A strategic view
Brighton & Hove’s secondary school system
In ten years this city has to run a school system for roughly 760 fewer children a year than it has now. This site is an attempt to say — from published data, in the open — what that actually means, and what the options are.
Watch the talk. Can we model our way out of this? at Brighton AI, Plus X Innovation, 4 November 2026. The video starts at the talk, 34 minutes into the evening’s recording. The slides are here.
Brighton & Hove has ten secondary schools, a cohort that peaked in 2019 and is falling, and a sequence of admissions decisions each argued about on its own. What has been missing is an account of how the whole system behaves: where the children are, how families choose, what each policy does to every other school, and which combinations leave the city with a viable set of schools in 2035.
That account is below, together with a simulator you can put your own scenarios through. Everything here is built from published data. No pupil records are used anywhere.
Start here
The main document
The analysis
Where the children live, how demand has moved since 2010, what makes a school attractive, the spatial interaction model of the whole city, and what each option on the table would do. Long, and meant to be argued with.
Interactive
The policy simulator →
Move a catchment, change an admission number, relocate a school, switch what families read when they choose — and watch the whole city re-allocate, this year and out to 2035.
The shape of the problem
760 fewer children entering Year 7 each year by 2033 than at the 2019 peak
10 secondary schools, in a city two miles deep between the sea and the Downs
0.95 R² — how closely the model reproduces the council’s actual September 2026 allocations
195 neighbourhoods, each routed to every school on the real bus timetable
Talks
November 2026
Nov 2026 Talk →
The travelling version: the problem, the tool, a live demo — and the argument it leads to. Why consultations test consent rather than options, and what would have to change for an authority to model honestly, in public, before it has decided.
November 2026 · Brighton AI, Plus X
Brighton AI Talk →
The problem, the tool, and a live demo: how AI collapsed the cost of building a city-scale model, and what else a city might model if it had the right data.
2026 · AI4CI
AI4CI talk →
The longer version, for the UKRI AI for Collective Intelligence hub: how the model was built, what it cost, and what supervision of that kind of work has to look like.
The evidence this rests on
Companion site
How to Pull the Right Lever →
The national study underneath all of this: every secondary school in England, five years of data, and what really predicts a cohort’s GCSE results. Absence is the largest school-level factor by a long way.
Secondary analysis
Exclusions and Attendance →
If attendance is the lever, is exclusion the problem? No — suspension is about 3% of secondary absence nationally and under 2% here. Illness-coded absence is 95% of what separates this city from England.
On the data. Everything on this site is built from published sources: DfE performance tables and the statistics API, the council’s own allocation factsheets going back to 2010, the 2021 Census, OpenStreetMap and the real bus timetable, and one Freedom of Information request. The code is open on GitHub. No pupil-level data is used anywhere.