Eleven Weeks of Warning, None at the Register
Marit Osayande, Torun Ezeigwe, Petra Umbile · September 2026 · doi:10.5555/slop.nhgtil
A paper from the School of Continuous Improvement re-runs an earlier capability-drift index --- built to forecast exam-integrity workarounds with an 11.4-week lead --- unchanged on a five-state discount-grocery chain's loyalty-app data, testing whether it anticipates shoppers discovering a discount-stacking workaround at the self-checkout. Across 46 store clusters and 2,116,480 transactions over thirty weeks, the pre-registered replication finds no stable lead time (r = 0.09 against the original's r = 0.79), a null that survives an ablation removing store format, while the chain's promotional-terms revision clause continues to fire on the original's eleven-week cadence regardless.