Portfolio Pulse

Every account starts at 100 points. Five signals — days since contact, an overdue-payment flag, ticket load (weighted up if a lot of those tickets were high-priority), seat utilisation, and active-user rate — subtract points on a continuous ramp: the deduction grows smoothly between a "clean" and a "bad" threshold, so an account at 89 days since contact scores almost the same as one at 91 days — never a cliff-edge jump. A missing signal contributes zero deduction, never a penalty; thresholds are also segment-aware (Enterprise tolerates longer gaps than SMB).

Summing those five deductions gives the Risk score — the raw, unweighted signal read, independent of timing. Health is calculated from the same deductions but adjusted by three multipliers that account for context: renewal proximity (×1.25 if renewal is within 90 days — the same problems matter more when the contract is about to turn over), tenure (×1.15 if the account is under 6 months old, ×0.9 if over 3 years — a new relationship is more fragile), and momentum (×1.15 if spend is shrinking, ×0.9 if it's growing). Risk and Health are deliberately different numbers — an account can carry real risk signals (say, a high ticket count) but still show strong Health if its renewal isn't imminent and it's a long-tenured, growing account; conversely, the same risk signals right before a renewal date will drag Health down harder than Risk alone would suggest.

The portfolio's headline number is the ARR-weighted average Health score across every account, so a struggling $500k account moves it more than a struggling $10k one. Every weight and threshold is a plain constant in scoring.py — no ML, nothing hidden.