BI-301h · Module 3

Automated Health Scoring

3 min read

Manual health scoring does not scale. An account manager who must manually assess twelve signals across six dimensions for each of their thirty accounts will do it once — and then never update it. Automated health scoring ingests signal data from systems that already collect it (CRM, support platforms, usage analytics, communication tools) and calculates health scores without manual input. The automation makes health scoring continuous rather than periodic and objective rather than subjective.

Do This

  • Integrate with data sources that already capture the signals — CRM meeting records, support ticket systems, product usage analytics, email systems — rather than creating new data entry requirements
  • Calculate scores daily and surface changes through alerts — the account manager should see score changes, not perform score calculations
  • Preserve manual override capability — the account manager may have information the automation cannot access, and their override should be logged but respected
  • Display both the automated score and the trend — the daily score is noise; the weekly and monthly trend is signal

Avoid This

  • Require manual data entry for health score inputs — any system that requires account managers to fill out forms will be abandoned within 90 days
  • Calculate scores so infrequently that they are always stale — a monthly health score calculated on the first of the month is two weeks out of date by mid-month
  • Remove manual override because "the model knows better" — the model knows the data; the account manager knows the context. Both matter.