How a Hotel Chain Cut Food Waste Costs by Tracking Every Plate
WasteLens replaces manual bin audits with computer vision — staff photograph waste, and the system identifies items, estimates quantities, and calculates financial impact in real time.
15% accuracy vs actual weight
The Real Cost of Guessing About Waste
Hotels and restaurant chains know they waste food. What they don't know is exactly what they're wasting, when, and how much it costs. The industry standard is bin audits — someone literally sorting through garbage — and estimation worksheets. Both are labor-intensive, inconsistent, and produce data that's weeks old by the time anyone reads it.
A hotel chain with 50 properties couldn't compare waste patterns across sites. Each property had its own spreadsheet, its own methodology, and its own definition of "acceptable waste." Corporate sustainability teams were flying blind.
A Camera at Every Waste Station
WasteLens replaces bin audits with a camera at each waste station. Staff take a photo of waste before disposing. Gemini 3.1 Pro Vision identifies items, estimates quantities, looks up unit costs, and produces real-time dashboards showing waste by category, shift, day of week, and location.
We tested both Gemini and GPT-4 Vision for food identification. Gemini outperformed on three dimensions: recognizing partially consumed items, weight estimation accuracy (within 15% of actual), and handling Spanish food vocabulary natively — critical for our Latin American markets.
The platform uses a four-level multi-tenant hierarchy (Program → Account → Site → Location) with role-scoped JWT tokens, supporting the B2B2B business model across waste consultancies, hotel chains, and individual properties.
“Gemini handled food items described in Spanish — pozole, chilaquiles, pan de muerto — without requiring English translation.”
From Weeks to Hours
Kitchen managers now identify waste patterns within hours, not weeks. The hotel chain can compare performance across 50+ sites from a single dashboard — no bin sorting required.
The system processes images asynchronously to avoid blocking kitchen workflows. Results flow into dashboards that sustainability teams actually use, because the data is current, consistent, and comparable across every property.
April 3, 2026