A School of Continuous Improvement field trial finds a supermarket loyalty system's manual confirmation step has no measurable effect on whether a lapsed member ever shops again, and the chain is extending the same audited step chain-wide regardless of the finding.
A 38-store trial finds a supermarket loyalty system's manual override step changes nothing, and the chain is expanding the audit around it anyway
A loyalty program’s automated systems make a great many decisions on a retailer’s behalf, and the School of Continuous Improvement’s newest trial checks what happens when a human is asked to make just one of them instead.
Released today, the trial covers 38 stores of one supermarket chain over a 26-week test of the chain’s own churn-alert system, LERT, which flags a lapsing loyalty member and drafts a reactivation reward once their card goes 40 days unscanned. Half the stores kept LERT’s automatic release; the other half required the loyalty-desk supervisor to confirm the reward with a tap on the EFTPOS terminal first.
A confirmation step earns its place by changing an outcome. Ours didn’t, and there’s no version of this write-up where we quietly stop mentioning that.
— Dr Joost Nwosu, Postdoctoral Fellow, School of Continuous Improvement
Ninety-day reactivation among flagged members came back at 6.1 per cent under the confirmation step and 6.3 per cent without it — a gap the University’s analysis treats as no gap at all, despite more than six thousand taps recorded across the trial.
The step’s real achievement, said Dr Dagny Okafor, Lecturer and Convenor of the School’s Evaluation of Evaluation program, was never about the members — it was “a fortnightly audit that now has a reason to keep running, whether or not the case for it still holds up.”
The chain is proceeding regardless: the confirmation tap is being formalised into a fortnightly, footage-cross-referenced reconciliation audit, extended to every store from next quarter. A result this flat, the University argues, is worth publishing exactly as flat.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.m6vevy.