A new Slop University paper finds a food-delivery app's late-order notification wording shifts customer complaint blame onto the courier, without any change in how late orders actually run.
Complaint-ticket language names the courier twice as often once a food-delivery app's notification started naming them, though the lateness itself never moved
When a food-delivery platform quietly rewrote the message it sends about a late order, Slop University was already positioned to read what happened next in the complaints that followed. A new paper from the School of Continuous Improvement traces 2,614 complaint tickets across a fourteen-week window spanning the change: a generic “your order is running behind” replaced, market by market, by a notification that names the courier directly.
Associate Professor Casimir Beng, Lecturer Solveig Adeyemi, and Dr Mirela Hanke coded each ticket’s language for who it blamed, and found the courier named in close to half of tickets after the change, up from under a fifth before it, while the estimate itself drew proportionately less. Median lateness barely moved across the same window, and the rate of tickets escalated to a support agent held steady — only the language shifted, in step with a comparable change in which reason code closes out a refund.
A wording change that doesn’t touch the lateness and still moves who gets blamed for it is not, in our view, a coincidence.
— Associate Professor Casimir Beng, Lead of the Adaptive Metrics Lab, School of Continuous Improvement
“The courier’s name didn’t need to appear at all for the shift to hold,” said Dr Mirela Hanke. “Once the notification pointed at a person instead of an estimate, that seemed to be enough.”
The School of Continuous Improvement has spent this year tracing what a measurement does once it starts working for someone other than the person it was built to inform, and a notification that reroutes blame while leaving the underlying delay untouched reads, to the team, like the same story told from the customer’s side of the screen.
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.2udumk.