# Coffee-queue ranking predictor fails to repeat at second site

> A new paper from the School of Emergent Priorities tests whether the coffee-cart queue-length index that appeared to lead the University's global ranking by two quarters holds at a second site, finding the correlation does not survive the move across campus while internal citations of the original pilot proposal keep climbing regardless.

A coffee-cart queue outside the Learning Commons made news on this site two
years ago: telemetry that appeared to lead the University's global ranking
by a stable two quarters, with a proposal to steer discretionary grant
strategy off the live number. A paper released this week ran the same test
at a second cart, on the Library concourse, to see whether the same queue,
somewhere else, tells the same story.

It doesn't. Across eighteen months of independent telemetry, the
correlation at the original two-quarter lag comes in at 0.09, against 0.78
at the first site, and no lag the authors checked clears significance.
Three robustness checks --- a term-time window, a different denominator, a
wider smoothing pass --- leave the null result exactly where they found it.

> A finding earns its keep by surviving somewhere it wasn't born. This one
> didn't, and a University trying to govern itself by evidence has to make
> room for that as readily as it made room for the original result.
>
> --- Professor Verity Marris, Director, Trajectory Analytics Group, School
> of Emergent Priorities

The paper's more striking finding may concern paperwork rather than queues:
internal committee papers kept citing the original pilot proposal at a
rising rate throughout the study, including after a briefing reporting the
non-replication did the rounds. Associate Professor Casimir Beng, whose own
earlier study proposed the pilot, said the result "raises exactly the
question a proposal should answer before it reaches a committee agenda a
second time, not after," adding he "would rather have had the answer
sooner."

The School of Emergent Priorities is treating the citation pattern, not the
failed correlation, as the harder problem to fix.

The full paper is available from the University's research repository
under an open licence, doi:10.5555/slop.gzh1b5.

Source: <https://slop.university/news/2026-07-15-it-didnt-cross-the-concourse/>\
Licence: [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)\
Cite as: Swift, B. (2026). _Slop University_ [Artwork]. https://doi.org/10.5281/zenodo.21199776
