UniRank
QS THE ARWU Methodology

How Are University Rankings Calculated: A Reader's Guide to the Indicators Behind the Numbers

A university rank is the end product of a publisher’s measurement model, not a direct reading of “quality.” unirank.io runs no student survey, publishes no agent or broker reviews, and does not manufacture a composite score or an overall reliability grade. The methodologies below show what can enter a published position; they do not declare a winning institution or ranking publisher.

The indicator coordinate map

The useful map separates the principal indicator families:

  • Research output and citations: used by QS, THE and ARWU, but with different measures and weights.
  • Reputation surveys: used by QS and THE.
  • Teaching and student-facing measures: used by QS and THE.
  • International outlook: used by QS and THE; no international measure is named in the ARWU 2025 table.
  • Prizes and per-capita performance: central to ARWU 2025, but not named as standalone indicators in the supplied QS or THE tables.

This map is not exhaustive. The QS World University Rankings 2026 also assigns 5% to Sustainability, covering environmental, social and governance factors and research tied to the UN’s 17 Sustainable Development Goals. THE World University Rankings 2026 assigns 4% to Industry, split equally between Industry income and Patents at 2% each. Neither category maps cleanly onto the principal families.

Research output and citations

For the QS World University Rankings 2026, the Research and Discovery lens carries 50%: Academic Reputation contributes 30% and Citations per Faculty contributes 20%. Although the latter divides citation volume by faculty count, the supplied methodology text does not disclose the underlying index or the complete indicator transformation formula.

For THE World University Rankings 2026, the Research environment pillar carries 29%: Research reputation 18%, Research income 5.5% and Research productivity 5.5%. Research quality carries 30%: Citation impact 15%, Research strength 5%, Research excellence 5% and Research influence 5%. THE standardises indicators using a cumulative probability function. Most metrics use a version of Z-scoring, while Teaching reputation, Research reputation, Research excellence, Research influence and Patents use an exponential component. The supplied evidence does not break down the definitions of Research strength, Research excellence or Research influence.

For ARWU 2025, Highly Cited Researchers carries 20%, papers published in Nature and Science carries 20%, papers indexed in SCIE and SSCI carries 20%, and Per Capita Performance carries 10%. ARWU assigns the highest-scoring institution 100 for each indicator; other institutions receive a percentage of that top score.

Citation counts reward indexed work that attracts citations. Relative measures can penalise institutions with fewer recorded citations, while subject, publication type and citation-window length affect comparability. THE normalises citation volume between subject areas, but normalisation does not turn a citation count into an observation of teaching.

Reputation surveys

The QS World University Rankings 2026 assigns 30% to Academic Reputation and 15% to Employer Reputation. Academic Reputation draws on the QS Academic Survey, which evaluates nominations for approximately 7000 institutions each year. That is a nomination universe, not a respondent count. Employer Reputation draws on the QS Employer Survey; the supplied evidence provides no employer response total.

For THE World University Rankings 2026, Teaching reputation carries 15% and Research reputation carries 18%. Its most recent Academic Reputation Survey ran from November 2024 to January 2025. THE combined the resulting 2025 data with the 2024 survey, producing more than 108,000 responses, and weighted responses to reflect the global distribution of scholars across disciplines and countries.

A reputation survey records respondents’ judgments. It cannot directly observe every teaching interaction, research process or graduate outcome. The QS evidence gives no response-count or sampling file, so its coverage cannot be assessed from the supplied material.

Teaching and student-facing measures

In the QS World University Rankings 2026, Employability and Outcomes carries 20%, divided between Employer Reputation at 15% and Employment Outcomes at 5%. Learning Experience carries 10%, entirely through the Faculty Student Ratio at 10%. The description of Employment Outcomes refers to employability and graduates’ impact, but the supplied evidence gives no graduate-tracking protocol with which to test that description.

In THE World University Rankings 2026, Teaching carries 29.5%: Teaching reputation 15%, Staff-to-student ratio 4.5%, Doctorate-to-bachelor’s ratio 2%, Doctorates-awarded-to-academic-staff ratio 5.5% and Institutional income 2.5%.

An aggregate staff-to-student ratio cannot reveal contact hours, class-size distribution, discipline-level teaching loads or an individual student’s experience. Similar ratios arise from different combinations of staffing and enrolment, so the ratios are not direct measures of those conditions.

