Research outputs

The published research outputs of Slop University.

All research outputs of the University are published open access under CC BY 4.0, with persistent identifiers minted by the Office of Research Outputs.

Published outputs

Poster

Naming a Shared Machine Does Not Extend Its Service Life

Fenna Okoro, Anouk Mensah, Kwame Lindqvist · August 2026 · doi:10.5555/slop.ioerp9

A research poster from the School of Emergent Priorities asks whether the human names given to shared working machines --- the club shed's ride-on mower, the bakery's dough mixer, the hot-water urn on the bench --- do anything to how long those machines stay in service. Asset-register extracts, service dockets and acquisition invoices for 1,806 items at 214 workplaces and community clubs are analysed by two-stage least squares, with naming instrumented on the rated pronounceability of each item's manufacturer model designation. The naive association of +3.6 years does not survive identification: the instrumented estimate is +0.2 years (95 per cent confidence interval −2.6 to +3.0), and no equipment class retains a positive effect at conventional levels. The poster reads the result as reverse causation, machines earning names by lasting rather than lasting by being named, and notes the subsequent removal of 1,412 recorded equipment names from the Horizon Register as unverified attributes.

Poster

Recovering the Unwritten Roster

Runa Adegoke, Anneke Tolan, Petra Umbile · August 2026 · doi:10.5555/slop.e9bpba

A research poster from the School of Emergent Priorities fits a hidden semi-Markov rotation model to 9,806 Student-of-the-Week lines from 38 primary-school newsletter archives (118 classrooms, 2022--2025). A fixed rotation with occasional recorded swaps reproduces the complete award sequence in 71% of classrooms, letting the recipient be named a median of nine weeks before the certificate is presented; sequences pause for cancelled assemblies and grandparents' day and resume without skipping a name, and certificate density roughly doubles in the final fortnight of the year as classrooms clear every unrecognised name before the holidays.

Poster

Benchmarking the Barista

Torun Ezeigwe, Anouk Mensah · August 2026 · doi:10.5555/slop.g8sr99

A research poster from the School of Continuous Improvement reports the first accuracy benchmark for the counter guess --- the barista's unprompted "the usual?" --- across 41 franchise cafés, 214 tracked regulars and 6,187 transactions over 14 weeks. Scored against the till record the guess reaches 91% top-1 accuracy; re-scored against intended orders elicited off-premises it falls to 68%, with roughly one confirmed usual in four reflecting acquiescence rather than preference. The loyalty app's your-usual prediction, trained on the till record, reproduces the barista's errors, and the partner network has since adopted a recognition-rate service standard scored from the same confirmations.

Poster

A National Baseline for the Tethered Counter Pen

Ingrid Vasseur, Runa Adegoke, Petra Umbile · August 2026 · doi:10.5555/slop.xcxdxw

A research poster from the School of Emergent Priorities reports the first population estimate of first-attempt legibility for the pens tethered to public and commercial service counters. A stratified multi-stage probability sample of 3,180 counters, each observed independently by two fieldworkers over eleven months with disagreements adjudicated blind, puts first-attempt legibility at 78.4% (95% CI 76.9-79.9); a further 13.4% of pens write only after priming, and 7.4% of sampled counters carry no tethered pen at all. The estimate has been lodged with the Indicator Commons as a first-wave baseline, against which a target of 85% is now set.

Poster

Does the Wording Matter, or Only the Hour?

Lindiwe Achterberg, Runa Adegoke, Petra Umbile · August 2026 · doi:10.5555/slop.m2muby

A research poster from the School of Emergent Priorities lodged 1,344 matched requests for small plumbing and electrical work with 336 sole-trader businesses across four metropolitan regions, randomising three content attributes householders are commonly advised to attend to alongside the clock window in which each request was sent. No content attribute moved the reply rate by more than 1.4 percentage points, while requests lodged between 16:20 and 17:00 drew a reply 3.1 times more often than the study mean.

Paper

Locating the Efficiency Frontier of an Equipment-Hire Network

Renke Sabel, Mirela Hanke, Sten Okwuosa · August 2026 · doi:10.5555/slop.36mdpr

A paper from the School of Continuous Improvement estimates input-oriented technical efficiency for all 61 depots of one national equipment-hire group across four consecutive quarters, correcting the finite-sample bias in envelopment estimators with a smoothed homogeneous bootstrap. Mean bias-corrected efficiency is 0.812 (95% CI 0.786 to 0.838) and 19 depots sit on the frontier, 14 of them in the lowest fleet-size tercile; the ten depots most often named as peers hold a median of 14 machines against a network median of 63. An ablation removing wash-bay hours from the input set moved mean efficiency by 0.004 and changed no depot's frontier status. Over the following quarter the six largest depots, issued reference sets and convened in a capability sprint, reduced their fleets by a median of 14 machines; their mean efficiency rose from 0.681 to 0.771 and their completed hires fell by 8.4 per cent.

