A new Slop University paper decomposes 486,312 residential plumbing jobs and finds a franchise network's first-visit resolution rate climbing 4.8 percentage points while falling inside every one of the nine categories it is averaged over.
A decomposition of 486,312 jobs puts the whole 4.8 point improvement, and more, down to a change in which work got dispatched
A figure can rise for four years while every part of it falls. Slop University has published a case where it did.
Nothing Improved Except the Average, released today through the Office of Research Outputs, examines the first-visit resolution rate a national plumbing and drainage franchise network publishes for each of its territories: the share of jobs closed without anybody having to come back. Across fourteen quarters the network figure climbed from 71.4 to 76.2 per cent. Inside all nine of the job categories it is averaged over, it fell.
Nothing here says the work got better or worse. It says the arithmetic is answering a question about the diary, and the diary is not what anybody reads the number for.
— Dr Petra Umbile, Postdoctoral Fellow, School of Emergent Priorities
Fourteen quarters of the network’s own scheduling extracts underpin the result: 486,312 completed residential jobs across 214 franchise territories, sorted and flagged by the network rather than by the team. Apportioning the change by established decomposition methods puts 8.6 percentage points on the shifting mix of work dispatched and 3.7 points against performance within the categories themselves. The mix moved once an online booking form began asking customers to pick a job type before an appointment was offered.
“What we would like is for somebody to run this over their own reporting before we get to it,” said Dr Bram Ntuli, Lecturer and Convenor of Improvement Grand Rounds in the School of Continuous Improvement.
The same rate decides which technician is offered the harder work. The University regards the arithmetic here as ordinary and the deposit as timely, and would rather its indicator research be judged on how quickly such cases reach the open record than on how rarely they arise. The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.p1t27w.