Solar Installers, Satellite Solar + Savings Report

Solar installation companies in sun-belt states currently spending $1,500-3,000 per closed deal through door-to-door canvassers, lead gen aggregators, or low-converting websites. Ideal for companies doing 5-30 installs per month who want to cut their cost-per-acquisition in half.

A homeowner enters their address. The app pulls a satellite image of their property, runs a vision model on it to identify their roof's usable solar area, sun exposure, orientation, and panel capacity. It then visually overlays solar panels onto the actual satellite image of their roof, calculates their estimated annual production based on local sun hours, projects 25 years of savings against their current electric bill, and presents three financing options (cash purchase, loan, lease).

The homeowner books a paid consultation with a $99 deposit. The entire experience feels like sophisticated engineering software, not a sales form.

This build is more complex than the Roofing tool because of the financial calculations and the multi-step backend logic. Expect 12-18 minutes for the initial build. The AI builder will create the Entities, Functions, and Pages, then wire them together.

Likely follow-up prompts you'll need to send: - The panel overlay rectangles might not align perfectly with the polygon coordinates from the vision model. Send: "The panel overlay rectangles on the satellite image aren't matching the polygon_coordinates from the vision analysis. Adjust the SVG rendering so each panel zone is drawn at exact pixel coordinates returned by the vision model, scaled to the displayed image size." - The savings chart may render flat or wrong.

Send: "The 25-year savings chart isn't showing the divergence between 'cost without solar' and 'cost with solar' correctly. The 'cost without solar' line should compound at the annual_electric_rate_increase rate. The 'cost with solar' line should be the system cost spread across year 1 (cash) or monthly payments (loan), then flat.

Shaded area between them is the savings." - The vision model JSON parsing might fail intermittently. Send: "Add a retry loop to the analyzeRoofForSolar function: if the vision model response can't be parsed as JSON, retry once with the prompt prefixed by 'Your previous response was not valid JSON. Return ONLY the JSON object, no markdown fences, no commentary.'" - The state lookup table needs to be seeded with realistic data.

Send: "Populate the StateData entity with all 50 US states. Use realistic averages: AZ=6.5 sun hours, $0.13/kWh; CA=5.7, $0.27; FL=5.5, $0.13; TX=5.5, $0.14; NY=4.2, $0.21; etc. I'll provide a full table if needed."