A School of Continuous Improvement impact report finds a home-service worker's completion photo barely outpredicted no photo at all, and tracks how fast a live check-in script travelled once platforms started asking for one instead.
Five years of the School of Continuous Improvement's work on the completion photo, the live check-in that followed it, and the group chat that got there first
Senior Lecturer Renke Sabel’s team spent the better part of two years doing something no home-services marketplace had thought to do on its own: checking whether the completion photo a worker uploads before payment actually predicts a job got done. The School of Continuous Improvement’s Impact Report 2021–2026, published today by the Office of Research Outputs, sets out five years of that checking, and what happened once the live video check-ins it recommended started catching what the photograph missed.
The photo, the report finds, tracked a service category’s ordinary completion rate about as well as no photo at all would have. Live check-in review, introduced across five categories from 2023, closed some of that gap and opened a new one: across the platforms studied, a script for passing a live check-in convincingly reached a worker’s own group chat in a median of eleven days, a regularity the School now reads as evidence that a check and the workaround to it arrive on much the same schedule.
We had built an entire appeals process around a photograph. Nobody had built anything around the job.
— Senior Lecturer Renke Sabel, Convenor, Indicator Commons
The University reads the finding as consistent with the audit brief the School set itself at founding: recommend an indicator, then keep checking it after somebody else adopts it. “Two weeks isn’t long enough to stop a script from circulating,” said Research Fellow Osei Vandermeer, whose Review Cadence Observatory built the early-warning model behind the report’s forecasting work. “It is long enough to stop being surprised by it.”
The full report is available from the University’s research repository under an open licence, doi:10.5555/slop.ub5opu.