CSRankings: Methodology, Metrics and Guide

CSRankings (csrankings.org) is a transparent, metrics-based ranking system designed to evaluate computer science institutions and faculty based on research publications at top-tier, highly selective conferences. Created as an open-source alternative to subjective reputation surveys such as U.S. News & World Report, the platform tracks faculty productivity across defined research areas using bibliographic data from computer science literature indexed in DBLP.

What it is

CSRankings is an open-source web application and institutional ranking directory for computer science departments globally. Unlike commercial ranking tables that rely on qualitative dean reputation surveys or composite institutional wealth metrics, CSRankings operates entirely on publication counts in elite, peer-reviewed venues.

The platform addresses a unique structural feature of computer science scholarly communication: in computer science, top-tier conference proceedings function as the primary archival venue for cutting-edge research, carrying prestige equal to or exceeding many archival journals. By aggregating verified faculty affiliations with conference indexing records, CSRankings provides an open baseline for comparing department research activity.

Who runs it and what it covers

CSRankings was created and is actively maintained by Emery Berger, Professor of Computer Science at the University of Massachusetts Amherst. The project evolved from earlier indexing experiments and datasets developed by Swarat Chaudhuri (UT Austin) and Alex Papoutsaki et al. (Brown University). The site source code and institutional dataset are publicly hosted on GitHub under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND 4.0).

The coverage of CSRankings is governed by strict inclusion parameters:

Data source: Publication records are pulled directly from public data releases of the DBLP Computer Science Bibliography (dblp.org).
Faculty eligibility: Inclusion is restricted to full-time, tenure-track faculty holding at least a 75% appointment for the academic year who can solely advise Ph.D. students in Computer Science. This includes qualifying faculty from related departments (e.g., Electrical and Computer Engineering or Information Sciences) provided they hold formal dissertation-advising rights for CS doctoral candidates. Adjuncts, lecturers, research scientists, postdocs, and industry fellows are excluded.
Venue selection: Conforming to research-centric ACM Special Interest Groups (SIGs), venues are selected based on community surveys and consensus among leading researchers. For an area to be tracked, at least 50 R1 research institutions must have published in its designated conferences over the preceding 10-year window.
Publication rules: Only peer-reviewed research track papers of at least 6 pages in length are counted. Keynotes, invited talks, tutorials, panel summaries, posters, and short abstracts are systematically filtered out.

How to search or use it

The CSRankings web interface allows researchers, prospective graduate students, and evaluators to generate customized rankings. To navigate the interface effectively, follow these operational steps based on the official documentation:

Select geographic region: Use the regional checkboxes at the top of the interface to filter institutions by country or continent (e.g., USA, Canada, Europe, Asia, or Australasia).
Define the evaluation window: Adjust the start and end year sliders to specify the publication date range (by default set to a rolling 10-year period).
Select subfield categories: Filter by high-level areas—AI (Artificial Intelligence), Systems, Theory, or Interdisciplinary Areas—or click individual checkboxes to select specific subdisciplines (e.g., Operating Systems, Computer Architecture, Cryptography, or Human-Computer Interaction).
Choose the ranking metric: Toggle between “All” (raw institutional score) or specific discipline filters to recalculate rankings based on the selected subfields.
Inspect institutional faculty: Click the triangular toggle icon next to any institution’s name to expand the row and view the specific roster of included tenure-track faculty.
Examine individual publication records: Click the publication count next to any faculty member’s name to view their attributed papers directly linked to their corresponding DBLP entry.
Verify faculty profiles: Select a faculty member’s name to navigate to their external academic homepage, Google Scholar profile, or DBLP author record.

CSRankings methodology: adjusted and average counts

To avoid distortions caused by varied author counts and differing field sizes, the csrankings methodology utilizes two core mathematical metrics: adjusted counts and geometric average counts.

Adjusted Counts (Fractional Authorship Credit)

A central challenge in bibliometrics is co-authorship inflation. If an institution received a full point for every co-authored paper, adding honorary or marginal co-authors would rapidly game departmental scores. CSRankings eliminates this incentive by enforcing fractional credit:

Each eligible publication contributes exactly 1.0 total unit of credit across all listed co-authors.
If a paper has N authors, each author receives 1/N credit, regardless of their status (faculty, student, or external collaborator).
An institution is credited with the sum of fractional shares held by its active tenure-track faculty. For detailed context on how authorship position is treated across publishing traditions, see our guide on CS conference vs journal authorship credit.
Credit remains portable: when a faculty member moves to a new university, their historical publication record transfers with them, reflecting the institution’s current research capacity rather than historical legacy.

Average Counts (Geometric Mean Area Normalization)

Subdisciplines within computing differ dramatically in annual publication output; for instance, machine learning venues accept thousands of papers annually, whereas theoretical computer science conferences accept far fewer. To prevent high-volume fields from swamping smaller subfields in composite rankings, CSRankings calculates the geometric mean of adjusted counts across selected areas:

For n selected areas, the score is calculated as:

Average Count = [∏_i=1..n (Adjusted Count_i + 1)]^(1/n)

Adding 1 prior to multiplication handles zero-publication areas and ensures fair cross-field weighting across disparate computational disciplines, similar to distinctions seen when evaluating computer science vs data science programs.

