The Churn Signal Hiding Outside Your Dashboard
The structural problem behind green accounts that churn, and what to build instead.
A CSM walks into a renewal conversation confident.
Usage looks stable. NPS came back at a 7. Three meetings logged last quarter. Exec sponsor replied to the last email.
Two weeks later, the account churns.
The CSM wasn’t incompetent. The health score wasn’t hacked.
It just measured the wrong things with complete precision. Nobody caught it until the cancellation notice arrived.
This is the most expensive blind spot in Customer Success right now. Most teams have no idea it's there until the cancellation notice arrives.
And when it does, the reason is almost always the same:
The health score was built to be easy to maintain, not hard to fool.
The Three Types of Broken Health Score
Name what you’re dealing with before you try to fix it.
Most broken health scores fall into one of three categories.
1. The Vanity Score
Built around engagement metrics because they’re easy to pull from the CRM.
Logins, email open rates, meeting counts, NPS. All visible. All exportable. None of them reliably connected to whether the customer is achieving anything.
The Vanity Score looks like a dashboard. It behaves like a decoration.
2. The Gut Feel Score
CSM-weighted, which means it reflects relationship warmth.
It’s a CSM’s greatest strength, and in this context, their biggest liability. The score trends green when the relationship feels warm. It trends red when the CSM is anxious.
This produces one consistent outcome: the accounts that churn are usually the ones nobody was worried about, because the relationship felt fine right up until it didn’t.
3. The Frankenstein Score
Built by committee.
CS wanted adoption metrics. RevOps wanted CRM activity. Product wanted feature usage. Leadership wanted NPS.
Everyone got a field. Nobody owns the logic.
The result is a score that changes meaning depending on who’s reading it and what narrative they need to tell.
Each one fails differently. But they all share the same root problem.
The Proxy on a Proxy Problem
A login doesn’t tell you if the user accomplished anything.
NPS tells you how someone felt on a Tuesday.
Email open rates tell you about your subject lines.
Support ticket volume tells you about your product’s friction, not your customer’s outcomes.
These are all signals that are one or two steps removed from the thing you actually care about: whether the customer is getting measurable value from the product.
When you stack proxies on top of proxies and call the result a health score, you get a number that is internally consistent but externally meaningless.
Green accounts churn. Red accounts renew.
And your CSMs spend 20 minutes pulling usage data, checking in with the AE, and reviewing account notes every time the score moves, because the score triggers a scavenger hunt instead of an action.
The signal problem is only half of it. The other half is timing.
Most health scores are built on lagging indicators. They tell you what already happened.
A customer can look perfectly healthy in your dashboard for 60 to 90 days while the real churn signal, the one a tool cannot capture, is already in motion.
What Qualitative Decay Actually Looks Like
The most predictive churn signals are not in your CRM.
They live in the texture of the relationship, and experienced CSMs feel them long before any dashboard reflects them.
Call it qualitative decay. It looks like this:
A champion who used to ask forward-looking questions about the roadmap stops asking. They stop pushing for new integrations. Their replies get shorter.
In QBRs, the energy shifts from “what’s next” to “this is fine”, and that shift from curiosity to comfort is one of the clearest early warning signals in Customer Success.
The internal advocate who used to cc colleagues on your emails quietly stops doing it. Their internal selling has gone dark. You still have access. You have lost reach.
The renewal conversation where the customer says “let’s just keep it the same.” That phrase sounds cooperative. It is inertia language, not value language.
A customer who is genuinely getting value from the product does not want to keep it the same. They want to do more.
These three signals have nothing in common with login data. They have everything in common with how a customer relationship actually ends.
The problem is that no health score in any tool today is built to capture any of this.
Which means the most reliable predictor of churn is sitting entirely outside your scoring system, dependent entirely on whether your CSM notices it and logs it somewhere.
A score is only as good as the CSM’s ability to act on what it cannot see.
This is why at-risk account frameworks need to sit alongside your health score. Not replace it, but cover the gap that the score structurally cannot fill.
What Actually Predicts Health
Stop asking “Is this account healthy?” Start asking, “Is this customer getting outcomes?”
The distinction matters because health is a state, and outcomes are a trajectory. A state can look fine while the trajectory is declining. Outcomes force you to answer a harder question.
Three outcome anchors that actually predict renewal direction:
1. The First Value Moment
Did the customer hit it, and how long did it take?
Time-to-first-value is one of the strongest leading indicators of long-term retention. Customers who take too long to reach their first meaningful win rarely recover the momentum.
2. The Recurring Value Motion
Are they using the features that deliver the outcome they bought for, consistently, not just at onboarding?
Shallow usage that covers only the most basic functionality is a risk signal even when the login count looks healthy.
3 The Expansion Signal
Are they asking for more, or just maintaining?
A customer who is genuinely getting value asks questions about what else is possible. A customer who is staying out of inertia goes quiet.
Underneath all three is a question most CS teams fail to answer cleanly at kickoff: did the customer ever define what success looks like?
If they didn’t, every health score you build is measuring activity against a blank target.
Fix goal identification at onboarding, and your health score immediately becomes more predictive, because now it has something real to measure against.
The structural issue with most health scores is this: they were designed to be reportable, not actionable. A score exists to answer one question: what do I do next?
If your score doesn’t produce a clear answer to that question, it’s not a health score. It’s a status update.
This connects directly to renewal risk systems that surface problems early. Goal clarity at onboarding is what gives your risk scoring its teeth.
You now know the three failure modes, the proxy problem, and what qualitative decay looks like before any tool flags it. What you don’t have yet is the system that runs it.
The architecture below is built into a ready-to-use Excel workbook.
Drop in your accounts, and your composite score, risk tier, and next action calculate automatically, including the qualitative signals your current tools can’t see.
If you manage a book of business and renewals are on the line, this is the tool that changes your Monday morning.
Upgrade to access the full architecture and download the template below→