The credibility problem
Every rep has seen it: a lead scored 94 that turns out to be a student doing research, while the VP who filled out the demo form sits at 41. Two experiences like that and the score becomes wallpaper — technically present, universally ignored.
Scoring fails when it’s built as a data project instead of a prioritization tool. The only question that matters: when two leads are in the queue, does the score pick the right one to call first?
Start with three signals, not thirty
The temptation is to weight everything. Resist it. Start with the three signals that your own closed-won history actually supports:
- Fit — are they the kind of company and role you close? Title band, company size, industry. This is the ceiling on the score; behavior can’t promote a bad fit into a good lead.
- Source — how did they arrive? A badge scan from a booth conversation, a pricing-page form, and a cold list import are not the same species of lead, and pretending otherwise is where most models go wrong.
- Recency — when did they last do anything? Interest decays fast. A hot signal from three weeks ago is a cold lead with good history.
Ship that. Add a fourth signal only when a rep can name the leads the current model ranked wrong.
Make the score explain itself
A bare number invites distrust. Show the why next to the score: “VP title · captured at DemoCon · visited pricing yesterday.” When reps can see the reasoning, they correct the model instead of abandoning it — “booth scans from that event were all students” is feedback you can encode.
Audit it quarterly, in one meeting
Pull last quarter’s closed-won and closed-lost. Check where they scored at entry. If wins clustered high and losses low, the model is earning its keep. If not, adjust the weights in the meeting, not in a backlog ticket that ships next year.
A useful score sorts the queue better than alphabetical order and keeps the team’s trust while doing it.
