01 HE Open Data Ranking 2027 Public Methodology v1 0
HE
Public Methodology v1.0
HE Open Data Ranking 2027 — Public Methodological Framework
This document is part of the official HE Open Data Ranking 2027 Methodology v1.0 documentation set.
Document control
Contents
1. Purpose and positioning
2. Core principles
3. Institution scope and eligibility
4. Research window and evidence model
5. Scoring architecture
6. The 24 indicators
7. Normalization and reliability
8. Missing data and zero values
9. Ranking outputs and interpretation
10. Transparency and reproducibility
11. Whole-source unavailability rule
12. Limitations
13. Methodology governance and versioning
1. Purpose and positioning
HE Open Data Ranking 2027 is a research-oriented university ranking designed to benchmark institutional research performance and academic impact using externally observable, structured academic evidence. It does not attempt to measure every mission of a university, and it should not be interpreted as a comprehensive measure of teaching quality, student experience, institutional finances, or reputation.
The 2027 methodology is designed for large-scale application, including universities that may not appear in traditional elite-focused rankings. Institutions are compared within a common scoring universe using the same definitions and rules.
2. Core principles
External evidence rather than self-reported performance data.
No reputation survey and no external ranking position as a scoring input.
Persistent institutional identity and explicit entity boundaries.
Inclusive coverage: no minimum research-output threshold.
A 10,000-point multidimensional architecture with transparent indicator weights.
Reliability adjustments for percentage indicators with small denominators.
Full-cohort normalization for most indicators, with a special zero-aware rule for sparse registry counts.
Missing, observed zero, and structural zero states are distinguished for audit even though all contribute zero under the scoring policy.
Scores and ranks are generated by rule; they are not manually adjusted to resemble other rankings.
3. Institution scope and eligibility
The ranking is built around institution-level identifiers supplied and verified by HE. Higher education institutions remain in the scoring universe even when their research output is zero. Institutions are removed only through an explicit HE exclusion decision. Review flags, missing country information, or a zero-output condition do not automatically remove an institution.
Entity boundaries follow the identifier assigned to the institution. Mergers, successor institutions, and major reorganizations may require an entity-continuity review because a newly created identifier may not carry the full historical research record of predecessor organizations.
4. Research window and evidence model
For the 2027 edition, the principal research-performance window is 2021–2025. Certain contextual scale measures, such as lifetime citations and current researcher-profile proxies, use a longer or current-profile reference period as explicitly stated in the indicator definition. The university website is not crawled and website traffic or visibility is not scored. Official institutional domains are used only as identity-resolution keys where appropriate.
5. Scoring architecture
The final institutional score is the sum of 24 KPI point contributions. Each KPI is normalized to a 0–100 score and then multiplied by its fixed maximum point value. The maximum conceptual score is 10,000 points.
KPI points_i = (Normalized KPI score_i / 100) × KPI maximum points_i
Total score = Σ KPI points_i, i = 1,…,24
6. The 24 indicators
Indicator definitions and mathematical formulas are published separately in the KPI Definitions & Formulas document. The weighting structure is fixed for the 2027 edition.
7. Normalization and reliability
The methodology combines scale, percentage, diversity, and sparse-count indicators. Scale indicators use logarithmic transformation and winsorization before percentile scoring. Percentage indicators that can be unstable at small sample sizes use an empirical-Bayes reliability adjustment. Diversity measures are shrunk toward the cohort median according to the size of their underlying evidence base. Sparse institutional-registry counts use a separate rule in which zero remains zero and positive observations are ranked only against other positive observations.
8. Missing data and zero values
There is no neutral 50-point imputation and no automatic exclusion for low coverage. Scoring coverage is reported so users can distinguish a low score from a data-coverage issue.
9. Ranking outputs and interpretation
Global rank based on the total score.
Country rank and continental rank using the same total score.
Overall score out of 10,000.
Global percentile within the edition’s ranked universe.
Five pillar scores and 24 KPI scores.
Data-coverage and audit information.
