School of Emergent Priorities researchers put a price on the per-line shelf-life forecast a supermarket chain runs its markdowns from, and find its ceiling sits below what it costs to produce.
A grocery chain's evening markdown forecast, valued against the fixed six o'clock rule it was bought to improve on
Three researchers spent half a year putting a price on a piece of software, and the price is the interesting part. Perfect Foresight, Same Stickers, released today by the School of Emergent Priorities, values the per-line shelf-life forecast one grocery chain runs its evening markdowns from, and values it the way an appraiser would: by what it changes about the decision it was bought to steer.
Dr Petra Umbile, Dr Runa Adegoke and Associate Professor Kwame Lindqvist worked from 1.18 million line-days of chiller and bakery records across 41 stores, setting the forecast-led markdown run beside a fixed six o’clock rule and simulating both across the whole range of forecast skill. Even perfect information about the following day’s sell-through, the team reports, would be worth $18.40 per store per week against $61.20 to produce the forecast; at the skill the chain actually achieves, the run does slightly worse than the rule.
We kept being asked whether the forecast was any good. That turned out to be the smaller question. The one worth the fieldwork is whether a better answer would have changed which trays got stickered.
— Dr Petra Umbile, Postdoctoral Fellow, School of Emergent Priorities
Associate Professor Lindqvist, who directs the Master of Priority Studies, said the appraisal “belongs a good deal earlier in the life of an instrument than anyone tends to put it”, and welcomed the questions the work raises about the University’s own commissioning habits. The University counts the study as an addition to its practice of appraisal as much as to its record of publication.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.mjevxm.