Roof Replacement Lead Gen from Satellite Imagery

This regional roofing company employed 8 canvassers who knocked on 200+ doors daily across suburban neighborhoods. Their close rate hovered around 2%—meaning 98% of conversations went nowhere. After storms, they'd blanket entire areas hoping to find damage, wasting weeks on properties that didn't need service. Sales reps were burning out, and customer acquisition costs were unsustainable.

We trained a convolutional neural network on thousands of labeled satellite images to identify visual indicators of roof distress: missing shingles, discoloration patterns, sagging sections, moss/algae growth, and storm damage signatures. The model scores every rooftop in a target geography and exports ranked lead lists with property details, owner information, and estimated roof age. Sales reps now knock only on doors with 70%+ replacement probability.

Conversion rate jumped from 2% to over 8%. The team reduced canvassing hours by 65% while generating more closed deals. Customer acquisition cost dropped by more than half. After major storms, they can now identify affected properties within hours instead of weeks.