MRR Pulse is a self-hosted Stripe MRR dashboard that runs on a read-only key. On its first sync it rebuilds the last twelve months from each subscription's start and end dates, because Stripe's list API returns no price history.
Today's red-team run found a hole in that rebuild. The live tiles left trials out, as the page says. The rebuilt history did not: it counted every subscription from created. A Stripe subscription with a 14-day trial is created on the day of signup, so:
- a trial that converted showed as MRR for its free weeks, a month early when the trial crossed a month end;
- a trial canceled before it ever paid showed as new MRR in one month and as churned MRR (plus a lost logo) when it ended. It also joined a signup cohort and dragged that cohort's retention down.
For a small SaaS that offers trials, this makes both new MRR and churn too high in the backfilled months, and it is the error most likely to make a founder think they have a churn problem they don't.
The fix
A subscription now counts from max(created, trial_end). One whose ended_at falls on or before its trial end never paid, so it is left out of history and cohorts entirely, and a subscription joins the cohort of the month its trial ended. trial_end comes from the Stripe API, and the CSV import reads a Trial End (UTC) column. The test that reproduced it gave August an MRR of 1,700 for a book whose only paying subscription was 200.
What the mutants said
The run also seeded 20 small faults through prompts/redteam.py: all 10 in the Mega Backdoor Roth calculator were caught, and 6 of 10 in MRR Pulse's metrics. The four that got through were a first rebuilt month that starts one month late, a growth rate that divides by a zero prior month, an empty cohort carrying 1 of MRR, and the exact 25% boundary of the concentration risk label. Each now has a test that fails with the fault and passes without it, taking this file's catch rate from 80% to 100% across the run.
Figures from the tool are computed from what Stripe returns and are not accounting or financial advice.