How we rank schools
We monitor 100+ data points for every school across the public web, survey schools directly and hear from parents who opt in. AI helps us understand the context behind the numbers; a published, weighted formula does the scoring. No model decides a rank.
- Schools scored
- 695
- Ranked
- 335
- Pillars
- 8
- Data points per school
- 100+
The index at a glance
- Type
- Deterministic weighted composite. Same inputs, same score, every time.
- Output
- Score 0 to 100 per school, a confidence figure, and a rank.
- Inputs
- 100+ data points per school from across the public web, direct surveys of schools, and ratings and reviews from parents who opt in.
- Where AI is used
- Reading the context behind the numbers, so every figure is understood in the round.
- Where AI is not used
- Pillar scores, weights, confidence, rank order. All fixed arithmetic, published below.
- Rank rule
- Ordered by score times confidence. At least 3 measured pillars to be ranked.
- Versioning
- Every formula change bumps the version. Each row records the version it was scored under.
How we get to know every school
Several ways of rating and reviewing a school, brought together in one place.
- 01
Monitor
We track 100+ data points for every school across the public web, and keep them up to date.
- 02
Ask
We survey schools directly, and parents who opt in share their experience through ratings and reviews.
- 03
Understand
AI reads the context behind the numbers, not just the numbers, so every figure is read in the round.
- 04
Match
We surface rich insights on every school and help match schools to your child.
Eight pillars, weights in the open
Each pillar scores 0 to 100 from the evidence. A pillar without evidence drops out of the score and the rest renormalise. The gap shows up in confidence instead, beside every score.
Published ranking position50%evidence 7% (50)
Position in an established national ranking, for the schools that hold one.
Signals- Published national ranking position
Formula1st scores 100, falling in a straight line to 80 at 50th. Absent for every other school, whose remaining weights renormalise to 100. Never counted toward confidence.
AI roleNo model step specific to this pillar.
Academic outcomesAI input13%evidence 42% (294)
Top-grade results at A level and GCSE, and offers from the most selective universities.
Signals- Share of A level entries at the top grades
- Share of GCSE entries at grades 9 to 7
- Offer rate from the two most selective universities
- Offer rate from research-intensive universities
FormulaUp to four signals averaged. With only one signal the pillar is blended halfway to a neutral 70, so a single strong figure cannot carry it.
AI roleLanguage models read results pages and summaries and extract the figures into a typed schema. Figures are reconciled with public exam statistics.
Destinations & outcomesAI input10%evidence 42% (295)
Where leavers go: university offers, school-published destinations and prep-school leaver offers.
Signals- University offers and places
- School-published leaver destinations
- Senior-school offers for prep leavers
- Alumni record
FormulaThe two strongest signals are pooled, so a strong result is not averaged away by a weak one. A single signal blends toward a neutral 60.
AI roleLanguage models read leaver-destination pages and extract offers by destination.
Future-readinessAI input7%evidence 14% (98)
Computing, AI, robotics, design and enterprise in the curriculum. 10 schools are classified so far; 88 hold a provisional tier 1 until they are.
Signals- Computing qualifications offered
- AI or data units and policies
- Robotics and making facilities
- Enterprise programmes
FormulaA 0 to 4 rubric tier maps to 40, 55, 70, 85 or 95. Hidden when not assessed.
AI roleAn AI classifier reads each school's curriculum text against the rubric and must cite its evidence. Marketing language alone scores tier 0.
Pastoral & wellbeingAI input7%evidence 69% (477)
Statutory inspection judgements on wellbeing, plus verified parent reviews once enough exist.
Signals- Statutory inspection judgements
- Verified parent reviews on pastoral care
FormulaExcellent 95, Good 80, Standards met 82, Sound 70, Below standard 50. Parent reviews, when present, are averaged in.
AI roleLanguage models read inspection reports and extract the judgement.
Access & breadthAI input6%evidence 61% (424)
Bursary depth, fully funded places and breadth of co-curricular life.
Signals- Bursaries and their maximum award
- Fully funded places
- Co-curricular breadth
FormulaAny bursary 70, awards worth half fees or more 80, fully funded places 90. Broad co-curricular provision lifts the floor to 75.
