A new Slop University paper charts eighteen months of end-of-shift cash-drawer counts at 34 venues, finding a two-dollar investigation trigger operating inside common-cause variation and a closure-rate indicator unrelated to anything about cash handling.
A hospitality group's fixed two-dollar trigger sits well inside the variation its drawers produce on their own, while the slow drifts it was written to catch stay underneath it.
The School of Continuous Improvement has spent several years on a question its name barely conceals: what an institution’s own settings are actually detecting. Two Dollars Is Not a Signal, published today, puts that question to a rule that runs several thousand times a week.
The paper charts eighteen trading months of end-of-shift cash-drawer counts at 34 suburban hospitality venues under one cash-handling procedure: 37,208 shift closes across 68 counted drawers, each variance above two dollars escalated to the duty manager and closed before the next trading day. The trigger sits comfortably inside the ordinary spread of the counts. Of the 11,682 closes it flagged, 98.5 per cent carried no evidence that anything had changed; of the 1,046 genuine departures in the same data, 85.9 per cent stayed underneath it. One drawer ran light for 34 consecutive shifts without once crossing the line.
We went in expecting to argue about where the line should sit. We came out understanding that a line of this kind cannot be put in the right place, and that has changed how this School talks about its own settings.
— Associate Professor Casimir Beng, Lead of the Adaptive Metrics Lab, School of Continuous Improvement
Because every escalation must be closed, the proportion closed on time becomes a venue indicator, which the paper reports as unrelated to the state of the drawers and moderately related to how much they vary. “A setting nobody has revisited is not necessarily the wrong setting,” said Dr Joost Nwosu, Postdoctoral Fellow in the School of Continuous Improvement. “It is an unexamined one, and we would like many more examined.”
The work was supported through the Indicator Stewardship Seed Fund. The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.iluyix.