Part of High-intent leads
Apply gates before points
Reject unsupported service scope, infeasible delivery and inappropriate contact contexts first. A strong phrase cannot compensate for these failures.
For remaining observations, assess need specificity, decision evidence and current status. Keep fit as a separate field. Do not use inferred company size to inflate purchase intent.
An illustrative ordinal rubric
| Dimension | 0 | 1 | 2 |
|---|---|---|---|
| Need clarity | Topic only | Problem described | Scoped outcome |
| Decision evidence | No decision stated | Options explored | Active provider or purchase request |
| Current status | Closed or obsolete | Unclear | Supported as open |
This rubric is an editorial example, not a validated model. A total of six does not mean a probability of purchase. Unknown status should trigger review, not be silently converted into strong evidence.
Record overrides and uncertainty
Store the source, reviewer date and reason for each rating. Allow a human override when new information changes the interpretation. Record why the override occurred.
A score should help organize review, not hide the evidence. If a teammate cannot explain why a request is ranked first, the model is too opaque.
Evaluate the rubric
Compare categories against later qualification and opportunity outcomes. Track false positives and review whether the weights serve the business’s actual decisions.
Do not repeatedly adjust weights to explain a tiny cohort. Start simple, keep criteria stable during an experiment and document meaningful revisions.