How to actually measure failed-payment recovery
Failed payments are one of the largest quiet losses in digital commerce — industry analyses put the cost around $129 billion in 2025, and involuntary churn (subscribers lost to payment failure, not choice) accounts for roughly 20 to 40% of all subscription churn depending on segment. So it is no surprise that every payments vendor now leads with a recovery rate.
Here is the problem: those rates are not comparable. Not because vendors lie, but because the industry never agreed on what the number measures. Two vendors reporting 50% and 80% may have identical performance. The difference is the math.
The three denominators
Recovery rate is a fraction, and the denominator is a choice. There are three in circulation:
1. Recovered / retried attempts. The friendliest math. If you only retry the failures you're most likely to win, the rate climbs as coverage falls. A 90% rate on this basis can mean you attempted almost nothing.
2. Recovered / recoverable failures. More honest, but "recoverable" is itself a judgment — hard declines (stolen card, closed account) are excluded, and where that line sits varies by vendor.
3. Recovered / all failed volume. The number a CFO actually feels. It is always the smallest of the three, which is why it is quoted the least.
Benchmarks only mean something on a stated basis. The 2025 published benchmarks put median recovery near 47.6%, with layered programs reaching 70 to 85% among top performers — useful reference points, provided you know which denominator they sit on and apply the same one to your own stack.
The window and the channel
Two more choices hide inside every headline rate. The window: a recovery counted within 7 days is a different achievement than one counted within 45, and for subscription businesses anything recovered after the cycle closes may already be a refund case. The channel: some recovery is invisible to the customer (an intelligent retry at a better moment, or the same transaction routed through a different provider), while some requires re-engaging the customer to update a card or complete a checkout. Both are legitimate. They are different capabilities, and a vendor quoting one blended number is quoting neither.
Why more retries is not more recovery
The tempting fix — retry harder — backfires in a specific way. Issuers score merchant retry behavior, and tightly spaced, undifferentiated retries push soft declines toward hard declines and damage the merchant's standing with the issuer. Recovery is a timing and routing problem: the decline reason, the issuer, the time of day, and the payment method all change when and whether a retry should happen at all.
What this looks like when it works
On Yuno's network, recovery runs as layered agents, and the numbers are quoted with their definitions. Smart routing with automatic retries recovers up to 30% of declined transactions — measured against declines, recovered invisibly, inside the transaction flow. NOVA, the recovery agent, re-engages customers whose payments failed and recovers up to 75% of failed payments, across 190+ countries, with no code to deploy — with "up to" doing honest work, and the basis stated on request for any merchant's own volume.
The layering matters more than either number: the cheapest recovery is the one the customer never sees, and the recovery agent should only receive what routing and retries could not already save.
Build your recovery scoreboard
1. Start from all failed volume, not from attempts. Pull every failed transaction for a quarter — declines, abandoned payment steps, failed renewals. That total is the denominator nobody wants; use it anyway.
2. Classify the failures. Hard declines (closed account, stolen card) out; everything else is addressable. Now you also have the honest middle denominator.
3. Compute the rate on all three bases. The spread between them is a map of where recovery underperforms: coverage, win rate, or both.
4. Fix the window to your billing cycle. A recovery that lands after the cycle closes is a different, smaller win. Count it separately.
5. Split invisible from customer-facing. Retry-and-reroute recovery and re-engagement recovery are different machines. Measure them apart, or you cannot tell which one is failing.
With the scoreboard in place, benchmark honestly — and when any vendor quotes a recovery rate, one question tells you whether it survives your denominators.
The takeaway
A recovery rate without its denominator is a marketing asset, not a metric. Build the scoreboard first; compare second. The teams that know their own number on all three bases are the ones no headline rate can dazzle.
Book a demo to see recovery measured on your own failed volume, on all three bases.
Sources
Slicker — 2025 Failed-Payment Recovery Benchmarks