Sales opportunity prioritization
- Industry
- Sales
- Type of work
- Predictive modeling
- Period
- 2026
Architecture
Goal
Decide which opportunities in the sales pipeline deserve effort, by estimating win probability and expected amount.
What I did
- Validated and cleaned 78k opportunity records, finding they were snapshots rather than history.
- Binned features with WoE/IV and checked monotonicity before modeling.
- Trained win-probability and amount models with XGBoost and LightGBM, explained with SHAP.
- Simulated prioritization and channel scenarios, and presented results in Quarto.