Lead Propensity Model
Scores each lead's conversion likelihood across four quarter horizons (CQ/NQ/NQ+1/NQ+2) using engagement patterns and journey sequences.
4
Quarter Horizons Scored
Daily
Scoring Frequency
The Problem
Sales teams had large lead lists with no per-lead conversion likelihood ranking. Pipeline projections at the lead level used flat segment-level rates rather than individual characteristics.
What Was Built
XGBoost Classifier trained on lead attributes, activity/engagement patterns, journey sequences, funnel stage timestamps, and derived velocity metrics. Multi-quarter scoring: separate probability per quarter horizon. Feature selection via Feature Importance + Chi-squared + ANOVA F-test. Heuristic fallback for low-data customers using segment conversion rates.
Business Impact
Replaced flat conversion rates with ML-scored per-lead conversion likelihoods, feeding the Pipeline Projection Engine with more accurate record-level contributions.
Tech Stack
Domain Tags
Details
- Role
- Primary Owner
- Status
- Production
- Tier
- Tier 1
- Period
- Apr 2022 – Dec 2024
- Employment
- ADA Asia