Tied rank
Several institutions may share a position. Top-100 tables can therefore contain more than 100 entries.
Methods before conclusions
This site reduces a large multi-provider ranking archive into a small analytical layer. It does not erase provider definitions, measurement windows, ties, missingness, or reuse restrictions.
Analytical consensus
Mean within-table percentile across the latest available broad overall editions; it is an analytical index, not a new ranking.
percentile = 100 × (1 − (rank − 1) / (ranked entries − 1))↓FOR EACH INSTITUTIONsignal = mean(provider percentiles)Source inventory
Each provider publishes on its own schedule, scope, and license. The table lists what each one contributes and how its data may be reused.
| Provider | Lens | Coverage | Latest edition | Reuse status |
|---|---|---|---|---|
| U.S. News | Publisher ranking | 2026 · 1 edition · 51 subject viewsCurrent global and subject rankings | 2026Archived Jul 19, 2026 | Provider-controlled |
| Times Higher Education | Publisher ranking | 2011—2026 · 16 editions · 11 subject viewsOverall and subject editions | 2026Archived Jul 19, 2026 | Provider-controlled |
| QS | Publisher ranking | 2018—2027 · 10 editions · 60 subject viewsOverall and broad subject editions | 2027Archived Jul 19, 2026 | Provider-controlled |
| Leiden Open Edition | Bibliometric ranking | 2023—2025 · 3 editions · 5 subject viewsPublication scale and field-normalized impact | 2025Archived Jul 19, 2026 | CC0-1.0 |
| OpenAlex | Derived research output | 2016—2025 · 10 editionsAnnual works output reconstructed from CC0 data | 2025Archived Jul 19, 2026 | CC0-1.0 |
| CWUR | Publisher ranking | 2012—2026 · 15 editionsOverall editions | 2026Archived Jul 19, 2026 | Provider-controlled |
| NTU Ranking | Bibliometric ranking | 2007—2026 · 20 editions · 33 subject viewsOverall, field, and subject editions | 2026Archived Sep 28, 2026 | Provider-controlled |
| ShanghaiRanking | Publisher ranking | 2003—2026 · 24 editions · 57 subject viewsARWU overall and GRAS subjects | 2026Archived Sep 28, 2026 | Provider-controlled |
| SCImago SIR | Bibliometric ranking | 2009—2026 · 18 editions · 19 subject viewsOverall and 19 research areas | 2026Archived Jul 19, 2026 | Provider-controlled |
| Nature Index | Research-output ranking | 2016—2026 · 11 editions · 17 subject viewsAll-sector and academic annual tables | 2026Archived Jul 19, 2026 | CC-BY-NC-SA-4.0 (numerical table data) |
| Webometrics | Web visibility ranking | 2025—2026 · 2 editionsOverall editions; July 2026 lists the top 15 per country | 2026Archived Sep 28, 2026 | CC-BY-4.0 |
Provider-controlled snapshots are included under separately confirmed permission. That permission does not replace publisher terms or automatically grant downstream redistribution rights. OpenAlex and Leiden are CC0; Webometrics is CC BY 4.0; Nature Index numerical data is CC BY-NC-SA 4.0. Edition years follow each publisher’s own label, so a “2027” QS table and a “2026” ARWU table can be released in the same summer. Publishers occasionally trim a release: Webometrics’ July 2026 edition lists only each country’s top 15, so web-visibility comparisons use the most recent complete edition.
Rank semantics
The site preserves publisher display values and uses a band’s lower bound only when a numeric position is required for plotting or percentile conversion.
Several institutions may share a position. Top-100 tables can therefore contain more than 100 entries.
The display stays a band. A trajectory uses 601 as a documented plotting convention, not as a precise rank.
THE reporters are counted in table coverage but excluded from ranked universes and consensus calculations.
Absence is not zero. It may mean outside the published table, not eligible, not collected, or not safely matched.
Research metrics
p allocates multi-institution papers fractionally. The site’s scale rank is derived from this value within the selected field.
MNCS normalizes citation impact by field and year. Top-10% share measures the fraction of work among highly cited papers.
Historical values are reconstructed from current institution snapshots. Counts are institution-work associations, not deduplicated global papers.
Share covers contributions to a selected journal set. Edition 2026 represents 2025 output; journal and method changes limit causal comparisons.
Entity resolution
Institution names change, translate, abbreviate, and collide. Shared ROR or OpenAlex identifiers are preferred where available; publisher consensus otherwise uses exact normalized names, country blocks, and a small reviewed alias map.
Preserve every raw source name and country.
Preserve source-specific identifiers and use shared ROR or OpenAlex IDs where datasets expose them.
Normalize Unicode, case, punctuation, and whitespace only to generate candidates.
Require a harmonized country, apply reviewed aliases for known variants, and keep ambiguous matches separate.
Responsible guidance
Look for recurring strength across methods and years.
Verify the exact department, program, faculty, and facilities.
Compare admissions likelihood, cost, aid, visas, and prerequisites.
Inspect class model, support, location, culture, and student outcomes.
Open the provider methodology and measurement period before interpreting movement.
Weight your constraints explicitly; do not substitute the consensus decimal.