A weekly churn risk review routine your team will actually keep
Every CS team says it watches churn risk. In practice it's often a health score nobody trusts and a scramble when a renewal goes sideways. A simple weekly routine beats a perfect model you check once a quarter.
Pick five signals you can actually measure
- Ticket volume up 50% or more week over week
- Severity: any Sev1 or Sev2 open more than 3 days
- Usage: weekly active users down 20% or more
- Executive silence: a sponsor email with no reply in 5 business days
- Renewal pressure: renewal inside 90 days with health not Green
Flag an account only when two signals fire together. One signal is noise more often than not. Two is a conversation worth having.
How the two-signal rule plays out
| Account | Signals this week | Flag? |
|---|---|---|
| Marlow Retail | Tickets up 60% | No, one signal. Watch. |
| Halden Freight | Usage down 24%, CFO silent 5 days | Yes |
| Larkspur Labs | Sev2 open 5 days, tickets up 80%, renewal in 61 days | Yes, top of the list |
Use fixed date windows
Agents don't remember last week. Spell out the windows in the prompt: this week is Monday to Sunday just finished, compared with the seven days before. That one detail stops most "why did it flag that?" confusion.
A sample digest
Oct 5 to Oct 11 vs Sep 28 to Oct 4 | Read: CRM, tickets, inbox, usage 40 accounts checked | 3 flagged 1. Halden Freight (2 signals) | Owner: Sam Executive silence: Gina Hale (CFO), Oct 7, no reply in 5 days. Usage: weekly active users 410 to 312 (down 24%). Next step: reply to Gina today with usage numbers, offer a short review. Skipped: usage for Marlow (no rows in the usage data).
The header says what was read. The Skipped line says what wasn't. Without both, a broken data connection looks exactly like a great week.
The prompt
Every Monday at 8am, review all accounts in [CRM view] for [last Mon to Sun] vs the 7 days before. Signals: [your five]. Flag accounts with 2 or more. For each: signals with source references, owner, one next step. Header: date windows, sources read, accounts checked, count flagged. Footer: Skipped, listing anything you couldn't check and why. Never contact customers. Draft only.
Make the meeting short
Fifteen minutes on Monday. Each owner says what they'll do on their flagged accounts this week. Anything flagged three weeks running goes to the team lead. That's the whole ritual.
Checking 40 accounts by hand at two minutes each is about 80 minutes a week. Reading a digest takes about 10. Those are my rough timings, not a study, so track your own. The full Monday Churn-Risk Digest with guardrails and tuning notes is play one in The CS Agent Playbook. Related: AI prompts for customer success managers.
Questions
What are early signs of churn?
Ticket spikes, long open high-severity issues, falling usage, a silent executive sponsor and a close renewal with weak health.
How often should we review churn risk?
Weekly works for most teams. Monday is a good choice so the week's actions are clear.
Why flag only accounts with two signals?
One signal is often noise. Two together usually means a conversation worth having.
Want the full routines?
The CS Agent Playbook has 14 ready-to-paste agent routines like this one, each with guardrails, a sample result and the failure to check for first. $39 for the PDF and a plain-text prompts file.
Related guides
- Churn reasons: a reason code list and exit interview that tell you what to fix
- How to build a customer health score your team will actually trust
- Renewal prep at 90, 60 and 30 days: a checklist that keeps surprises out
- Churn risk email templates for when an account starts slipping
Free: prompt templates | churn risk calculator | churn rate calculator | CSM capacity calculator | escalation email generator | health score calculator | NRR and GRR calculator