BI-301h · Module 1

Health Score Normalization

3 min read

Raw health scores are not comparable across accounts without normalization. A technology startup with weekly executive meetings has a different engagement baseline than an enterprise bank with quarterly executive reviews. If both are scored on the same absolute scale, the startup will always appear healthier than the bank — not because the relationship is stronger, but because the baseline behavior is different. Normalization adjusts scores relative to the account's own baseline, its segment peers, or its historical trend.

Do This

  • Normalize engagement signals against the account's own historical baseline — a customer who typically meets monthly and shifts to quarterly is showing the same decay signal as one who typically meets weekly and shifts to monthly
  • Segment normalization for accounts of similar size, industry, and relationship maturity — comparing a 90-day-old relationship against a 5-year relationship produces meaningless scores
  • Apply trend normalization to distinguish between stable scores and volatile scores — a consistently 70 is healthier than a score that oscillates between 50 and 90

Avoid This

  • Use raw absolute scores for cross-account comparison — the CEO of a 50-person startup responds to emails faster than the CEO of a 50,000-person enterprise, and that is not a health signal
  • Normalize so aggressively that all scores cluster around the midpoint — the normalization should preserve meaningful variation while removing structural bias
  • Skip normalization because "simpler is better" — unnormalized scores systematically misallocate intervention resources toward accounts with naturally high baseline engagement