Research papers

Research papers published by 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

Paper

Eleven Seconds at the Gate

Torun Ezeigwe, Marit Osayande, Sten Okwuosa · August 2026 · doi:10.5555/slop.c62txw

A paper from the School of Continuous Improvement times 1,147 gate gear checks across 18 rallies at nine pony clubs in one metropolitan-fringe zone, recording the number of riders waiting at the start of each check and re-checking every fourth horse-and-rider pair against a 14-item defect schedule fixed before the season, by a second accredited assessor blinded to the gate outcome. Check duration falls an estimated 3.1 seconds for each additional rider waiting (95% CI 3.6 to 2.6), from a median of 46 seconds at queues of two or fewer to 11 seconds at queues of nine or more; detection follows duration, with checks over 30 seconds recording 58 per cent of the defects the blinded re-check confirmed, checks under 15 seconds recording 11 per cent, and the pooled figure across all observed checks reaching 31 per cent. Every one of the nine clubs returned between two and five times its gate-recorded defect rate on re-check, while the association's compliance indicator held between 99.1 and 100 per cent across all 18 rallies and showed no relation to detection at any of them. Adding rally start-time punctuality to the model moves the queue-length coefficient by 0.4 per cent.

Paper

Was Like That When We Moved In

Thandiwe Solberg, Ingeborg Nwachukwu, Marek Solheim, Petra Umbile · August 2026 · doi:10.5555/slop.8ezzbp

A paper from the School of Continuous Improvement codes 61,884 free-text descriptions from 4,312 rental entry condition reports lodged with one property-management group over six years, developing a nine-category frame whose largest category --- condition described by reference to a previous occupant, the age of a fitting, or the state in which the property arrived --- accounts for 28.4 per cent of what tenants write. A mapping panel of the group's own staff could place 81.2 per cent of that category on none of the six lines the bond is settled against, leaving 30.6 per cent of all coded description mass unassignable; after the free-text field was replaced with a drop-down built from those six lines, the category fell from 27.9 to 4.6 per cent of recorded content while the volume of content rose.

Paper

The Price Nobody Pumps

Ingrid Vasseur, Kwame Lindqvist, Anouk Mensah · August 2026 · doi:10.5555/slop.1qy5e3

A paper from the School of Emergent Priorities pairs 12,431 board-price observations across all 34 retail fuel sites in one metropolitan region with the same sites' wet-stock reconciliation of litres dispensed by grade, over eighteen weeks. The grade a pole board carries in its largest numerals accounts for a median 31.4% of litres sold per site (IQR 24.8 to 38.1), and at no site does it reach half. Recomputing the region's published price-competitiveness index with grade weights drawn from dispensed volume rather than from the advertised grade alone displaces sites by a median of 7 rank positions of 34, moves 11 of the 34 more than ten places, and drops the site published as the region's most competitive to fourteenth, with agreement between the two orderings at Kendall's tau-b 0.58. Displacement concentrates among sites with the widest spread between their cheapest and dearest posted grade, and three pre-registered restrictions of the sample shift median displacement by no more than 0.4 places.

Paper

Neither of Us Has Ever Raised It

Ronja Oyelaran, Lindiwe Achterberg, Runa Adegoke, Fenna Okoro · August 2026 · doi:10.5555/slop.8ihxgf

A paper from the School of Emergent Priorities surveys both owners of 512 shared-driveway pairs across four metropolitan councils, obtaining complete pairs for 318 of them, and recodes each owner's free-text answer on who would arrange resurfacing into the party it nominates so that the two answers within a pair can be compared directly. The two owners nominate the same party in 34.0 per cent of pairs (Cohen's kappa 0.12, 95% CI 0.05 to 0.19), each nominates the other in 41 pairs, and at least one nominates the council, which holds no maintenance duty over the strip, in 58. Two raters working blind to the survey agreed on the same driveways' surface condition substantially better than the owners agreed on responsibility, and 57.0 per cent of discordant pairs' driveways rated poor or failed against 24.9 per cent of concordant pairs'. Both owners judged the surface to need work within two years in 194 pairs, and in 163 of those neither reported ever having raised it, while the four councils' registers record 19 shared-access matters over six years and close 18 of them. Excluding every pair containing a non-resident owner moves concordance by 0.6 percentage points.

