Computer-Vision Housekeeping Verification Tool for a Multi-Property Hotel Group

Overview

Built a computer-vision verification tool for a multi-property hotel group's housekeeping team to confirm rooms meet cleaning and staging standards before being marked ready for the next guest. Housekeepers photograph a finished room, and the system checks against the brand's standard setup — checking key areas like made beds, cleared surfaces, and properly arranged amenities — flagging likely misses for a supervisor to double-check. This gave the group an objective, consistent quality layer across properties with different housekeeping teams and turnover rates, rather than relying entirely on spot-check inspections. The tool was explicitly scoped to flag likely issues for human review, not to certify a room ready on its own. A multi-property hotel group relied entirely on spot-check inspections to catch housekeeping quality issues, with different teams and turnover rates across properties leading to inconsistent room readiness. We built a computer-vision tool where housekeepers photograph a finished room, checking it against the brand's standard setup — beds, surfaces, amenities — and flagging likely misses for a supervisor to double-check, explicitly scoped to flag issues for human review rather than certify a room on its own. We defined the brand's standard room setup criteria together with housekeeping leadership, then trained the flagging model against a set of known-good and known-flawed room photos. It ran alongside manual spot-checks at one property for several weeks before supervisors relied on it as the primary review trigger. Guest complaints related to room readiness dropped as the group now catches an estimated majority of housekeeping misses before a guest ever checks in, instead of relying on periodic spot-check inspections.