A new Slop University research poster compares scripted and unscripted fraud-risk interviews across 46 retail stores, finding confirmed-fraud catch rates hold flat between protocols while scripted stores report longer queue times.
Forty-six stores and fourteen weeks find a scripted fraud-risk interview catches confirmed fraud at the same rate as an unscripted one, and takes longer doing it
Forty-six stores, one fixed twelve-question script, and fourteen weeks of matched refund decisions are behind the School of Continuous Improvement’s newest research poster, and the finding undercuts the assumption behind most of them: a script standardises what a desk asks, not what fraud gives away.
A script tells the desk what to ask. It has nothing to say about what a scripted answer is worth once fraud has already decided not to give one.
— Dr Bram Ntuli, Lecturer and Convenor, Improvement Grand Rounds, School of Continuous Improvement
Dr Solveig Adeyemi and Dr Ntuli held each store’s protocol, scripted or unscripted, fixed for the full trial window, matching confirmed-fraud outcomes against loyalty-linked purchase history rather than staff say-so, and timing every interview from the fraud-risk flag’s first alert to the refund decision. Across the fourteen weeks and seven store-format clusters, the scripted and unscripted protocols caught fraud at statistically indistinguishable rates, while the scripted desks ran a longer queue in most of the formats measured.
“The desk was never short of confidence in the flag,” said Dr Adeyemi. “What it lacked was any evidence a script does anything the flag doesn’t already do on its own.”
Whether the twelve questions survive past the trial is now a rostering decision, not a research one; the School has flagged the queue-time cost for whoever makes it.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.vza4af.