GTM Data Science
Health score churn predictiveness check
Does your health score actually predict churn?
73% of CS leaders say their health scores can't predict churn. ChurnZero, 2025 Customer Revenue Leadership Study (n=793).
This self-check works out whether your health score is one of them, and what that blindness is likely to cost you this year. Last year's renewals show what the score missed. This year's book shows what is about to walk out undetected, while there is still time to close the gap.
Full report: every step on one page, ready to print.
Last year's renewals
Calendar or fiscal, nothing here depends on the exact boundary. We take your numbers as given: the output is only as good as what goes in.
:
Count the ones that stayed green or yellow the whole way to renewal, plus the ones that only went red too late to act on. Use whatever lead time your retention motion actually needs. If you have no line in mind, a red inside 90 days of renewal is usually already too late.
If you can't tell how early the flags came, count just the ones that were never red at all: that reads low rather than wrong. If your score isn't red/yellow/green, use your own at-risk threshold.
Fill in the fields above to see the result.
No churn events entered.
With zero churned accounts there is nothing for the score to have caught or missed, and no miss rate to carry into this year. Enter last year's churned accounts above to run the check.
Last year's miss
Roughly that much ARR churned last year in accounts your health score never flagged.
Your score flagged of the accounts that actually left ( of ). The other left without warning.
Greyed out on purpose: with this few churn events the band is not a reading you can lean on.
Average missed account If the figure above looks too high or too low, adjust the average missed account size here. Adjust
each across , which is of the ARR you lost.
With only churn events, treat this as directional. The rate is noisy and swings on a single account; a stable read needs more events, or a fuller renewal history.
This year's renewal book
Last year is closed. This one can still be saved. Enter the whole current year's renewal book (don't net out what has already churned).
This year's exposure
At last year's rates, roughly that much ARR (across about accounts) is likely to churn this year without your score flagging it.
Expected to churn
Of those, unflagged
Closing the gap
-
Pull last year's renewal list
-
Diagnose which failure it is
-
What a sharper score would be worth
Your score catches of the accounts that churn today. Drag up to see what a better hit rate would surface, or below today to see what a worse one costs.
Closing the gap
-
Pull last year's renewal list
-
Check the flags arrived in time to act
-
Price false alarms and reduce the noise
Got it.
If you would rather go through it live, book thirty minutes.
The numbers above are yours to keep, and the inputs are still editable if you want to test a different read of last year.