Ask ten CSMs which number they trust more, their own gut or the health score, and nine will pick their gut.
That's not a hunch. It showed up almost exactly that way in a 30-day check-in I ran recently with a client, at the end of their first month on a new engagement. We surveyed their CSM team on what they actually trust when they need to know if an account is at risk. Nearly everyone pointed to their own relationship read: call tone, response speed, how engaged a champion sounded on the last call. Barely a third said the same about their official health score.
The easy read is that this is an adoption problem. Get the CSMs to trust the tool, run some training, tighten the scoring model, move on.
I don't think that's what's happening.
When almost every CSM on a team independently arrives at the same conclusion, that's not resistance. That's data. The team is telling you, in near unanimous terms, exactly what's missing from the model.
Here's why it's missing. Most health scores lean almost entirely on usage: logins, feature adoption, ticket volume, maybe an NPS score layered on top. Usage data is easy to pull automatically from the product, so it's what gets built first and weighted heaviest. Stakeholder relationships and business outcome realization take deliberate, ongoing work to track, so most companies quietly skip them, or leave them to a comment field nobody scores.
That's the gap CSMs are describing. A champion who's gone quiet, an executive sponsor who's stopped showing up to QBRs, an outcome that was promised at the sale and never actually realized, none of that shows up in a usage chart. It shows up in stakeholder management and business outcome tracking, which is exactly the layer a health score needs to actually predict renewal. Usage tells you whether people are logging in. It doesn't tell you whether the business case is holding, or whether the person who championed the deal still has the standing to renew it.
I've written before about health scores that go stale because nobody circles back to verify them. This is a different failure. That one is a discipline problem. This one is a design problem: a score built primarily on usage was never going to catch what the CSMs are catching, because the components that would catch it, stakeholder mapping, relationship strength, outcome realization, were never built into the model.
So the fix isn't asking your team to trust the dashboard more. It's building stakeholder management and business outcome tracking into the score itself, with the same rigor and the same weight you'd give usage. Not a sentiment field. A structured, tracked, scored input.
Because a score built mostly on usage isn't a health score. It's a usage score wearing a health score's name, and your CSMs already know the difference.