The Health Score That Lied

Health score: 78. Green. "Healthy," according to the dashboard your team checks every Monday. Ninety days later, the account doesn't renew.

Nothing about the number was wrong, exactly. Support tickets were normal. Usage hadn't cratered. On its own axis, 78 was a fair read. The problem is that a single rolled-up number can only tell you the average temperature of an account — and averages are exactly where compounding risk goes to hide.

Three quiet signals, one loud outcome

Here's the pattern we see most often, across AI chip companies, server OEMs, and data centers alike: no single event is dramatic enough to move a health score on its own.

• The champion who used to reply in an hour starts taking three days.
• A quarterly business review gets "pushed to next month" — and then pushed again.
• A competitor quietly starts a proof-of-concept inside the account.

Individually, a CSM might reasonably shrug off any one of these. Together, they're not three yellow flags — they're a pattern that has a name, a track record, and a next move.

A dashboard that only reads the number misses all of it. The account was never quiet. It was talking the whole time.

Why the pattern matters more than the number

A rolled-up score is built to answer "how are we doing overall," which is a fine question for a Monday standup and a bad one for renewal risk. Renewal risk isn't an average — it's a specific combination of things happening at the same time to the same account. Catching it means watching for the combination, not waiting for the average to drop.

This is the exact scenario that shows up most often in our own account base: a health score sitting comfortably above 70, a renewal inside 90 days, and a compound pattern building underneath it that the score can't see. It's also, not coincidentally, the trigger condition for what we call the Renewal Safeguard playbook — health below 70 within 90 days of renewal fires one response, but the accounts worth worrying about are often the ones that haven't dropped below 70 yet.

In the account we retained using exactly this pattern, the compound signal was flagged roughly sixteen weeks before the renewal date — early enough to run a real recovery motion instead of a last-minute save. The full $3.2M renewed.

What to watch instead of the average

You don't need a new dashboard to start thinking this way. Start with three questions for any account that "looks fine":

• Has engagement quietly slowed — response time, meeting cadence, attendance — even if no one's said anything is wrong?
• Has a scheduled touchpoint been deferred more than once, especially a QBR or renewal conversation?
• Is there any signal of competitive activity — a new stakeholder asking pointed comparison questions, a procurement process that started without you?

Any one of these, alone, is noise. All three, at the same time, in the same account, is a pattern — and patterns are catchable long before a health score ever moves.