Paper

Everyone Passes, Differently

Torun Ezeigwe, Marit Osayande, Solveig Adeyemi · July 2026 · doi:10.5555/slop.czhkle

A paper from the School of Continuous Improvement treats the 24-item observation checklist behind an accredited first-aid practical assessment as a measurement instrument, calibrating 9,148 completed assessments from six registered training providers over nineteen months under a dichotomous Rasch model. Nineteen of the 24 items are endorsed on 97.8 per cent or more of sheets, item difficulties span 1.1 logits, and person separation reliability is 0.19; in a pre-registered substudy of 612 assessments independently re-observed by a blind marker, the same items span 4.3 logits and separation rises to 0.74. Eleven items show differential item functioning across providers, four beyond the conventional large-effect threshold, and re-equating on the 13 DIF-free items reorders the providers whose first-attempt competency rates a workplace-safety scheme publishes side by side, the highest-rated placing fourth of six and the lowest-rated second (Spearman's rho -0.31, 95% CI -0.88 to 0.60). Six pre-specified model variants move person separation by no more than 0.04.

Paper

Six Minutes, Unless You Press

Marit Osayande, Dagny Okafor, Anneke Tolan · July 2026 · doi:10.5555/slop.e6c924

A paper from the School of Continuous Improvement runs a pre-registered matched mystery-shopper audit of the two channels by which a customer raises an assistance request on a hardware store floor --- 32 trained shoppers, 1,164 scripted requests across 19 stores of one metropolitan network over fourteen weeks, crossing channel with request type under an incomplete block design, alongside a structured observation of 2,918 naturally occurring requests. Median time to first substantive assistance is 1 min 52 s via the in-aisle call button and 6 min 41 s via direct approach (difference in medians 4 min 49 s, 95% CI 4 min 02 s to 5 min 36 s), while the button accounts for 8.1 per cent of observed assistance events. Across the 19 stores the chain's published service-availability score, which feeds its rostering formula, is not appreciably associated with the audited direct-approach wait (Spearman's rho = 0.09, 95% CI -0.38 to 0.52), and four pre-registered ablations move the channel estimate by no more than 8 s.

Paper

The Number Arrives Either Way

Runa Adegoke, Kwame Lindqvist, Ingeborg Nwachukwu · July 2026 · doi:10.5555/slop.idvq3t

A paper from the School of Emergent Priorities tests whether a metropolitan water utility's read-completeness rate, which has risen from 93.1 to 99.4 per cent across a transmitting-meter rollout and is used to size the field-reading crew, carries any information about the manual reading a non-reporting connection still requires. All 96 field readers across the utility's fourteen depots walked a standardised twenty-installation route of pre-characterised registers, fourteen routine and six diagnostic. Depot read-completeness does not predict mean route accuracy (r = 0.02, 95% CI −0.52 to 0.54), and a two one-sided tests procedure against a pre-specified smallest effect of interest supports equivalence rather than non-detection. Routine registers are transcribed correctly at 0.94 against 0.42 for the diagnostic items and 0.28 for the two readings implausible against the property's own history, and four pre-registered sensitivity analyses move the primary coefficient by no more than 0.02.

Paper

The Tick Has Two Boxes

Iben Chikere, Anneke Tolan, Solveig Adeyemi · July 2026 · doi:10.5555/slop.bbeucv

A paper from the School of Emergent Priorities follows fourteen volunteer-run community repair cafés in one metropolitan council area for fifteen months, combining 226 hours of bench observation at 71 sessions, 58 semi-structured interviews, a documentary pass over 4,908 paper intake slips, and a twelve-week telephone follow-up with 318 visitors. Reflexive thematic analysis yields a nine-type typology of the states an item is actually in when it leaves the bench; re-coding 1,204 observation-paired slips finds the recorded two-state tick reading repaired on 66.4 per cent of them and spanning five of the nine types, against 43.6 per cent meeting a sustained-use criterion at twelve weeks. Recomputing the council's quarterly diverted-mass tally under four coding schemes moves the median per-session figure by no more than 1.4 kg, the tally being assembled from the item-class and weight fields and never from the outcome box.

