around 47.6%, processing errors at 80–90%.
We looked for the origin of those numbers and could not find one. They appear in blog posts published by companies selling recovery software, citing each other, with no sample size, no period, no definition of what counts as a recovery and no link to underlying data. We are not accusing anyone of inventing them. We are saying we could not verify them, and a number we cannot trace is not a number we will build an estimate on.
There is also a definitional trap in them. A recovery rate of 70% means something very different depending on whether the denominator is all failed invoices, only the ones a tool attempted, or only the ones it attempted more than once. Baremetrics is explicit that it reports attempted recovery rate for exactly this reason. Most of the circulating figures do not say.
What the evidence does support
Two things are documented rather than asserted, and both shaped our coefficients.
- Hard declines are not retryable. Stripe documents nine decline codes it will not retry at all, including lost_card, stolen_card and transaction_not_allowed. The scheduled attempts keep counting but no charge is made. A zero coefficient here is not pessimism, it is the documented behaviour of the platform.
- Expired cards have a partly automatic fix. Stripe's automatic card updater refreshes stored details when an issuer reissues a number, which recovers a slice of the card-update family without anyone being emailed. That is part of why we rate this family highest, and also why the rate is well short of total.
Everything beyond those two points is inference. We have told you which is which.
How to measure your own rate instead
Our estimate is a starting point, not a substitute for your own data. If you want the real number for your account, it is a few minutes of work and it is free.
- 1.Take the invoices from a period that has fully played out — twelve months ago rather than last month, so the retries have finished.
- 2.Count the ones that failed at least once and were eventually paid. That is your recovered set.
- 3.Divide by all the invoices that failed at least once. That is your real recovery rate, on the honest denominator.
- 4.Split the failures by decline family and repeat. Now you have your own coefficients, measured rather than assumed.
If your numbers differ meaningfully from ours, yours are right. We would genuinely like to hear about it.
Questions people ask
- What percentage of failed payments can be recovered?
- There is no trustworthy single answer. The one public figure with a stated methodology is Baremetrics' median attempted recovery rate of 12.7%, from 119 of its own US B2B SaaS customers in May 2026. Higher figures circulate widely but we could not trace any of them to a sample, a period or a definition.
- Where do Cardnudge's 60/25/0 recovery rates come from?
- They are our own conservative assumptions, set deliberately low. They are not measurements of your account or of anyone's. The zero for hard declines is documented behaviour — Stripe refuses to retry nine decline codes — but the sixty and twenty-five are judgement calls, and we would rather say so than dress them up.
- Are expired cards or insufficient funds more recoverable?
- We rate expired and incorrect cards as more recoverable, because the money generally exists and Stripe's automatic card updater resolves part of the family on its own. Insufficient funds depends on a balance that was empty once being full later, which is less reliable than it sounds.
- Why do you count unclassified declines as zero?
- Because we do not know why they failed. Assuming an unexplained decline is recoverable would inflate the estimate in exactly the direction that flatters us, so we count none of it.
Sources
- Baremetrics — Subscription Payment Recovery Benchmarks (2026) — published 3 June 2026, updated 10 July 2026, checked 7 October 2026
- Stripe Docs — Smart Retries, non-retryable decline codes — checked 7 October 2026
- Stripe Docs — Revenue recovery, automatic card updates — checked 7 October 2026