How attribution works
The full rules behind every proof report, in plain words. Every monthly report will show the numbers that went into them. The price is a flat $49 a month and never depends on these numbers.
The paid recovery service is not open yet, so there are no proof reports yet. This page describes how they will work. Email hello@renewvane.com to hear when the paid service opens.
The problem we are solving
When a subscription payment fails, several things try to fix it at once. Stripe retries the card. Stripe may automatically update a reissued card through the card networks. The customer may notice and fix it on their own. And our emails ask them to. If we simply counted "paid after our email", we would be taking credit for all of it.
Step 1: random assignment
When recovery is on, every invoice that newly fails is assigned at random, independently, before anything is sent:
- Emailed group, about 85% by default: gets our recovery emails.
- Holdout group, about 15% by default: gets no emails from us at all. We do not even look up the subscriber's email address. Everything else stays as it was: your Stripe retries, Stripe's own emails if you use them, automatic card updates, your in app banners.
You can choose a holdout share from 10% to 50%. It can never go below 10%. The assignment uses a secret random key per account, so nobody, including us, can steer an invoice into either group. It is recorded the moment the invoice is first seen and never changed afterwards, even if you later change the percentage; a new percentage only applies to invoices that fail after the change.
Step 2: the evidence gate, invoice by invoice
An emailed invoice counts as evidence checked only when all of these are true:
- The subscriber clicked our link. A click only counts when a real browser presses through our page, so email security scanners that open links do not count.
- The invoice was paid within 7 days after that click.
- The payment was made through the page we opened for them, in one of two ways: with a new card that was added no earlier than the click and no later than 2 hours after it, and that never failed on this invoice; or by completing their bank's 3D Secure confirmation on the Stripe invoice page we sent them to.
Anything else is recorded and not credited to us: a charge on the card that already failed (an automatic retry), a card that existed before the click, a card added days later, a payment with no click, a click after the payment, a click more than 7 days before it, or an invoice paid outside Stripe. Automatic retries on the old card never count.
Step 3: only matured invoices are compared
An invoice is matured once it is 21 days past its failure. That covers our 9 day email sequence, the 7 day click window, and a margin for quiet hours. Both groups get exactly the same 21 days. A payment made later than that counts for neither group.
Step 4: the group comparison and the safety margin
For each group we work out the share of failed dollars that came back within 21 days, over a trailing window of at least 60 days ending at the end of the month. If the holdout has fewer than 20 matured invoices in that window, the window grows in 30 day steps up to 180 days.
Measured lift = emailed group's rate minus holdout group's rate.
Even if our emails did nothing, the two rates would differ a little by chance, and taking credit for that noise would be misleading. So we subtract a safety margin of 1.645 times the statistical uncertainty of the lift. In plain words: we only claim lift we are about 95% sure is real.
Cautious lift = measured lift minus the safety margin. If that is below zero, it counts as zero.
Step 5: cautious added recovery
Cautious added recovery = the smaller of (a) the evidence checked payments for the month and (b) the month's emailed failed dollars times the cautious lift. US dollars only. The report sets it next to your $49 plan price.
The group check stops us claiming what Stripe or your subscribers would have done anyway. The invoice check stops us claiming lift we cannot tie to a subscriber actually using our link. The report credits us with whichever is smaller.
If even 180 days do not hold 20 matured holdout invoices, we cannot measure lift. Then the report says “not measured yet” and shows the raw counts instead of a dollar figure. We never claim recovery we have not shown our emails added.
Step 6: provisional, then final
A month's report stays provisional until every invoice that failed in it has matured, which is 21 days after the month ends. Then it is final and never changes. The report is never a bill: your plan is a flat $49 a month whatever it says.
What you see every month
- Count and dollars of matured failed invoices in each group, and how many are not yet matured
- Recovered dollars and rate in each group over the comparison window, and the window's dates
- The measured lift, the safety margin and the cautious lift
- The evidence checked payments, the cautious lift in dollars, the cautious added recovery and the value compared with your plan
- Notes such as "extended window", "not measured yet" and "non USD excluded"
Your data export lists every recovered invoice ID, its group, and whether it passed the evidence check and why, so you can check each one in your own Stripe dashboard.
Honest limits
- The report usually understates the real effect. The safety margin is large when accounts are small. That is the trade we chose.
- Stripe does not label a payment as an automatic retry. We infer that a subscriber acted from the card being new or from the bank confirmation, not from a Stripe flag.
- Final reports do not change. A refund or late payment after a report is final does not change it.
- Changing your Stripe settings mid month makes the comparison less clean. Tell us first.
- US dollars only. Other currencies are listed but not included in the comparison.
Questions about a report? Reply to it. A person reads every reply.