Rank movement across editions should be interpreted together with score, percentile, and cohort size. The ranking universe is expected to expand substantially over time; an institution can move down in exact rank while improving its relative percentile if many new institutions are added.
10. Transparency and reproducibility
HE publishes the methodology version, weights, indicator definitions, scoring rules, and institutional identity information. External rankings may be used to validate broad convergent behavior, but no external ranking position is read by the production scoring engine. HE does not manually alter an institutional score to align it with another ranking system.
11. Whole-source unavailability rule
If an entire external registry used by a scoring component is unavailable for the full ranking cohort at the official data-freeze date, that component remains defined but contributes zero uniformly to all institutions. Its points are not redistributed to other indicators. This prevents a temporary source-access problem from silently changing the weight structure. If the source becomes reproducibly available before the data freeze, activation requires a full-cohort regression check; otherwise activation is deferred to a later edition and documented.
12. Limitations
Open scholarly databases are not perfectly complete or geographically uniform.
Institutional mergers and successor identifiers can create discontinuities in historical attribution.
A composite rank necessarily reflects normative choices about indicator weights.
Some indicators capture research behavior that differs structurally across disciplines and national systems.
Small score differences should not be interpreted as proof of a substantive institutional difference.
The ranking measures the construct defined by HE; disagreement with another ranking is not, by itself, evidence of error.
13. Methodology governance and versioning
Methodology v1.0 is frozen for the inaugural 2027 edition. Corrections to code, data acquisition, memory handling, or file export that do not change indicator definitions or weights are implementation updates, not methodology changes. Any future substantive methodology change must be documented before the next edition, sensitivity-tested, versioned, and recorded in the Methodology Change Log.
References and source notes
HE Open Data Ranking 2027 Scoring Engine v0.10.6 (internal technical source; frozen basis for Methodology v1.0).
HE Open Data Ranking 2027 Validation Report v1.0.
HE Open Data Ranking 2027 KPI Definitions & Formulas v1.0.
Version | Methodology v1.0
Status | FROZEN — 2027 INAUGURAL EDITION
Effective date | 13 August 2026
Classification | Public reference
Purpose | Public-facing explanation of what the ranking measures, how scores are constructed, and how results should be interpreted.
Methodological basis | Frozen from scoring engine v0.10.6 after calibration and validation on 600-, 150-, and 1,000-institution cohorts.
Methodology freeze | 13 August 2026
Scoring architecture | 10,000 points · 5 pillars · 24 KPIs
Research window | 2021–2025 for the 2027 edition
Governance rule | External ranking positions are validation evidence only and are never scoring inputs.
Positioning — A global open-data ranking of university research performance and academic impact. The ranking is designed for broad institutional coverage and does not require a questionnaire, institutional data submission, or reputation survey.
Pillar | Weight | Maximum points
Research Performance & Open Knowledge | 50% | 5,000
Internationalization & Global Networks | 18% | 1,800
Research Capacity & Academic Breadth | 15% | 1,500
Societal Relevance & Sustainable Research | 7% | 700
Cross-Sector Research Engagement | 10% | 1,000
# | Indicator | Wt. | Pts | Public definition
1 | Research Output Scale | 11% | 1100 | Research Outputs
2 | Total Citation Impact Scale | 7% | 700 | Lifetime citations
3 | Research Output Momentum | 3% | 300 | Annual research outputs
4 | Top 10% Research Excellence | 9% | 900 | Top-10% citation-normalized share
5 | Top 1% Exceptional Research Excellence | 12% | 1200 | Top-1% citation-normalized share
6 | Open Research Excellence | 1% | 100 | OA share among Top-10% works
7 | Open Access Performance | 1% | 100 | Open-access share
8 | Open Knowledge Infrastructure | 6% | 600 | Eight-component open-knowledge composite
9 | International Research Collaboration | 4% | 400 | Internationally coauthored share
10 | Multilateral Research Collaboration | 4% | 400 | 3+ country works / international works
11 | Geographic Collaboration Strength | 1% | 100 | Country breadth + partner-country evenness