AI roleLanguage models extract bursary terms from fees and admissions pages.
Independent review coverage4%evidence 52% (362)
How many established independent reviewers have written the school up.
Signals- Independent editorial reviews
Formula70 plus 10 for each reviewer. Weighted lightly: coverage tracks age and size as much as quality.
AI roleNo model step specific to this pillar.
Value-add3%evidence 31% (213)
Outcomes set against how selective the school is at entry.
Signals- Academic outcomes
- Selectivity at entry
FormulaThe academic score, adjusted up for non-selective schools. A proxy until entry and exit data exists, so its weight stays low.
AI roleNo model step specific to this pillar.
Published weights. They apply in full when every pillar has evidence. A pillar without evidence always drops out the same way.
The benchmark pillar carries 50% for the 50 schools that hold a place in an established published ranking, so the two lists don’t describe different worlds. It is left out of confidence in both directions: holding a position never raises confidence, and lacking one never lowers it.
How much we know, shown beside every score
Confidence is the weighted share of a school’s applicable pillars backed by real evidence. Prep schools are not marked down for senior-only pillars such as A level results or university destinations.
Rank uses score times confidence, so a school with one brilliant figure cannot outrank a school we can see in full. Below 3 measured pillars a school is listed with the evidence it has and no rank.
Schools can raise their confidence by claiming their listing and supplying verified data. It earns a badge, never a different formula.
What the index will not do
- No paid placement
Schools cannot pay to change a score or a rank.
- No hidden formula
Every weight and rule is on this page.
- No single number
Every score breaks down into its pillars.
- No replacing a visit
Use it to shortlist, then go to the open day.
Model versions
- v2026.6.1current
- Junior and infant sections that share a government registration with their senior school are scored on their own evidence only. Exam results, university destinations, alumni and the inspection judgement belong to the whole registration, so they no longer count toward a junior section's rank.
- Sections left with fewer than three measured pillars are listed without a rank.
- v2026.6.0
- Published ranking position becomes its own pillar at 50%, for the schools that hold one. Every other school is scored on its own evidence with weights renormalised.
- Independent review coverage split out at 4%. A review says a school was written about, not where it placed.
- Benchmark pillars sit outside confidence in both directions and do not count toward the three-pillar floor for a rank.
- Rank agreement with the published benchmark rose from 0.35 to 0.74 (Spearman).
- v2026.5.0
- Pupil flourishing and Parent sentiment removed. Neither had fired for any school, so their 17 points went to pillars with real data.
- Future-readiness hidden where it was not measured, instead of showing a default.
- A rank now needs at least three measured pillars.
- v2026.4.0
- Confidence floor and sparse-pillar penalty tightened.
- Parent reviews count only from five verified reviews.
- v2026.3.0
- Three sub-indices added: Academic, Whole-child, Future-ready.
- Thin-signal penalty on academic and destinations.
- Destinations pools the top two signals instead of taking one maximum.
- v2026.2.0
- Academic pillar rebuilt from raw results and offer rates, no longer from an external rank.
- Future-readiness becomes a 0 to 4 rubric set by an AI classifier, replacing a keyword scan.
- v2026.1.0
- Initial release. Nine pillars, with academic approximated from a published rank.
Questions
Does AI decide where a school ranks?
No. AI reads school publications and turns them into structured facts, and it classifies curriculum text on a published rubric. The score is a fixed weighted formula over those facts, and every weight is on this page.
Why does a published ranking count for half the score?
Families compare lists, and a ranking that ignores the established ones describes a different world. So where a school holds a published position, it counts 50%. A school without one is scored on its own evidence and is never marked down for the omission.
What does the confidence figure mean?
The weighted share of a school's applicable pillars backed by real evidence. Prep schools are not marked down for senior-only pillars. Rank uses score times confidence, so thin evidence cannot buy a high place.
Can a school pay to rank higher?
No. Schools can claim their listing and supply verified data. Verified data earns a badge and can raise confidence, but the formula is the same for every school.
Where does the data come from?
From several directions at once. We monitor 100+ data points for every school across the public web, we survey schools directly, and parents who opt in share ratings and reviews. Each school's page shows the facts we hold for it.