Approval and dunning.

Recurring revenue dies in two places: the authorization, and the retry you never designed.

Approval: +13.36%; Braintree to Stripe
Revenue: +$600K / mo; from the approval lift
Involuntary churn: −10%; custom retry strategy

The work

I led the Braintree to Stripe migration for a subscription product with more than three million users. Approval rate is not a dashboard number. It is whether the card the customer already gave you still works in their market, on that network, at that moment.

A 13.36 point lift is +$600K a month. That is what a processor change is for. Not a cleaner dashboard. Not a preferred vendor.

Failed payment as a state machine

Soft decline, hard decline, insufficient funds, do-not-honor, and issuer timeout are not the same event. Treating them as “payment failed” produces involuntary churn you will later call churn. The custom retry strategy cut that involuntary churn 10%.

I also built the discount and campaign platform that ran marketing across those three million users. Promos are payment-state problems: they have to be correct on capture, refund, and the next invoice.

Why I would rather route than migrate Resume