Predictive Lead-Scoring Model for a Growing Realty Franchise

Overview

Built a custom machine learning model that scores inbound buyer and seller leads based on engagement patterns, listing price range, and historical conversion data from the franchise's CRM. The model runs automatically on every new lead and surfaces a priority score directly inside the CRM agents already use, so no new tool was added to their workflow. We trained and validated the model against two years of the client's historical lead data before rollout, then tuned it with agent feedback during a pilot phase. The goal was to help agents spend their limited follow-up time on the leads statistically most likely to close, rather than working every lead with equal effort. Agents were working every inbound lead with roughly equal effort, despite the franchise's own historical data showing that a relatively small share of leads accounted for most closings. We built a machine learning model trained on two years of the franchise's CRM history that scores every new lead and surfaces the priority directly inside the CRM agents already use, so no new tool was added to their workflow. We trained and validated the model against two years of historical lead and closing data before writing any integration code, then ran a pilot with a small group of agents to tune the score against their real-world judgment. Feedback from that pilot shaped the final scoring thresholds before the model rolled out franchise-wide. Agents now prioritize the leads statistically most likely to close instead of spreading equal effort across every inbound lead, improving response time on high-value leads without adding any new software to learn.