Strengths and limits

CSRankings provides distinct advantages for academic evaluation, but also introduces significant limitations that researchers and administrators must weigh:

Strengths

Transparency: All underlying code, faculty rosters, venue filters, and parsing scripts are publicly visible and reproducible on GitHub.
Resistance to citation cartels: By counting primary publications in highly selective conferences rather than citation counts, CSRankings avoids citation rings, self-citation inflation, and Google Scholar profile manipulation.
Current faculty reflection: Because faculty data reflects current institutional rosters, departments cannot coast on past historical prestige if productive researchers have departed.
Customizable evaluation: Evaluators can tailor rankings to specific subdisciplines rather than accepting an arbitrary institutional aggregate.

Limits and Critiques (The CACM Debate)

The platform has been the subject of prominent academic critique, notably by Professor Mark Guzdial in the Communications of the ACM (CACM) blog (“Why I Don’t Recommend CSRankings.org: Know the Values You are Ranking On”):

Goodhart’s Law: When CSRankings conference lists become the target metric for hiring and tenure, departments may prioritize incremental conference papers over longitudinal, high-risk research.
Total exclusion of journals: High-impact archival journals—such as the Journal of Machine Learning Research (JMLR), IEEE Transactions, and ACM Transactions—are omitted because DBLP conference lists are the primary source. This penalizes research groups where journals represent the gold standard.
Penalty on interdisciplinary collaboration: Papers published in cross-disciplinary venues (e.g., medical informatics, computational biology, or educational technology) do not count toward CSRankings if the venues sit outside core ACM/IEEE CS conferences.
Department size bias: Although fractional credit prevents individual paper padding, total department scores still scale with faculty size. Large departments frequently outrank elite smaller departments in aggregate views.
Inapplicability to undergraduate education: Prospective undergraduate students often mistakenly consult CSRankings, despite the fact that doctoral publication volume bears little correlation with undergraduate teaching quality, curriculum design, or mentorship, areas central to computer science vs software engineering pedagogy.

Alternatives

Institutions and prospective scholars typically balance CSRankings against several alternative ranking models and bibliometric databases:

Ranking / Platform	Primary Data Source	Methodology	Key Strengths	Key Weaknesses
CSRankings	DBLP publication dumps	Count of selective conference papers with fractional co-author credit and geometric normalization	Open-source, fully transparent, metrics-based, immune to survey reputation bias	Excludes journals and interdisciplinary venues; favors large faculties in total score
U.S. News CS Rankings	Reputation surveys	Subjective peer assessment survey scores from department chairs and graduate directors	Captures broad institutional reputation and historical prestige	Closed methodology, slow to reflect faculty turnover, vulnerable to perceptual lag
Research.com	OpenAlex / Crossref / Google Scholar	Faculty D-index (Discipline H-index) and cumulative citation thresholds	Measures long-term citation impact across both conferences and journals	Vulnerable to citation gaming, name disambiguation errors, and historical bias
Google Scholar Metrics	Google Scholar web crawl	H5-index and H5-median for computer science publication venues	Broad venue coverage across international conferences, workshops, and journals	Uncurated preprints, vulnerability to spam indexing, lack of institutional normalization

FAQ

Is CSRankings reliable for choosing an undergraduate university?

No. CSRankings measures research publication volume by tenure-track faculty in selective graduate conferences. It does not measure teaching quality, undergraduate curriculum rigor, student-to-faculty ratios, career services, or internship placement.

Why are peer-reviewed journals excluded from CSRankings?

CSRankings relies on DBLP conference records because conferences historically represent the primary publication track in computer science. The platform creator chose to omit general science journals (such as Nature or Science) and domain journals due to indexing differences and historical focus on conference parity.

How does CSRankings handle faculty changes and moves?

CSRankings maintains institutional faculty rosters via an open-source GitHub repository updated through a self-service form. Submissions are processed on a quarterly schedule. When a faculty member switches institutions, their entire historical publication record transfers to their new institution.

Can adding more co-authors boost an institution’s CSRankings score?

No. CSRankings uses adjusted counts, allocating exactly 1/N credit for each paper with N authors. Adding co-authors dilutes individual and institutional credit, preventing author-padding inflation.

Who is eligible to be listed as faculty on CSRankings?

Only full-time tenure-track faculty (at least a 75% appointment for the full academic year) who have the sole authority to advise Ph.D. students in Computer Science at that campus are eligible for inclusion.

Sources

Berger, Emery. “CSRankings: Computer Science Rankings FAQ.” csrankings.org. Read November 7, 2026. https://csrankings.org/faq.html
Berger, Emery. “CSrankings: A metrics-based ranking of top computer science institutions.” GitHub Repository (emeryberger/CSrankings). Read November 7, 2026. https://github.com/emeryberger/CSrankings
Guzdial, Mark. “Why I Don’t Recommend CSRankings.org: Know the Values You are Ranking On.” Communications of the ACM (CACM) Blog, Association for Computing Machinery. Read November 7, 2026. https://cacm.acm.org/
DBLP. “DBLP Computer Science Bibliography: Open Data and Standards.” Schloss Dagstuhl – Leibniz Center for Informatics. Read November 7, 2026. https://dblp.org/