12 | Institutional Network Strength | 1% | 100 | Institution breadth + partner-institution evenness
13 | International Research Leadership | 2% | 200 | Corresponding-institution share among international works
14 | High-Impact International Collaboration | 6% | 600 | Top-10% share among international works
15 | Researcher Base | 5% | 500 | Authors with institution as last-known affiliation
16 | Research Productivity Efficiency | 4% | 400 | Research outputs per last-known author
17 | Research Identity & Discoverability | 3% | 300 | ORCID coverage among last-known authors
18 | Research Field Diversity | 3% | 300 | Normalized field-distribution entropy
19 | SDG Research Share | 1% | 100 | Share of outputs aligned with ≥1 SDG
20 | SDG Research Diversity | 1% | 100 | Normalized Shannon diversity of SDG groupings
21 | High-Impact SDG Research | 3% | 300 | SDG excellence advantage vs institutional baseline
22 | Open SDG Research | 2% | 200 | SDG openness advantage vs institutional baseline
23 | Cross-Sector Research Engagement Composite | 6% | 600 | Company + government + healthcare + nonprofit shares
24 | High-Impact Industry Collaboration | 4% | 400 | Top-10% share among industry-collaborative works
Why reliability adjustment matters — A university with one highly cited paper should not automatically receive the same excellence signal as a university with thousands of papers. Reliability adjustments preserve performance differences while reducing instability caused by small denominators.
Data state | Meaning | Scoring treatment
OBSERVED | A numerical value was successfully measured. | Value is scored normally.
OBSERVED_ZERO | The source was successfully checked and the measured value is zero. | 0 contribution where applicable.
STRUCTURAL_ZERO | A conditional indicator is undefined because its measured denominator is zero. | 0 contribution; not counted as missing coverage.
MISSING | A required numerical value was not available from the source. | 0 contribution and reported as missing in audit outputs.
Freeze decision — No additional weight tuning is authorized for the 2027 inaugural edition merely to improve agreement with other rankings or to move individual institutions toward expected positions.
Public Methodology v1.0
HE Open Data Ranking 2027 — Public Methodological Framework
This document is part of the official HE Open Data Ranking 2027 Methodology v1.0 documentation set.
Document control
Contents
1. Purpose and positioning
2. Core principles
3. Institution scope and eligibility
4. Research window and evidence model
5. Scoring architecture
6. The 24 indicators
7. Normalization and reliability
8. Missing data and zero values
9. Ranking outputs and interpretation
10. Transparency and reproducibility
11. Whole-source unavailability rule
12. Limitations
13. Methodology governance and versioning
1. Purpose and positioning
HE Open Data Ranking 2027 is a research-oriented university ranking designed to benchmark institutional research performance and academic impact using externally observable, structured academic evidence. It does not attempt to measure every mission of a university, and it should not be interpreted as a comprehensive measure of teaching quality, student experience, institutional finances, or reputation.
The 2027 methodology is designed for large-scale application, including universities that may not appear in traditional elite-focused rankings. Institutions are compared within a common scoring universe using the same definitions and rules.
2. Core principles
External evidence rather than self-reported performance data.
No reputation survey and no external ranking position as a scoring input.
Persistent institutional identity and explicit entity boundaries.
Inclusive coverage: no minimum research-output threshold.
A 10,000-point multidimensional architecture with transparent indicator weights.
Reliability adjustments for percentage indicators with small denominators.
Full-cohort normalization for most indicators, with a special zero-aware rule for sparse registry counts.
Missing, observed zero, and structural zero states are distinguished for audit even though all contribute zero under the scoring policy.
Scores and ranks are generated by rule; they are not manually adjusted to resemble other rankings.
3. Institution scope and eligibility
The ranking is built around institution-level identifiers supplied and verified by HE. Higher education institutions remain in the scoring universe even when their research output is zero. Institutions are removed only through an explicit HE exclusion decision. Review flags, missing country information, or a zero-output condition do not automatically remove an institution.
