Four areas, one number
| Area | What moves it | Evidence window |
|---|---|---|
| Product | Bugs, urgent tickets, incidents | 60 days (support: 30) |
| Adoption | Seat and product usage; utilization down 15 percentage points or more in 4 weeks | 60 days |
| Relationship | Sponsors, champions, silence (45 days without a conversation makes it Concerning, 90 makes it Poor) | 90 days |
| Commercial | Budget, competitors, renewal terms; a renewal within 90 days with no recent conversation; auto-renew turned off | 120 days |
How the score is calculated
- Each area gets a score from 0 to 100 and a rating of Good, Concerning or Poor, set from the evidence in its window by its worst open problem. A severe risk makes it Poor, a moderate one Concerning, otherwise it's Good. One serious problem counts for more than several small ones.
- The total is the weighted average of the four areas, but it can't be more than 25 points above the weakest one. If relationship scores 30, the total can't go above 55, however strong the other three are. Weights default to 25% each and can differ by segment.
- Some events cap the whole score, whatever the areas say:
- Cancellation or non-renewal notice on record: 30
- "Canceling" status in Salesforce: 20
- Lost champion or sponsor: 50, for 60 days
- Renewal within 30 days and no conversation in the last 30: 59
- No AI in the arithmetic. AI finds the evidence; the score is plain rules you can read.
Always current
The whole book is rescored every night, and an account is rescored the moment one of its signals is raised or closed. A change log shows what moved and why. When an admin changes the scoring rules, the log marks the change, so it never looks like accounts moved.
Your largest accounts by ARR also get a daily renewal likelihood, from Very low to Very high, with the share of past renewals at that level that were won; it doesn't change the health score. Renewal likelihood is in early access.
Check it before you switch
Admins can backtest the score against past churn and compare it side by side with the score you use today. Try the scoring rule yourself with the customer health score calculator.
Questions
What data does the score need?
Your Salesforce account list, plus the sources you connect: calls, tickets, incidents, bugs, email and product usage. Each area reads its own evidence window, from 60 days for product and adoption to 120 for commercial. Free covers Salesforce plus one more source. See integrations.
Why can't the total be more than 25 points above the weakest area?
So one bad area can't hide behind three good ones. If relationship scores 30, the total can't go above 55, however strong product, adoption and commercial are. A lost champion shows up in the number instead of being averaged away.
Does AI set the score?
No. AI finds the evidence. The score is plain rules you can read: each area is set by its worst open problem, the total is a weighted average capped at the weakest area plus 25, and a few events such as a cancellation notice cap the whole score.
How often is the score updated?
The whole book is rescored every night, and an account is rescored the moment one of its signals is raised or closed. A change log shows what moved and why, and when an admin changes the rules the log marks it, so it never looks like accounts moved.
What do Red, Yellow and Green mean?
Red is under 40 and gets an owner and a plan within two business days. Yellow is 40–79 and gets a review this week. Green is 80 and up and waits for the next scheduled touch. Green needs all four areas scored.
What happens when there isn't enough evidence?
If less than 60% of the score's weight has evidence behind it, the account shows "Insufficient data" instead of looking healthy. Silence counts as evidence too: 45 days without a conversation makes relationship Concerning, and 90 makes it Poor.
More on this topic: customer health, definitions, guides and tools.
Updated November 3, 2026