Start a project

Research

What if?

Good questions have a habit of leading somewhere interesting. We explore ideas, test assumptions and follow the evidence to see what’s possible.

Started 0.00001 apart
The Lorenz equations, run twice from starting points 0.00001 apart. For a while the runs agree. Then they don’t.

01 · Latest white paper

White paperOctober 2026Seating and inventoryApplied statistics

Structure-aware adaptive sampling for venue seat inventory verification

How to show a digital seat inventory matches the venue without checking every seat: census the structure, sample the seats.

Read the white paper Download PDF31 pages
  • The questionDoes every seat on sale match a real seat in the venue?
  • The answerTest every block and row, sample seats for what those tests cannot see, inspect known problems directly, and escalate by rule.
  • Why it mattersA seat that is not where the inventory says can be sold twice, or not exist at all. Random checks miss exactly those errors.

02 · Publications

Written down, with the working shown.

Everything we have published, newest first. Read each one here in full, or take it away as a PDF.

202604 entries

  1. White paperOctober

    Structure-aware adaptive sampling for venue seat inventory verification

    How to show a digital seat inventory matches the venue without checking every seat: census the structure, sample the seats.

    Seating and inventoryApplied statistics PDF · 31 pages
  2. PaperSeptember

    Forecasting sell-outs in advance ticket sales with a Bayesian mixture model (opens research.loft.ug)

    When will a ticket type sell out? A model of how advance demand arrives, more accurate out of sample than the pickup method.

    Applied statistics
  3. PaperSeptember

    Modelling crowd arrival at stadium turnstiles (opens research.loft.ug)

    When spectators arrive against kick-off, and the number of turnstiles at which no queue forms.

    Crowd dynamicsCapacity and safety
  4. White paperJune

    What we have learned about stadium access control

    What running gates at high-pressure matches taught us, and why small losses in scan quality become large losses in capacity.

    Access controlCapacity and safety PDF · 18 pages

03 · Try the question

How many gates before nobody queues?

The question behind our arrival paper, simplified enough to play with. Change the crowd, the gates and how fast they scan.

Gates for no queue36Arrivals peak at 531 / min.
Longest wait5minFor the unluckiest person in the line.
Everyone in2 min after kick-offWhen the last queue clears.
Gates open29

With 29 gates, a queue builds from 47 min before kick-off and peaks at 2,252 people.

Arrivals a minuteMore than the gates takeQueue
Assumptions

One arrival curve for every crowd, peaking close to kick-off. Every gate admits the same number a minute, and nobody arrives after kick-off. Minute by minute and without randomness, so real queues start a little earlier. The paper models all of this properly: Modelling crowd arrival at stadium turnstiles ↗

Corrections welcome

Think we’re wrong?

Good. Research gets better when ideas are challenged. Tell us what we’ve missed.

Talk to us