OpenAlex Metrics (OAM): citation impact relative to comparable papers

OpenAlex Metrics (OAM) shows whether a set of papers was cited more or less than comparable papers worldwide. It gives field-normalised citation indicators for every paper in scope, including the mean normalised citation score (MNCS) and the share of papers among the 10% and 1% most cited. The indicators can be computed for a university, a lab, a researcher, a country, a funder's portfolio or a list of DOIs. This record holds the prebuilt data, so it can be queried without building anything from the OpenAlex snapshot. OpenAlex already gives most works a citation score, the FWCI. It compares each work with everything OpenAlex indexes, and it counts citations from anything, including preprints, datasets and theses. OAM compares a paper only with peer-reviewed research papers like it, and counts only the citations that come from such papers. In this dataset, a paper qualifies when it passes all seven rules: It is in a core journal or a conference proceedings. It has a reference list. It has at least one institutional affiliation. It is in English. It has an OpenAlex subfield. It is not paratext, such as an editorial, erratum or table of contents. It is not retracted. Anyone can build another dataset with different rules using the code at https://github.com/epfl-p-data/OpenAlexImpactMetrics. The data comes from the public OpenAlex snapshot of 14 October 2026, which has 464,765,485 works. Of these, 62,493,369 (13.45%) pass the rules. Each paper is compared with papers of the same subfield, publication year and venue type (journal or conference). This version counts every citation up to the snapshot, self-citations included. A dataset with a fixed citation window, for example the publication year and the three following years, or without self-citations, can be built with the code. OpenAlex's own FWCI and percentile are included next to OAM's scores, unchanged. The archive oam-2026-10-14.tar (11.9 GiB) holds Parquet tables: the per-paper scores the scope status of every work lookup tables for selecting papers by institution, country, author, ORCID, funder, DOI, source or field A manifest, bundle.json, records the scope definition, the citation counting and a checksum for each file. Install the Python package with pip install openalex-metrics and load the data with oam.download(). The file edges-2026-10-14.tar (9.2 GiB) holds the citation links: 2,171,482,565 links, each one paper citing another, with the citing paper's publication year. Install them with oam.download(edges=True). The scores are unchanged. An older version of the package ignores this file and still installs the main archive. The scores archive is the file built with OAM 0.2 and needs openalex-metrics 0.2 or newer. The citation links were written with OAM 0.5 and need that version or newer. OAM 0.1 reads only version 23493155 of this record, which holds the same snapshot scored with a four-year window and without self-citations. The data is derived from OpenAlex and released under CC0. The code and documentation are at https://github.com/epfl-p-data/OpenAlexImpactMetrics.