International outlook

The QS World University Rankings 2026 assigns 15% to Global Engagement: International Faculty Ratio 5%, International Research Network 5%, International Student Diversity 0% and International Student Ratio 5%. Its International Research Network indicator counts countries represented and sustained partnerships, defined as partnerships producing three or more joint papers within a five-year period.

For THE World University Rankings 2026, International outlook carries 7.5%. The supplied methodology refers to country-normalised international measures but does not provide their names or individual weights. THE also says its study-abroad metric currently carries 0%. It changed normalisation to account for national population because staff and students from large countries can face different cross-border opportunities from those in smaller countries.

International ratios consequently reflect geography, country size and institutional composition as well as cross-border activity. The supplied evidence does not document every denominator QS uses, so these are not direct measures of internationalisation quality.

Prizes and per-capita performance

The ARWU 2025 assigns 10% to alumni winning Nobel Prizes or Fields Medals and 20% to staff winning those awards. More recent recipients receive greater weight than earlier recipients. ARWU uses no reputation survey and draws its prize information from Nobel Prize and Fields Medal records.

Prize indicators can identify award recipients associated with an institution, but they do not observe current teaching, student support or day-to-day performance. Their long measurement windows also differ from the recent citation windows used elsewhere.

ARWU 2025 calculates Per Capita Performance by dividing weighted indicator scores by full-time-equivalent academic staff. Where a country’s staff figures are unavailable, ARWU substitutes the average staff number for universities in the world top 1000. A per-capita result therefore depends on both numerator and denominator: it can change when staffing changes even if the underlying award or publication total does not. Raw counts, meanwhile, tend to increase with institutional scale.

Where the data come from

Bibliometric sources observe indexed publications and recorded citations, but not their full social or educational value. THE World University Rankings 2026 uses Elsevier’s Scopus data, covering 2020–2024 publications and 2020–2025 citations, with subject-volume normalisation. The QS World University Rankings by Subject methodology, updated 25 March 2026, also sources citation data from Scopus and applies subject-specific publication thresholds.

ARWU 2025 declares Web of Science SCIE and SSCI, Nature, Science, Clarivate’s Highly Cited Researchers list and official prize records as principal sources. By contrast, the supplied evidence does not identify the index behind overall QS Citations per Faculty. It therefore cannot establish whether every remaining non-bibliometric value is self-reported, verified or modelled.

Coverage and publication limits

ARWU 2025 says more than 2500 universities are ranked but publishes only the best 1000. THE World University Rankings 2026 may exclude universities that do not teach undergraduates, have fewer than 1000 relevant 2020–2024 publications with a minimum of 100 a year, or concentrate 80 per cent or more of research output in one of 11 subject areas.

The QS World University Rankings by Subject methodology, updated 25 March 2026, covers 55 subjects across five broad areas and says its list is not extensive. It names Academic Reputation, Employer Reputation, Research Citations per Paper, H-index and International Research Network, but supplies no numerical weights.

Rankings also require edition-level checks. QS World University Rankings 2026 warns that corrections may be made after publication. THE World University Rankings 2026 records a correction to country-normalised scores for Northern Cyprus universities.

A practical reading procedure

Before interpreting any position:

  • Identify the exact table and edition. Do not mix an overall QS table with the QS subject table, THE World University Rankings 2026 or ARWU 2025.
  • Decompose the result into its published families and indicators. A total position is not evidence about teaching, research, internationalisation or awards separately.
  • Audit the source and period. Separate survey perceptions, bibliographic records, prize records, ratios and institutional inputs. Do not call an undocumented input self-reported merely because an institution may have supplied it.
  • Check coverage and denominators. Ask whether the table is exhaustive, whether per-capita staff figures were observed or substituted, and whether disciplinary or country normalisation applies.
  • Read the rank label literally. An exact rank is an ordinal position; a published band is an interval and must not be converted into a midpoint. The supplied evidence contains no complete institutional tables or systematic rank-band dataset, so no institutional positions are estimated here.
  • Check corrections and then match the indicator to the question. Use research families to inspect research signals, reputation indicators to inspect surveyed perceptions, teaching measures to inspect resource or outcome proxies, international measures to inspect cross-border composition, and prize measures to inspect recognised elite achievement. No single published rank supplies an overall reliability score, and the publishers’ systems weight these families differently.