A payment failure is a technical event. A cancellation is a customer decision. A broad dunning sequence can connect the two by asking every customer to reconsider a subscription before the payment mechanism has been diagnosed.
My view comes from five years leading billing and payment product work. The model on this page separates the payment failure from the customer's engagement state, then shows how both change the recovery decision. Its outputs are authored indices, not production recovery rates or retention probabilities.
DECISION FRAME
Two signals decide the route
The payment mechanism determines what can fix the charge. The customer state determines whether contact helps or reopens a cancellation decision. Pick any cell to send it to the model.
Along a row the route holds: the mechanism decides what can fix the charge.
Down a column the contact posture holds: the customer state decides whether to message.
Reissued credential · DormantStart with Updater first. Quiet routes first: in the model, a dunning message moves the risk of reopening the decision by +28 points for this customer. Silent repair limits decision risk, but it does not solve the underlying lack of use.
Separate mechanism from customer state
An expired card, insufficient funds, and an issuer decline fail for different reasons. Active, at-risk, and dormant customers respond differently to contact. Mixing those categories creates a taxonomy that cannot explain why a route changes.
The technical mechanism should decide what can fix the charge. The engagement state should decide whether contact helps, when it should happen, and what decision risk it creates.
Recovery is the first outcome, not the last
Payment recovery answers whether a charge eventually succeeded. It does not answer whether the subscription survived.
A program can recover the invoice and still lose the customer weeks later. That is why the model separates immediate recovery from 30-day, 60-day, and 90-day survival. The survival curve is illustrative, but the measurement principle is operational: recovery and retention need separate scoreboards.
Route on the failure mechanism
- Expired credential: account updater is the first technical route. A retry does not change the expiration date.
- Insufficient funds: timing matters. A retry guided by current advice or a known pay cycle can be more appropriate than credential repair.
- Issuer decline: processor and network advice should determine whether retry is appropriate. Stripe documents advice codes such as `do_not_try_again`, `try_again_later`, and `confirm_card_data`, and warns that unsuitable additional retries can be viewed as potential fraud. Stripe decline guidance
- Reissued credential: updater is the first route when the network has the replacement. Visa describes account updater as a way for participating issuers to provide updated account information for credentials that expired or were reissued, lost, or stolen. Visa Account Updater
These are starting routes rather than blanket rules. Coverage, consent, issuer advice, and service criticality can change the decision.
Let engagement state govern contact
An active customer may accept a brief payment message because the product is already present in their routine. A dormant customer receives a different prompt: decide again whether this subscription is worth keeping.
That does not mean a team should hide a failed payment. It means contact should be designed as a retention intervention, measured as one, and used after the quiet technical routes have been considered.
How to read the model
Choose a failure mechanism, an engagement state, and a recovery program. The technical route sets immediate recovery and route fit. The customer state changes downstream survival and decision risk. Every displayed value is calculated from the base pairing and the selected state modifier.
The model covers 48 combinations. It is not fitted or calibrated. Its job is to expose why the decision changes and to make the assumptions available for challenge.
INTERACTIVE MODEL
Route a failed payment
Change the payment mechanism, customer state, and recovery program. The model separates technical recovery from downstream subscription survival.
Showing Updater first for a dormant customer with reissued credential. Immediate recovery 74, 90-day survival 53, decision risk 15.
ILLUSTRATIVE INDICES / 0 TO 100
Recovery scoreboard
Did the charge go through?
This is the mechanism-led route.
Does the program match the payment mechanism?
Retention scoreboard
Is the subscription still alive at 30, 60, and 90 days?
Updater first21 points lost between recovery and day 90
Higher suggests the intervention is more likely to reopen the subscription decision.
Potential benefit
Account updater is designed to receive replacement credentials when available.
Where it can fail
There can be a gap before the new credential propagates.
Customer-state effect
Silent repair limits decision risk, but it does not solve the underlying lack of use.
All four programs for this payment and this customer
| Program | Recovery | Day-90 survival | Decision risk | Route fit |
|---|---|---|---|---|
| 5 | 0 | 20 | 7 | |
| 44 | 13 | 90 | 24 | |
| Mechanism-led | 74 | 53 | 15 | 98 |
| 85 | 55 | 48 | 78 |
WHAT THE ROUTES REST ON
- FACTStripe returns advice codes such as
do_not_try_again,try_again_later, andconfirm_card_data, and warns that unsuitable additional retries can be viewed as potential fraud. Stripe decline guidance - FACTVisa Account Updater lets participating issuers provide updated account information for credentials that expired or were reissued, lost, or stolen. Visa Account Updater
- MODELStarting routes rather than blanket rules. Coverage, consent, issuer advice, and service criticality can change the decision.