04 · Innovation Lab
Lead Scoring Simulator
Change a hypothetical lead. A predictive model re-scores it instantly and shows which factors moved the number.
Illustrative: A demonstration on synthetic data. Not a real system or result.Synthetic weights that demonstrate the technique. This is not a reproduction of any proprietary or production model.
Tier A · Route to sales now
Assign to the territory owner with a 1-hour response SLA and a pricing-ready task.
Typical lead: 13% · This lead: 7.4×
- Requested pricing+1.35
- Industry fit+0.75
- Lead source+0.75
- Timeline+0.55
- Recency+0.36
- Role+0.35
- Site engagement+0.29
- Company size+0.15
- Email opens+0.02
- Budget confirmed0
Values are log-odds. Green pushes the score up, amber pulls it down.
Activity-only score
34/100
Counts clicks and opens.
Predictive score
93/100
Weighs fit, intent and recency.
Quiet on activity, strong on fit and intent. Activity scoring would miss this one.
How it works
A logistic model
Each characteristic adds to or subtracts from the lead's log-odds of converting. The total passes through a sigmoid to become a probability. In production the weights are learned from won and lost history; here they are set by hand.
Explainable by design
Because the model is additive, every score breaks down into contributions. The bars compare this lead with a typical one, so a rep can see why it ranks where it does and challenge it if it looks wrong.
Scores drive action
A score only matters if it changes what happens next. Tiers map to routing rules: who gets the lead, how fast, and which nurture track it enters.
Why activity scoring fails
Points for clicks and email opens reward curiosity, not buying intent. A student researching the category can out-score a decision-maker who visited once and asked for pricing. Weighing fit and intent alongside behaviour fixes that; the comparison panel shows the difference.
What changes in a real deployment
Weights are trained on the CRM's own outcomes, validated on held-out data, monitored for drift, and written back to the CRM so routing and reporting use the same number. See the predictive scoring project.