Methodology — model v3.0
How every table on this site is built
Each ranking is a single weighted composite scored out of 100. The same six indicators are applied to every table on the site — the world index, thirteen national tables and the specialist tables — so a law school, a management masters and a whole institution are judged on comparable terms. Everything below is published before the data for an edition is closed.
- Indicators
- 6
- Model version
- 3.0
- Current edition
- 2026
- Editions published
- 5
The composite in one paragraph
Six indicators are collected for every entry, converted to a z-score within their own table, rescaled onto 0–100, multiplied by a fixed weight and summed. The result is reported to one decimal place. Ties break on graduate outcomes, then on selectivity. Nothing else enters the score.
| Indicator | Weight | What it answers |
|---|---|---|
| Graduate outcomes | 30% | Does the qualification lead somewhere? |
| Salary uplift | 20% | What does the qualification change financially? |
| Faculty & research strength | 15% | Who actually teaches, and what do they produce? |
| Selectivity | 15% | How hard is it to get in? |
| Employer reputation | 10% | Do the people who hire actually rate it? |
| Cohort & network reach | 10% | What do you get after you graduate? |
| Total | 100% | One score, one decimal place |
Principles we hold the model to
Weights are fixed before the data is seen
The weighting is published at the start of each cycle and cannot be changed once collection has begun. If a weighting is wrong, it is corrected in the next edition and disclosed, never retro-fitted to produce a preferred ordering.
One composite, every table
The same six indicators and the same weights produce the world table, every national table and every subject table. We do not reweight a table because its result looks counter-intuitive.
Nothing is for sale
No institution pays to be assessed, to appear, to move, or to be excluded. There is no premium listing, no paid profile and no advertising sold against a specific table position.
Method before result
Where a result surprises us, we publish it and explain the mechanism in the entry note. The alternative — quietly adjusting until the table matches expectation — is what makes most rankings unreadable.
Estimates are labelled as estimates
Scores here are our own editorial estimates, not official figures released by the institutions named. Every page says so, and adjacent scores within about a point should be read as a rough equivalence.
Corrections are permanent and visible
A corrected entry keeps a note of what changed and when. We do not overwrite history to make a past edition look better than it was.
Read the tables against the method
Every entry on the site carries its six indicator scores, so any position can be read back against the model on this page.
See the tables