Skip to content

Documentation

Which of my Stripe customers are about to churn?

Kicklace marks a paying customer as likely to churn for one of four reasons: the subscription is set to cancel, a payment failed and Stripe is still retrying, nobody on the account has signed in for thirty days, or, where there are no sign-ins to go on, nothing at all has happened for thirty days. It is a rule over facts Kicklace already holds, and it says which of the four is true rather than guessing a score.

The rule, and the real numbers

Four reasons, checked in this order, and only for somebody who is paying right now: cancelling — a live subscription is set to stop at the end of its period; payment_failed — Stripe is still retrying a card that failed; no_sign_in — nobody on the account has signed in for thirty days, where Kicklace has sign-ins to go on at all; quiet — nothing at all has been recorded for thirty days, where there is nothing but the timeline to go on. Thirty is one constant the whole rule shares. A customer nothing has ever been recorded for does not wear the chip: unmeasured is not the same as quiet, and a spreadsheet of old customers should not read as a thousand people about to leave.

The same rule is a population, Likely to leave, on the Populations page — its count, its members, and the exact fact that put each one there: "No activity in 34 days; the rule says 30." It has one number you can change, how many quiet days count, and it can only be raised past thirty, never lowered, because thirty is already the floor the rule underneath it keeps. Say it in a sentence — "wait 45 days instead of 30" — and every member is recomputed against the new number within a minute.

Kicklace also turns the same facts into one line on the person's own record: "$49 a month is at risk: the card failed after paying since Aug 2 and Stripe is still retrying." — with the figures it is built from listed beside it, never printed alone. See Payments from Stripe for how the subscription itself gets here, and the lifecycle for the four states — paying, trial, churned, free — the rule reads before it ever asks about churn.

Set it up

  1. Connect Stripe: Connections → Stripe → Connect, paste the address into Stripe's own webhook settings, bring the signing secret back.
  2. Paste a read-only restricted key into the wizard's third step and press Read the history — this scores the customers you already have, not only the ones who pay from today on.
  3. Open People, narrowed to the Likely to leave population, or open the Populations page for the whole list with why each person is on it.
  4. If thirty days reads too aggressive for your business, open that population and change the number in a sentence.
  5. Write an automation that answers it — "when somebody enters Likely to leave, email them" — or press Automate this straight from the finding.

What it does not do

It never predicts a probability or a score: it names which of the four facts is true, or it says nothing. It never emails anyone by itself — a send is either a person's own, or an automation a person switched on. And it needs something to measure: a customer with no sign-in data and nothing on their timeline is never marked as leaving, because nothing about them has been recorded at all.