Entity boundaries follow the identifier assigned to the institution. Mergers, successor institutions, and major reorganizations may require an entity-continuity review because a newly created identifier may not carry the full historical research record of predecessor organizations.
4. Research window and evidence model
For the 2027 edition, the principal research-performance window is 2021–2025. Certain contextual scale measures, such as lifetime citations and current researcher-profile proxies, use a longer or current-profile reference period as explicitly stated in the indicator definition. The university website is not crawled and website traffic or visibility is not scored. Official institutional domains are used only as identity-resolution keys where appropriate.
5. Scoring architecture
The final institutional score is the sum of 24 KPI point contributions. Each KPI is normalized to a 0–100 score and then multiplied by its fixed maximum point value. The maximum conceptual score is 10,000 points.
KPI points_i = (Normalized KPI score_i / 100) × KPI maximum points_i
Total score = Σ KPI points_i, i = 1,…,24
6. The 24 indicators
Indicator definitions and mathematical formulas are published separately in the KPI Definitions & Formulas document. The weighting structure is fixed for the 2027 edition.
7. Normalization and reliability
The methodology combines scale, percentage, diversity, and sparse-count indicators. Scale indicators use logarithmic transformation and winsorization before percentile scoring. Percentage indicators that can be unstable at small sample sizes use an empirical-Bayes reliability adjustment. Diversity measures are shrunk toward the cohort median according to the size of their underlying evidence base. Sparse institutional-registry counts use a separate rule in which zero remains zero and positive observations are ranked only against other positive observations.
8. Missing data and zero values
There is no neutral 50-point imputation and no automatic exclusion for low coverage. Scoring coverage is reported so users can distinguish a low score from a data-coverage issue.
9. Ranking outputs and interpretation
Global rank based on the total score.
Country rank and continental rank using the same total score.
Overall score out of 10,000.
Global percentile within the edition’s ranked universe.
Five pillar scores and 24 KPI scores.
Data-coverage and audit information.
Rank movement across editions should be interpreted together with score, percentile, and cohort size. The ranking universe is expected to expand substantially over time; an institution can move down in exact rank while improving its relative percentile if many new institutions are added.
10. Transparency and reproducibility
HE publishes the methodology version, weights, indicator definitions, scoring rules, and institutional identity information. External rankings may be used to validate broad convergent behavior, but no external ranking position is read by the production scoring engine. HE does not manually alter an institutional score to align it with another ranking system.
11. Whole-source unavailability rule
If an entire external registry used by a scoring component is unavailable for the full ranking cohort at the official data-freeze date, that component remains defined but contributes zero uniformly to all institutions. Its points are not redistributed to other indicators. This prevents a temporary source-access problem from silently changing the weight structure. If the source becomes reproducibly available before the data freeze, activation requires a full-cohort regression check; otherwise activation is deferred to a later edition and documented.
12. Limitations
Open scholarly databases are not perfectly complete or geographically uniform.
Institutional mergers and successor identifiers can create discontinuities in historical attribution.
A composite rank necessarily reflects normative choices about indicator weights.
Some indicators capture research behavior that differs structurally across disciplines and national systems.
Small score differences should not be interpreted as proof of a substantive institutional difference.
The ranking measures the construct defined by HE; disagreement with another ranking is not, by itself, evidence of error.
13. Methodology governance and versioning
Methodology v1.0 is frozen for the inaugural 2027 edition. Corrections to code, data acquisition, memory handling, or file export that do not change indicator definitions or weights are implementation updates, not methodology changes. Any future substantive methodology change must be documented before the next edition, sensitivity-tested, versioned, and recorded in the Methodology Change Log.
References and source notes
HE Open Data Ranking 2027 Scoring Engine v0.10.6 (internal technical source; frozen basis for Methodology v1.0).
HE Open Data Ranking 2027 Validation Report v1.0.
HE Open Data Ranking 2027 KPI Definitions & Formulas v1.0.
