Churn surprises aren't really surprises. In almost every case I've seen, when a customer churns unexpectedly, a retrospective look at their account history shows clear signals that something was wrong — signals that were present weeks or months before the decision to leave. The signals weren't invisible. They just weren't being looked at.
Churn prediction is the RevOps practice of systematically looking at those signals before it's too late to act on them. This post covers what the signals are, how to build a system that surfaces them, and what to do when an account is flagged as at-risk.
The Two Categories of Churn Risk
Churn risk signals fall into two categories: leading indicators (signals that appear before the churn decision is made) and lagging indicators (signals that appear around or after the decision). Lagging indicators tell you what happened. Leading indicators give you time to intervene.
Most CS teams focus on lagging indicators — NPS scores, renewal dates, explicit escalations — because those are the easiest to track. The companies that retain customers at high rates are tracking leading indicators: product usage patterns, support ticket frequency, stakeholder engagement, and time-to-value metrics.
The Leading Indicators Worth Tracking
Product Usage
Usage drop is the most reliable early churn signal in SaaS. A customer who was logging in daily and is now logging in weekly has changed their relationship with your product. A customer whose team has gone from 8 active seats to 3 is telling you something important.
If your product has usage analytics, set up alerts: any account where usage drops more than 40% week-over-week should trigger a CS notification. Any account that hasn't logged in for 14 days in a product that's supposed to be used daily is at serious risk.
Support Ticket Volume and Sentiment
A spike in support tickets often precedes churn — customers who are frustrated enough to reach out repeatedly are telling you their experience isn't meeting expectations. But the metric isn't just volume. A customer who submits a lot of tickets and gets them resolved quickly is often a highly engaged power user. A customer who submits tickets and gets slow or unsatisfying resolution is accumulating frustration.
Track support ticket resolution time and customer sentiment in support interactions alongside ticket volume. A rising volume of unresolved or negatively-rated tickets is a meaningful churn signal.
Executive Sponsor Changes
When the executive sponsor who championed your product leaves the customer organization, the risk profile of that account changes immediately. The person who fought for budget, who understood the ROI, who built internal adoption is gone. Their replacement has no skin in the game and may have preferred a competitor.
Track contact-level job changes for key accounts. Tools like LinkedIn Sales Navigator, Salesforce, or Bombora can alert you when a key stakeholder changes roles. Make it a CS protocol to reach out and establish a relationship with new stakeholders before renewal.
Non-Renewal of Multi-Year Contracts
A customer who renews annually on a multi-year contract and then declines to renew the multi-year is telling you they're evaluating the relationship. That's not necessarily churn, but it's a meaningful signal that should trigger a conversation about what changed.
Building a Churn Risk Score
Individual signals are useful. A composite churn risk score is more powerful — it combines multiple signals into a single view that CS can act from without having to check five different dashboards.
The simplest version of a churn risk score weights your leading indicators and produces a score for each account. A customer with declining usage (25 points) + two unresolved support tickets this month (15 points) + an executive sponsor change last quarter (30 points) has a score of 70 — which might put them in your "high risk" bucket requiring immediate CS outreach.
You don't need sophisticated ML to start. A weighted scoring model in a spreadsheet or CRM calculated field, reviewed weekly, is more effective than no score at all. As you accumulate more data on what signals actually predict churn in your customer base, you can refine the weights and eventually move to a model-based approach.
What CS Does With At-Risk Accounts
Identifying at-risk accounts is only useful if CS has a defined playbook for responding to them. Without a playbook, CS gets the flag and doesn't know what to do — or each CSM handles it differently, making it impossible to know what interventions actually work.
A basic at-risk intervention playbook includes: immediate outreach to the primary contact acknowledging any issues, executive outreach if the risk is high enough, a structured success review focused on value delivered vs. value expected, and a clear save offer if the customer is genuinely considering leaving.
Track which interventions happen for at-risk accounts and which accounts churn anyway. That data is your feedback loop for improving the playbook.
The handoff infrastructure that creates churn risk in the first 90 days is covered in the sales-to-CS handoff guide. The expansion opportunity that sits alongside churn risk is covered in the revenue leakage guide.
Build a churn prediction system that actually catches risk in time.
Take the free RevOps Scorecard to see how your customer health monitoring and CS infrastructure stack up.
Take the Free Scorecard →