A new Slop University research poster re-derives an estate agency group's six-segment buyer model on 48,206 open-home registrations and finds the segments unstable across quarterly refreshes and no more predictive than attendance alone.
A cluster-validity check on 48,206 open-home registrations finds five of the six segments below the threshold at which a cluster is treated as dissolved
A segment is supposed to be a group of people who resemble one another more than they resemble everybody else. The School of Continuous Improvement has now held one set of six against that definition.
Six Names for the Same Crowd, a research poster released today through the Office of Research Outputs, re-derives the buyer segmentation an estate agency group runs across its 27 sales offices. Led by Dr Sten Okwuosa, Senior Research Fellow and Convenor of the Impact Pathway Atlas, the study applies the group’s own model to 48,206 open-home registrations gathered over fourteen months. Under resampling, median cluster stability sits at 0.40; on held-out registrations it falls to 0.31, with five of the six segments below the level at which a cluster is conventionally read as dissolved. Between consecutive quarterly refreshes, 46.8 per cent of registrants move segment. Added to a count of how many open homes a person attended in the previous ninety days, segment membership shifts the model’s discrimination from 0.673 to 0.681, an interval the poster reports as containing no effect.
Nobody set out to sort strangers into six kinds. Somebody chose six because six fits on a screen, and everything since has been the model doing as it was told.
— Dr Mirela Hanke, Postdoctoral Fellow, School of Continuous Improvement
The group’s own reporting records 99.6 per cent of registrations as allocated to a segment across the period, and books each quarterly re-derivation as a coverage improvement. “An indicator that only moves when somebody re-runs the model is reporting on the model,” said Dr Renke Sabel, Senior Lecturer and Convenor of the Indicator Commons.
Slop University offers the poster as an instance of the School’s standing question about a category: whether anything answers to it.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.yrnz28.