Version | Methodology v1.0
Status | FROZEN — 2027 INAUGURAL EDITION
Effective date | 13 August 2026
Classification | Public reference
Purpose | Public-facing explanation of what the ranking measures, how scores are constructed, and how results should be interpreted.
Methodological basis | Frozen from scoring engine v0.10.6 after calibration and validation on 600-, 150-, and 1,000-institution cohorts.
Methodology freeze | 13 August 2026
Scoring architecture | 10,000 points · 5 pillars · 24 KPIs
Research window | 2021–2025 for the 2027 edition
Governance rule | External ranking positions are validation evidence only and are never scoring inputs.
Positioning — A global open-data ranking of university research performance and academic impact. The ranking is designed for broad institutional coverage and does not require a questionnaire, institutional data submission, or reputation survey.
Pillar | Weight | Maximum points
Research Performance & Open Knowledge | 50% | 5,000
Internationalization & Global Networks | 18% | 1,800
Research Capacity & Academic Breadth | 15% | 1,500
Societal Relevance & Sustainable Research | 7% | 700
Cross-Sector Research Engagement | 10% | 1,000
# | Indicator | Wt. | Pts | Public definition
1 | Research Output Scale | 11% | 1100 | Research Outputs
2 | Total Citation Impact Scale | 7% | 700 | Lifetime citations
3 | Research Output Momentum | 3% | 300 | Annual research outputs
4 | Top 10% Research Excellence | 9% | 900 | Top-10% citation-normalized share
5 | Top 1% Exceptional Research Excellence | 12% | 1200 | Top-1% citation-normalized share
6 | Open Research Excellence | 1% | 100 | OA share among Top-10% works
7 | Open Access Performance | 1% | 100 | Open-access share
8 | Open Knowledge Infrastructure | 6% | 600 | Eight-component open-knowledge composite
9 | International Research Collaboration | 4% | 400 | Internationally coauthored share
10 | Multilateral Research Collaboration | 4% | 400 | 3+ country works / international works
11 | Geographic Collaboration Strength | 1% | 100 | Country breadth + partner-country evenness
12 | Institutional Network Strength | 1% | 100 | Institution breadth + partner-institution evenness
13 | International Research Leadership | 2% | 200 | Corresponding-institution share among international works
14 | High-Impact International Collaboration | 6% | 600 | Top-10% share among international works
15 | Researcher Base | 5% | 500 | Authors with institution as last-known affiliation
16 | Research Productivity Efficiency | 4% | 400 | Research outputs per last-known author
17 | Research Identity & Discoverability | 3% | 300 | ORCID coverage among last-known authors
18 | Research Field Diversity | 3% | 300 | Normalized field-distribution entropy
19 | SDG Research Share | 1% | 100 | Share of outputs aligned with ≥1 SDG
20 | SDG Research Diversity | 1% | 100 | Normalized Shannon diversity of SDG groupings
21 | High-Impact SDG Research | 3% | 300 | SDG excellence advantage vs institutional baseline
22 | Open SDG Research | 2% | 200 | SDG openness advantage vs institutional baseline
23 | Cross-Sector Research Engagement Composite | 6% | 600 | Company + government + healthcare + nonprofit shares
24 | High-Impact Industry Collaboration | 4% | 400 | Top-10% share among industry-collaborative works
Why reliability adjustment matters — A university with one highly cited paper should not automatically receive the same excellence signal as a university with thousands of papers. Reliability adjustments preserve performance differences while reducing instability caused by small denominators.
Data state | Meaning | Scoring treatment
OBSERVED | A numerical value was successfully measured. | Value is scored normally.
OBSERVED_ZERO | The source was successfully checked and the measured value is zero. | 0 contribution where applicable.
STRUCTURAL_ZERO | A conditional indicator is undefined because its measured denominator is zero. | 0 contribution; not counted as missing coverage.
MISSING | A required numerical value was not available from the source. | 0 contribution and reported as missing in audit outputs.
Freeze decision — No additional weight tuning is authorized for the 2027 inaugural edition merely to improve agreement with other rankings or to move individual institutions toward expected positions.