Paper

Five Gates, No Dawn Gap

Petra Umbile, Marek Solheim, Verity Marris · July 2026 · doi:10.5555/slop.n23z4k

A paper from the School of Emergent Priorities repeats an earlier dog-park gate-counter study at five further enclosures, one in each council now applying the 1.72 dawn correction factor the original fed, completing 694 paired observer sessions over 24 weeks per site. The pooled dawn-minus-midday agreement difference is -2.1 percentage points against 33 in the original, inside the pre-registered equivalence bounds, with the only attenuated site the only one sharing the original's sensor mounting.

Paper

Lodged, Not Displayed

Marit Osayande, Solveig Adeyemi, Thandiwe Solberg · July 2026 · doi:10.5555/slop.fp74me

A paper from the School of Continuous Improvement releases FORECOURT, an open corpus of 96,318 roadside fuel-price board observations captured from windscreen cameras on a 24-vehicle courier fleet across 412 sites over nineteen weeks, with 8,214 double-coded reference readings and an open reader that recovers the exact posted price string for 94.6 per cent of held-out readings. Pairing each observation with a price-notification register's contemporaneous state, board and register hold the same price at 83.6 per cent of observations (95% CI 82.9-84.3), with a median absolute divergence of 2.4 c/L where they disagree. The board trails a notified decrease by a median 4 h 12 min and a notified increase by 11 min, while the scheme's published notification-timeliness rate, computed entirely from lodgements, did not fall below 99.1 per cent across the same period.

Paper

Nothing Improved Except the Average

Bram Ntuli, Fenna Okoro, Petra Umbile · July 2026 · doi:10.5555/slop.p1t27w

A paper from the School of Continuous Improvement decomposes the first-visit resolution rate a national plumbing and drainage franchise network publishes for each of its territories, using 486,312 completed residential jobs from 214 territories and 4,918 technicians across the fourteen quarters from 2023 Q1 to 2026 Q2. The published rate rises from 71.4 to 76.2 per cent while falling in all nine of the network's own job categories, by between 0.9 and 6.8 percentage points; a symmetric components-of-a-difference decomposition apportions 8.6 percentage points to the composition of dispatched work and minus 3.7 points to category-specific performance. The composition shift dates from an online booking triage introduced in the fifth quarter, which lifted the combined share of tap-and-washer and toilet-and-cistern jobs from 18.5 to 42.4 per cent. Technicians in the top quartile of complex-job share carry a published rate 9.4 points below the network mean, and the dispatch rule allocates complex work by published rate. Twenty-four alternative specifications move the composition component by no more than 0.6 percentage points.

Paper

Where the Volume Went

Torun Ezeigwe, Sten Okwuosa, Joost Nwosu · July 2026 · doi:10.5555/slop.rjrkg0

A paper from the School of Continuous Improvement audits 1,204 grocery lines delisted across 38 stores of one regional chain over two range-review cycles, pairing basket-level point-of-sale records with the planning system's own substitute nominations and estimating per-line volume transfer against a synthetic control drawn from stores where the same line was retained. Observed transfer averages 41.3 per cent of pre-delist units against a fixed planning parameter of 85 per cent, with 112 of 1,204 lines meeting or exceeding it; at household level 46.8 per cent of prior buyers purchase the nominated substitute within 26 weeks, 21.4 per cent more than halve their category units, and 6.1 per cent cease buying the category at that store. Three alternative substitute-matching rules move the mean estimate by less than 1.5 percentage points. The parameter is not observable in the reporting the review produces, each cycle recomputing category performance on the range that survived the previous one.