Score leads on behaviour — and show the working

A growth marketer inherits a lead list and one question: who do I call first. A score answers it only if the score is trustworthy, and a score is trustworthy only when you can see what moved it.

HeadshotMarketing already models the score as a first-class field on the lead rather than a spreadsheet column. The leads view renders it as a bar per row, buckets the list into tiers, counts how many leads clear the high-intent threshold, and charts the distribution so a marketer can see whether the list is bimodal or flat. Sorting and filtering by score work today, and the score travels with the lead through conversion into a contact and a deal.

What does not exist yet is the derivation. The score is a value the team maintains, not a number the platform computes: nothing in the service reads a lead's email opens, page visits, form fills, or profile completeness and updates it. So the histogram is honest about the data it was given and silent about where that data came from.

The designed behaviour is a scoring pass that runs over the signals the platform already stores — campaign engagement, social interaction, content views, form submissions, and how complete the lead record is — writes a score with a per-signal breakdown, and keeps the previous value so a change is visible as a change. On top of that, a plain-language explanation: this lead moved from 48 to 76 because they opened three emails from the launch campaign and requested a demo. That explanation is the point. An unexplained model score gets ignored by the salesperson it was meant to help.

Threshold behaviour follows from it: crossing a score boundary should be a trigger, routing the lead into a nurture journey or onto a rep's desk automatically, with the reason attached so the handoff carries context.

This is roadmap, described as roadmap. The data model, the CRM surface, and the automation triggers it would feed are in place; the scoring engine and its explanations are not built yet, and nothing in the product should be read as claiming otherwise.