A new Slop University paper builds a checkout-history score to flag which tool-library members are likely to return equipment late, then tests it against a volunteer roster's own blind shortlist, finding the two name the same members about as often as they disagree, and score just as well as each other by season's end.
A season-long, three-branch tool library trial finds a checkout-history score and a volunteer roster's blind shortlist tied on precision, agreeing on half their names
A season spent testing a checkout-history score against a volunteer roster’s own hunches has left Slop University’s tool libraries about where they started: trusting the hunches, provisionally, a little more carefully.
A shortlist that matches the roster isn’t nothing — it’s a start on knowing which half of the ledger the roster was already carrying.
— Dr Renke Sabel, Senior Lecturer and Convenor, Indicator Commons, School of Continuous Improvement
The School of Continuous Improvement team built a Return Punctuality Score from each member’s own checkout history at three neighbourhood tool libraries, then asked each branch’s volunteer coordinator, before either list was compared, to name the members they expected to chase up over the coming season. Across a full 22-week season, the score’s shortlist and the coordinators’ blind shortlist caught late returners at statistically indistinguishable rates and named more than half of the same people. Stripping the score’s single most heavily weighted factor out afterwards left its performance exactly where it started.
“None of this happens without three branches’ worth of volunteers willing to hand over a shortlist they knew might be checked against a formula,” said Dr Osei Vandermeer, Research Fellow. “That’s the part of the season we’re proudest of.”
The University reads the result as its evaluation agenda intends such results to be read: not a verdict on the score, but a record of what checking one, honestly, actually costs.
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.dq8md6.