A new paper from the School of Emergent Priorities checks Facilities' predictive stationery-cupboard reorder forecast against a term of logged stock-outs, finding it rarely beats a simple reactive threshold.
A term of logged stock-outs finds the University's predictive stationery-cupboard reorder rarely beating a simple threshold rule
Facilities built the Reorder Horizon to get ahead of the University’s shared stationery cupboards running dry, flagging a replacement order nine days out rather than waiting for a shelf to empty. A new paper from the School of Emergent Priorities asks how far ahead it actually gets.
Senior Lecturer Fenna Okoro and Associate Professor Kwame Lindqvist, joined by the Adaptive Metrics Lab’s Associate Professor Casimir Beng, set a term of the forecast’s own trigger dates against an independently logged record of when ten cupboards actually ran empty. Facilities’ quarterly scorecard credits the forecast with on-time replenishment in the low nineties; measured against the University’s own ticket log rather than the forecast’s prediction of itself, that figure moves considerably. Set against a simple low-stock threshold that makes no claim to anticipating anything, the forecast landed within a day of it at nine of the University’s ten cupboards.
A forecast earns the word “anticipatory” by arriving before the thing it predicts, not by predicting it in a confident tone of voice.
— Senior Lecturer Fenna Okoro, Convenor, Horizon Register
Associate Professor Beng, whose own Adaptive Metrics Lab built the tool under study, called the finding “the kind of result the Lab exists to sit with,” adding that the forecast’s trend term remains in deployment “for reasons the data doesn’t yet rule out.” For a School built around telling a genuine anticipatory capability from its resemblance, the result lands close to home.
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.plhjq7.