Paper

Towards a Discrete-Event Model of Container-Refund Depot Throughput

Fenna Okoro, Anouk Mensah, Solveig Adeyemi · August 2026 · doi:10.5555/slop.nkxa2l

A paper from the School of Emergent Priorities and the School of Continuous Improvement builds a discrete-event model of the over-the-counter container-refund transaction from 14,208 timed handovers at eleven privately operated depots, decomposing service time into presentation, opening, counting and settlement stages. The opening stage --- in which the load is made available to the operator --- accounts for a mean 31.4% of handover duration and 54.9% of its variance, and across a sweep of 10,000 replications per configuration a second counting lane returns 6.0% more throughput against 23.3% for moving the arrival stream to its fastest-observed presentation class. An ablation removing the arrival-batching component shifts mean throughput by 0.5% (n.s.) and is retained for completeness.

Paper

Does a Shorter Interval Hold the Second Half?

Thandiwe Solberg, Sten Okwuosa, Joost Nwosu · August 2026 · doi:10.5555/slop.h6skfa

A paper from the School of Continuous Improvement reports a pre-registered stepped-wedge cluster-randomised trial in which eleven volunteer-staffed community theatre venues crossed over from a twenty-minute to a fifteen-minute advertised interval across one season, with two observers per performance timing 4,118 returns to seat over 34 performances. Return by the five-minute mark was 96.2 per cent under the longer interval and 96.0 per cent under the shorter, an adjusted difference of 0.2 percentage points that falls inside the pre-registered equivalence bound; bar takings per patron, late returns and a bell-protocol ablation moved no further, and return times measured backwards from curtain-up were near-identical in every wedge step. Curtain-down came 5.1 minutes earlier, the only effect to reproduce at every venue. The Indicator Commons has since ratified a Standard Interval Duration of twenty minutes and added interval observance to each venue's annual attestation.

Paper

Multi-Objective Fixture Construction for a Suburban Darts League

Verity Marris, Ingrid Vasseur, Anouk Mensah · August 2026 · doi:10.5555/slop.2q3v7p

A paper from the School of Emergent Priorities formulates one metropolitan darts association's 18-round season fixture as a fourteen-term multi-objective integer program, eliciting the terms from twenty seasons of fixture-committee minutes and solving 10,000 weight vectors drawn from the fourteen-simplex. No term varies by more than 4.6 per cent of its own optimum between the 5th and 95th percentile of draws, the median solution shares 92.6 per cent of its assignments with the one the ratified weights produce, and lifting the single-night availability window at one venue is worth 3.4 times the travel spanned by the entire weight simplex.

Paper

Repeatability and Reproducibility of the Pre-Move Volume Estimate

Casimir Beng, Torun Ezeigwe, Kwame Lindqvist · August 2026 · doi:10.5555/slop.9ae0mt

A paper from the School of Continuous Improvement and the School of Emergent Priorities runs the removalist's walk-through volume estimate through a fully crossed gauge repeatability-and-reproducibility study: eight estimators from independent operators each quoted the same twelve furnished households three times, blind to their own earlier figures, against a reference volume measured from the loaded vehicle at pack-out. Of the variation in 288 estimates the estimator and its interaction with the household account for 59.9%, against 35.4% for the household itself; combined repeatability and reproducibility reaches 80.4% of total study variation and the system resolves one distinct data category, so the eight estimators' means on the same twelve houses run from 21.1 to 36.7 cubic metres. Mean bias is +3.70 cubic metres, a bedroom-count rule of thumb tracks the reference more closely on all three agreement measures, and restricting the panel to estimators with ten or more years in the trade moves the headline figure by 1.8 percentage points (n.s.).

Paper

SLUMP

Solveig Adeyemi, Iben Chikere, Thandiwe Solberg · August 2026 · doi:10.5555/slop.k0jc8n

A paper from the School of Continuous Improvement describes and deploys SLUMP, a specimen-lineage service that carries concrete test cylinders from domestic and small-commercial pours --- the carport slab, the shed floor, the shopfront ramp --- through five attested custody events to the crush bench. Across nine months and 214 participating sites the service handled 1,388 cylinders and 6,940 custody events, raising lineage completeness from 17.2 per cent under the incumbent paper docket to 94.6 per cent. An ablation removing the tamper-evident seal module left 28-day compressive strength distributions indistinguishable (F(2,1385) = 0.41, p = 0.66), and 61.2 per cent of owners requested a lineage certificate that carries no standing with any authority.

Paper

CAUTION

Petra Umbile, Fenna Okoro, Kwame Lindqvist · August 2026 · doi:10.5555/slop.hgk1nh

A paper from the School of Emergent Priorities describes CAUTION, a sensing platform that fits each portable folding hazard cone with an ultra-wideband tag, a duty-cycled accelerometer and a hinge-state switch, and decodes its position and deployment state against the tenancy floor plan. Across a fourteen-week deployment of 412 tagged cones and 61 anchors at three regional shopping centres, the decoder reaches a macro-F1 of 0.941 against 486 hours of direct observation, with median map-matched positional error of 0.58 m. The 8,614 recovered episodes are strongly bimodal on a log scale, with a median of 41 minutes, a 90th percentile of 6 h 21 min and a longest episode of 34.2 days; removing the accelerometer channel changes macro-F1 by 0.004, and the trace is released under an institutional schema intended to support settlement of facilities-services contracts on cone-hours.