CHURN IS DEAD
I Stopped Tracking Health Scores. Here's What I Found Instead.
7 min read · Strategy
Archive note: This issue predates the evidence ledger introduced in August 2026. Treat uncited benchmarks and examples as editorial analysis, not independently verified findings.
The $2M Account That Broke My Faith in Health Scores
The head of enterprise accounts pulled up the dashboard. Customer Health Score: 95/100. Usage trending up 23% quarter-over-quarter. Three successful QBRs. Active support tickets: zero. Expansion pipeline: $400K.
"This is our healthiest enterprise account," she told the executive team on Tuesday.
On Thursday, they received the cancellation notice.
Fifteen-minute call. New CFO. Vendor consolidation policy. $2M annual contract terminated effective next quarter. No negotiation. No renewal discussion. Just gone.
The health score never moved. It stayed at 95 until the contract ended.
This happened at a $200M SaaS company with a sophisticated CS platform, machine learning-powered health scoring, and dedicated customer success operations. Their "predictive" system had 18 months of behavioral data, engagement patterns, and usage analytics.
It predicted exactly nothing.
From Prediction Theater to Attribution Reality
I stopped believing in health scores that day. Not because the math was wrong, but because the problem was.
We're trying to predict the future when we should be understanding the past. We're building fortune-telling machines when we need autopsy reports.
Health scores excel at one thing: confirming churn decisions after they're made. They're rear-view mirrors painted to look like crystal balls.
The real breakthrough isn't predicting churn. It's understanding what you could have prevented versus what was always outside your control.
What Health Scores Actually Measure
Every CS platform vendor will tell you their AI-powered health scores predict churn 60-90 days early. Here's what they actually track:
Product engagement. Login frequency, feature adoption, API calls. Useful for identifying disengaged users, not accounts making strategic decisions.
Support interactions. Ticket volume, resolution time, satisfaction scores. Reveals product issues, not budget constraints.
Relationship health. Meeting cadence, response times, stakeholder changes. Captures communication patterns, not executive priorities.
Meanwhile, the factors that drive 60% of enterprise churn remain invisible:
- Budget cuts and spending freezes
- Executive turnover and strategic shifts
- Competitive displacement
- Billing failures and payment issues
- Regulatory changes
- Company acquisitions
Your health score algorithm doesn't know the new CFO has a mandate to cut SaaS spending by 30%. It can't predict that your champion just accepted a role at another company. It won't flag that your customer's biggest competitor just acquired them.
These aren't edge cases. They're the majority.
The Churn You Can Control vs. The Churn You Can't
Last year, I worked with a VP of Customer Success whose team was getting blamed for missing their retention targets. Their health scores were "85% accurate at identifying at-risk accounts." Their churn rate was still climbing.
We spent three weeks analyzing their last 50 churned accounts. Not their health scores. Not their usage patterns. The actual reasons customers left.
Here's what we found:
23% churned due to budget cuts. CS couldn't have prevented these. Economic conditions, company restructuring, and spending freezes don't show up in product usage data.
18% churned due to competitive displacement. A competitor offered better pricing or features. CS was never in the conversation.
16% churned due to billing failures. Credit card expirations, payment processing issues, invoice disputes. Pure operational problems.
12% churned due to executive changes. New leadership with different vendor preferences. CS relationships evaporated overnight.
31% churned due to preventable factors. Poor onboarding, lack of adoption, unresolved product issues, missing use cases.
The revelation: 69% of their "churn problem" wasn't a Customer Success problem at all.
Their team was optimizing health scores and running retention playbooks for churned accounts that were never saveable by CS. Meanwhile, the 31% they could have influenced received the same generic "high-touch" treatment as everyone else.
The Four Types of Churn
This analysis revealed a simple framework that changed how they approached retention:
1. Preventable by CS
Poor onboarding experiences, low feature adoption, unresolved use cases, relationship gaps, renewal process failures. These require CSM intervention and process improvement.
2. Preventable by Product
Missing features, performance issues, usability problems, integration gaps. These need product roadmap prioritization and engineering resources.
3. Preventable by Sales
Wrong ICP fit, unrealistic expectations, pricing misalignment, competitive positioning failures. These require better qualification and positioning.
4. Unpreventable (Market Forces)
Budget cuts, executive changes, company acquisitions, regulatory shifts, competitive disruption. These require acceptance and pipeline planning.
Once you classify churn this way, your team's priorities become crystal clear. Stop wasting cycles on unpreventable churn. Double down on the categories you can actually influence.
How Attribution Changes Everything
Six months after implementing churn attribution analysis, that VP's team hit their retention targets for the first time in two years.
They didn't improve their health scores. They stopped trying to.
Instead:
They focused CS resources on preventable CS churn. Better onboarding processes, proactive adoption programs, dedicated renewal management.
They escalated product churn to engineering. Feature requests became business cases backed by churn data. The product team finally prioritized integration improvements when they saw it prevented $800K in annual churn.
They adjusted sales qualification. When they proved that poor ICP fit drove 12% of churn, sales changed their qualification criteria.
They stopped blaming themselves for market forces. Budget cuts still happened, but they didn't trigger team retrospectives and process changes.
Most importantly, they shifted from reactive firefighting to strategic prevention.
Your 90-Day Churn Attribution Audit
Here's how to implement churn attribution at your company:
Week 1-2: Data Collection
Pull your last 50 churned accounts. Don't rely on CRM notes or internal assumptions. Call or email the actual decision-makers who left. Ask one question: "What was the primary reason you decided not to renew?"
Week 3-4: Classification
Classify each account into the four categories: Preventable by CS, Product, Sales, or Unpreventable. Be honest about what you could have realistically influenced.
Week 5-8: Pattern Analysis
Look for patterns within each category. Which CS-preventable churn happens during onboarding vs. renewal? Which product gaps appear most frequently? What sales qualification misses repeat?
Week 9-12: Resource Reallocation
Stop spending time on prediction. Start investing in prevention. Build playbooks for your highest-impact preventable churn categories.
The goal isn't perfect attribution. It's resource allocation based on reality instead of algorithm confidence scores.
Health scores will keep improving. AI will get better at pattern recognition. Platforms will promise more accurate predictions.
But until they can predict budget cuts and executive changes, they're measuring the wrong things.
The accounts you lose tomorrow were probably always going to leave. The question is: are you focused on the ones you can actually save?
By Kuber Sethi · All issues · Subscribe