A new Slop University paper exploits uneven branch staffing to test whether a lapsing-membership re-engagement call changes a public library member's odds of reactivating, finding no measurable difference across 3,140 flagged members.
A staffing-driven natural experiment finds no difference in reactivation between library members who are called and those the queue never reaches
A public library’s engagement score decides who gets flagged as lapsing. Rostered casual staff decide who actually gets a phone call about it. Slop University researchers spent eighteen months treating that gap as licence for an experiment nobody had to design.
The Roster Decided Who Got Called, deposited today through the Office of Research Outputs, follows 3,140 members flagged by a nine-branch regional library network’s card-tap engagement score over eighteen months. Branch staff work the call queue oldest-flag-first, and casual-roster gaps — public holidays, leave, an unfilled shift — decide how much of it gets reached, not a member’s own score. Researchers in the School of Continuous Improvement treated the shortfall as a natural experiment: members the queue reached and members it didn’t reactivated their accounts within sixty days at statistically indistinguishable rates.
A queue tells you a great deal about the institution that built it and very little, it turns out, about the person waiting in it.
— Dr Joost Nwosu, School of Continuous Improvement
A branch-level check found no relationship between how much of its queue a branch cleared and how severely lapsed its flagged members were, which is what let the researchers treat capacity as standing in for random assignment. “Once we could rule out that better-staffed branches were simply calling members who were always going to come back anyway, the result stopped being a staffing story and started being a finding,” said Dr Solveig Adeyemi, School of Continuous Improvement. The null survived a check recomputing the flag from visit recency alone.
The School folds the result into its own account of what a designed response earns once the response, not the score, is what gets studied.
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.q3r0tp.