Why it matters
Without a plan, a failed payment either turns into a cancellation or sits unpaid. Dunning is the plan, a set of steps that runs automatically from the first failure until the payment is collected or the subscription ends.
Good dunning protects revenue without annoying the customer. Most people whose card fails did not decide to leave, so the aim is to fix the card quickly and keep them in the product while you do it. Measuring each step tells you which part of the plan is earning its place.
The formula
In words: recovery at a step is the customers recovered at that step, divided by the customers who entered dunning. The step rates add up to the overall recovery rate.
In symbols:
step_recovery = recovered_at_step / entered_dunning x 100
total_recovery = sum of step_recovery- Every step uses the same denominator, the customers who entered dunning.
- A customer is counted at the step where the payment was finally collected, and at no other step.
The four parts of dunning
- Retries: new attempts to charge the saved card, spaced over days so a temporary problem has time to clear.
- Messages: emails or in-product notices that tell the customer a payment failed and link to a simple card update page.
- Grace period: the time the customer keeps access while the payment is still outstanding.
- Final action: what happens when the last step fails, such as pausing or cancelling the subscription.
A worked example
Example numbers (made up for illustration): 400 subscribers entered dunning in a month. The table counts how many were recovered at each step.
| Step | Recovered at the step | Share of those who entered dunning |
|---|---|---|
| Retry on day 1 | 60 | 15.0% |
| Retry on day 3 | 40 | 10.0% |
| Email with a card update link | 32 | 8.0% |
| Final retry on day 7 | 28 | 7.0% |
| All steps | 160 | 40.0% |
Each step is measured against the 400 who entered dunning, so the step rates add up to the total: 160 recovered is 40.0%. The table shows which step earns its place and which one adds little.
What published data says
Stripe documents a recommended retry schedule for its Smart Retries feature. It is a starting point for your own settings, which should follow your own results.
- Stripe Smart Retries recommended default: 8 tries within 2 weeks. Stripe documentation recommends 8 tries within 2 weeks for Smart Retries. Read on 18 Oct 2026. Source.
How to improve it
Treat dunning as a short sequence you can read step by step, then adjust the weakest part.
- Spread retries across days and weekdays, and stop retrying hard declines.
- Write messages that name the amount, say what failed and link straight to a card update page.
- Choose a grace period long enough to fix a card and short enough that unpaid use does not pile up.
- Warn the customer before the final action, with the date it will happen.
- Review recovery by step each month and remove steps that add nothing.
Common mistakes
- Retrying hard declines, which wastes attempts and cannot succeed.
- Sending one email and nothing after it, so a missed message ends the subscription.
- Ending the grace period without warning, which turns a fixable card problem into an angry cancellation.
- Writing messages that read like threats. A calm note with a clear link recovers more goodwill than it costs.
- Not measuring recovery by step, so a step that does nothing runs for years.
Work it out with your own numbers
The failed payment recovery calculator lets you test how a different recovery rate changes the revenue you keep. Open the failed payment recovery calculator and enter your own figures. It runs in your browser and shows each step of the sum.
Read next
- Payment recovery rate: the overall result that dunning produces.
- Involuntary churn: the customers lost when dunning does not work.
- Failed payment rate: how many payments enter dunning to begin with.
Questions
What are the parts of dunning?
Retries, messages, a grace period and a final action. Together they cover how you try again, how you tell the customer, how long they keep access and what happens at the end.
How long should a grace period last?
Long enough for a customer to notice the message and update a card, and short enough that unpaid use does not accumulate. Set it from your own recovery results by step.
Should I retry every failed payment?
No. Retry a soft decline, since it may clear. After a hard decline, request a different card.
How do I know a step is worth keeping?
Measure the customers recovered at each step against everyone who entered dunning. A step with a very small share is a candidate to change or drop.