Scoring models fail socially before they fail technically. The first time a rep opens a 92-point lead and finds a student doing research, the score stops being used.
Make the inputs visible
Every score should be able to show its components: fit, behaviour, recency. When the number is wrong, someone can point at which part was wrong — and that is what makes it fixable rather than dismissed.
Decay is not optional
const score = fit * 0.5 + behaviour * 0.5 * Math.pow(0.9, weeksSinceLastActivity);Without decay, a burst of activity in March keeps a contact at the top of a queue in September. Decay makes the list reflect now.