Haa

The Lead Intelligence Engine transforms raw CRM data into actionable admissions intelligence, enabling teams to prioritize outreach and improve enrollment rates.

Admissions teams were manually reviewing thousands of leads with no systematic prioritization. High-value prospects received the same attention as unlikely conversions, resulting in poor conversion rates and wasted resources.

Engineered a feature-rich ML pipeline: extracted behavioral features from CRM data (response time, touchpoints, source channel), trained ensemble models (Random Forest + XGBoost) with precision-recall optimization, and built a scoring API for real-time lead ranking.

• Class imbalance (5% enrollment rate) required SMOTE and weighted loss functions
• Feature engineering from messy, inconsistent CRM data
• Calibrating model thresholds for business-appropriate precision/recall tradeoff

Optimized for recall (capturing all likely enrollees) rather than precision, aligned with business requirement to not miss potential students. Ensemble approach outperformed single models by 12% on F1.

• 78% precision, 85% recall on enrollment prediction
• 35% improvement in sales team conversion rate
• Reduced lead review time by 60%
• Analyzed 20,000+ lead records

Business context must drive model optimization choices. A technically "better" model optimized for accuracy was worse for the actual business problem.

• Real-time scoring API integration with CRM
• NLP analysis of lead communication quality
• A/B testing framework for outreach strategies