Churn you see coming
Risk surfaces while the account is still savable, with the behaviour and the replay that explain it — not a percentage after the fact.
Where you are today
Revenue you already earned is walking out while you sleep.
Someone stops logging in and nothing tells you. Something breaks where you can’t see it. They cancel with two words that aren’t the reason. A card expires at 3am. By the time a chart adds it all up, the money and the customer are both gone.
Read-only Stripe access · no card required
Act I — Today
The signal nobody watches
Day 47 · no ticket, no complaint
She stops showing up, and nothing tells you.
Logins thin out. The integration she connected in week one quietly drops. Three of her five seats go untouched for a month. None of it reaches you, because none of it is the kind of thing anyone builds an alert for. The cancellation is six weeks away and already decided.
Account signals
UnwatchedWeekly sessions
week 3 → week 7
Seats active
3 never returned
Integration
day 31, silently
Support tickets
nothing to answer
No alert exists for any of this.
The ninety seconds you can't see
Day 51 · two sessions, 40 seconds each
Something broke for her. You'll never know what.
She opened the app twice, got as far as one screen, and left. Nothing threw an exception, so your error tracker has nothing. Support tries it and it works fine for them. All you are left with is a hunch, and a number at the end of the month.
Session · 14 Feb
Not capturedno recording available
Reproduced internally · works fine
Error tracker · nothing thrown
Last event logged · page_view
The cancel you never understood
14:41 · Tuesday the 11th
She cancels on a page you don't control.
A hosted billing portal takes it from there: one button, no questions worth asking, no offer, and a free-text box she fills with “too expensive.” It wasn't. Everything that would have told you the real reason was sitting in your own data, unread.
Hosted billing portal
What you gotWhy are you canceling?
“too expensive”
What was true
Stopped at onboarding step 3 of 6
4 sessions in 90 days
0 teammates invited
Never connected an integration
The night shift you don't work
03:12 · Tuesday the 4th
A card expires while you're asleep.
Stripe retries on its own schedule, sends a template nobody opens, and eventually gives up. A week later the subscription cancels itself. This customer was not unhappy at all — their bank declined a card and no one was awake to notice.
Invoice INV-4471 · $490
StripeAttempt 1
Tue 03:12card_expired
Attempt 2
Fri 03:12card_expired
Attempt 3
Mon 03:12card_expired
Subscription canceled
Thu 03:12unpaid
Nobody read a single one of these.
$0
MRR out the door this quarter · illustrative
You never had a churn problem.
You had a capacity problem.
Every one of those saves was possible. Each just needed somebody to notice in time, work out the real reason, and send the right message on the right day — for every account, every day, without stopping. That is a full-time job you have never been able to justify hiring for.
The turn
So you connect Stripe.
Read-only. Five minutes. That is the entire ask.
S
Stripe connected
Restricted key · read-only
Read-only to start
No code to install
Nothing sends without your say-so
Act II — From here on
And the night shift starts working.
Same four moments. Same timestamps. Someone is awake for all of them now.
Act II — With Churnless running
The signal, caught
Day 47 · no ticket, no complaint
The risk surfaces six weeks before she decides.
Churnless is already reading the behaviour you couldn't watch — sessions, seats, integrations, billing history, the tone of every reply — and scoring the account against how your customers have actually churned before. On day 47 it flags her, shows the evidence behind the score, and starts working the account while there is still something to work with.
Health score · Northwind Labs
Live34/ 100
High riskWhy, with the evidence
Sessions down 83%
week 3 → week 73 of 5 seats dormant
30 daysIntegration disconnected
day 31Renewal in 43 days
$1,200 MRRThe ninety seconds, on tape
Day 51 · two sessions, 40 seconds each
You watch the exact moment it broke.
Privacy-masked session replay, attached to her timeline and cued to the session the risk score flagged. She reached the import screen, waited eleven seconds for something that never resolved, and left. Nobody had to reproduce it, and nobody had to guess.
Session replay · 14 Feb
MaskedImport · spinning 11s
Cued to the session the score flagged
Inputs and PII masked before capture
Attached to her account timeline
The cancel, understood
14:41 · Tuesday the 11th
She cancels inside your product — and the flow already knows why.
Not a hosted billing page. The flow is embedded in your app, and it opens holding her whole history: the stall at step three, the eleven-second wait, the seats nobody used. So it doesn't reach for a discount she never asked about. It offers the help she actually needs, and it remembers every offer she has already turned down.
Cancel flow · in your app
EmbeddedBefore you go —
Imports stalled for you on the 14th. Want 20 minutes with an engineer to finish the migration — on us?
Book a sessionNo thanks, continue
Why this offer, not a discount
Replay shows the import stall, not price
Discount already declined in November
Score fell on usage, not on billing
The night shift, worked
03:12 · Tuesday the 4th
The card declines. Churnless is already on it.
It reads why the payment failed — an expired card, not an empty account — and schedules the retry for when that reason actually clears. The gentle email goes first, in your voice. The urgent one waits until it is earned. She updates her card at 8:15am from her phone, and neither of you thinks about it again.
Recovery plan · INV-4471
AgentRead the decline reason
card_expired · not insufficient_funds
Scheduled the retry
Thu 09:00, her local time
Sent the gentle email
Urgent tone held back
Card updated · $490 kept
08:15 — no ticket, no chase
$0
MRR kept this quarter · illustrative
Act III — Six months later
The part you notice isn’t the number.
It’s what stops being your job.
Saves that run without you
The cancel flow makes the right offer inside your product, and every failed payment gets worked on the day it is most likely to clear.
Your Fridays back
No exporting Stripe into a spreadsheet. No writing dunning emails one at a time. No guessing which accounts deserve the call this week.
A forecast you would defend
Every at-risk account carries a score, the evidence behind it, and a record of what has already been tried on it.
Your week
9hours a week
Chasing failed payments, chasing cancellations, and reconstructing why either happened. What replaces all of it is one message on Monday telling you what was saved and what it took.
Illustrative — your own before-and-after comes out of the audit.
What is running while you are not looking
- Health scores that move on real behaviour — sessions, seats, integrations, billing, feedback — with the evidence you can open
- Cancel flows embedded in your own product, that adapt to the account and remember every offer already turned down
- Privacy-masked session replay, cued to the sessions that moved the score, attached to the customer's timeline
- Payment recovery that fits the retry and the dunning tone to the actual decline reason
- An exit interview that asks the follow-up question and reports back what it learned
- Timelines, a REST API, and an MCP server, so your own agents can read all of it
Start here
Start on the honest end of this.
Connect Stripe read-only and take the leak audit first — your real failed payments, your real cancellations, your real exposure. Nothing installs and nothing sends until you decide it should.
Read-only Stripe access · no card required · your 14-day trial starts only when you turn protection on