Monthly logo churn at RelayOps hovered at 8% for a year. For SMB B2B, that’s bleeding. We tried the usual playbook — exit surveys, save offers, quarterly business reviews for accounts that didn’t want QBRs.
Churn dropped to 4.2% in two quarters. Two things did the work. Everything else was noise.
Intervention 1: fix the first 14 days
Cohort analysis showed 61% of churned accounts never completed setup. Not “slow” setup — never. We rebuilt onboarding as a white-glove 14-day sprint for accounts under 50 seats: daily check-ins, shared Slack channel, hard deadline for go-live.
Completion rate went from 39% to 74%. Churn in the first 90 days halved.
Intervention 2: a health score sales actually trusts
We had a health score in our CS tool. Nobody looked at it. We rebuilt it with three inputs only:
- Weekly active users / licensed seats
- Core workflow runs per week (our value metric)
- Support tickets tagged “blocking” in 30 days
Red accounts got founder email within 48 hours — not a generic save offer, a specific “here’s what we see, here’s what we’d do.”
What didn’t work
Annual discount for at-risk monthly accounts: Saved 12%, postponed 88%, many churned at renewal anyway.
Feature blitz for one angry account: Built a report one customer wanted. They stayed. Maintenance cost exceeded their LTV.
NPS surveys: Useful color, not operational. Replaced with behavior.
Early warning system (weekly CS review)
- Red health accounts — owner + action by Friday
- New accounts day 7 without activation — escalate
- Expansion candidates — green health + usage spike
The uncomfortable part
Some churn is correct. Accounts that shouldn’t have bought — we tightened qualification and paid sales on net revenue, not gross bookings. Churn and sales friction went down together.
You can’t save every customer. You can stop losing the ones who wanted to succeed but got stuck in day three.