Client Project · Data Analysis & Predictive Modeling
A client running a senior living community had a sales problem: too many leads, no reliable way to tell which families were actually ready to move. Their sales team was spending equal time on cold inquiries and warm prospects. I built an AI-driven model to fix that.
Using NLP to process unstructured CRM notes from HubSpot and MatrixCare, I engineered intent-based features by extracting clinical urgency signals — phrases like "hospital discharge" or "immediate placement" — that reliably predicted conversion. The result was a propensity model scoring every lead from 0 to 1, with a 0.98 AUC — meaning it correctly ranked high-intent families at the top of the list nearly every time.
The practical outcome: sales counselors could sort their CRM by propensity score and focus their energy on the top 20% of the funnel, where the highest concentration of families ready to make a move actually lived. The model also identified email marketing as the highest ROI channel and flagged paid search keywords for refinement.
All data was fully anonymized and handled in compliance with HIPAA standards.
