Lead Prioritization Engine

Context

A B2B SaaS company with a high-volume inbound motion was drowning in leads. Their sales team of 50 reps received thousands of new leads weekly, but had no systematic way to prioritize beyond basic firmographic rules.

The Problem

Simple scoring rules (company size, industry) produced too many false positives. Sales reps spent hours qualifying leads that never converted, while high-potential leads sat in queue. Conversion rates were flat despite increasing marketing spend.

Why generic AI wouldn't work

Generic lead scoring tools use shallow signals like email opens and page views. For complex B2B sales, intent signals are nuanced — a VP visiting the pricing page twice has different meaning than an intern doing competitive research. The company's unique sales motion and ICP required custom modeling.

The System We Designed

• Multi-signal feature engineering from CRM, product analytics, and web behavior
• Gradient boosting model trained on 2 years of historical conversion data
• Account-level aggregation to capture buying committee signals
• Dynamic scoring that updates in real-time as new signals arrive
• Integration with CRM to surface scores and explanations to reps

Human-in-the-Loop & Explainability

Scores come with explanation of top contributing factors. Reps can flag scoring anomalies through the CRM interface. Weekly review sessions identify patterns the model is missing, feeding back into feature engineering.

Outcomes

• 3.2x improvement in conversion rate from top-scored leads
• Sales cycle reduced by 23% for high-score cohort
• Rep satisfaction increased (less time on unqualified leads)
• Marketing gained clear signal on which channels produce quality