# Tracking customer health: 7 signals that predict churn > Most customer health scores use the wrong inputs. Here are the 7 signals that actually predict B2B SaaS churn, and how to track them without engineering work. Published: 2026-05-22 · Author: BlueHill Team · Tags: churn, customer-health, playbook A good customer health score predicts churn 60–90 days before it happens. Most don't, because they over-weight product usage and ignore the human signals. ## The 7 signals (ranked by predictive power) ### 1. Champion departure Far and away the strongest signal. When the buyer leaves the company, churn risk doubles. Track exec-sponsor changes via LinkedIn or email-bounce signals. ### 2. Drop in interactions Not just product usage — total interaction volume (emails, calls, support tickets, portal logins). A 50% drop in interactions over 30 days predicts churn 80%+ of the time. ### 3. Onboarding milestone misses If the customer didn't hit Day-7 / Day-14 / Day-30 milestones, year-one churn risk is 3–5× higher than the cohort baseline. ### 4. Reduced exec engagement Exec sponsor stops attending QBRs or accepting meeting invites. Customers don't downgrade communication when things are going well. ### 5. Support escalations A spike in support volume — especially escalations to "speak to a manager" — is a leading churn indicator even when CSAT remains decent. The escalations represent friction, not just unresolved tickets. ### 6. Failed expansion conversation You asked about adding seats / tier upgrade / new module. They said "we'd love to but..." — that's a hedge that often becomes "we're not renewing." ### 7. Procurement / contract pushback Late payment, requests to renegotiate terms mid-contract, or new procurement-team involvement at renewal. These are bureaucratic signals of decision-maker discomfort. ## Signals that *don't* predict churn (despite being popular) - **Pure feature usage** — useful but lagging; can be high for soon-to-churn customers who are still onboarding new users - **NPS** — too noisy, too gameable, too lagging - **Login frequency** — heavy-users can still churn; light-users can still renew ## How to build a health score with these 7 signals Weight them: champion departure 25%, interaction drop 20%, onboarding misses 15%, exec engagement 15%, escalations 10%, expansion hedges 10%, procurement signals 5%. Update weekly. Anything that moves to amber/red triggers a save play: outreach to the new champion, exec-to-exec meeting, value scorecard sent. ## How BlueHill helps The customer record aggregates the interaction signals (emails, calls, notes, escalations) automatically. Onboarding milestones are first-class objects. Status reports surface trending-red accounts every Monday. ## Related reading - [Customer health score glossary](/glossary/customer-health-score) - [What is churn?](/glossary/churn) - [How to scale CS without hiring](/blog/scale-customer-success-without-hiring)