Extended definition
Bibliometric analysis is the quantitative mapping of a field, subarea, or topic’s scientific output, from publication metadata: authors, institutions, countries, keywords, citations, journals. Contemporary formalization of the practice is Donthu et al. (2021), who organized the methodology into two large families: performance analysis (productivity, citations, impact factor aggregated by author, institution, journal) and science mapping (coauthorship networks, co-citation, bibliographic coupling, keyword co-occurrence). Primary databases are Scopus (Elsevier, more inclusive) and Web of Science (Clarivate, more conservative). Computational tools include Bibliometrix in R (Aria & Cuccurullo, 2017), VOSviewer (visual maps), CiteSpace (temporal evolution), and the Sci2 Tool. Bibliometric analysis is a methodology in itself — it does not replace systematic review, but offers a panoramic view that traditional reviews do not capture in fields with thousands of publications.
When it applies
Bibliometric analysis is appropriate for mapping the structure of a broad field (hundreds to thousands of papers), identifying central authors and institutions, detecting emerging or declining fronts, and computing comparative indicators (aggregate h-index, average impact factor, international coauthorship pattern). It is often required in doctoral projects as a contextualization chapter, in integrative reviews that complement qualitative analysis with quantitative overview, and in applied research for science management (research funding policy, institutional planning). It is now a standard chapter in master’s and doctoral proposals in many fields.
When it does not apply
Bibliometric analysis does not apply as a substitute for systematic review (which has explicit PRISMA criteria and answers a specific research question), nor as the sole basis for substantive conclusions — bibliometrics maps patterns, it does not synthesize findings. It is not appropriate when the field is small (few dozens of publications), where direct reading is more informative than coauthorship networks. It is not reliable in fields poorly indexed in Scopus/WoS — humanities, regional social sciences, literature in non-English languages. It does not replace expert validation: patterns detected by algorithm need interpretation by those who know the field.
Applications by field
— Management and administration: territory of explosive use in the last decade; Q1 journals publish dozens of bibliometric analyses per year. — Public health: mapping subareas (mental health, chronic diseases, COVID-19), international collaboration networks. — Engineering: analysis of technological evolution, identification of emerging topics in AI, advanced materials, energy. — Education and social sciences: conceptual maps of dominant theories, citation networks among schools of thought.
Common pitfalls
The first pitfall is treating bibliometric analysis as “automatic review” — it does not answer substantive questions about the field, only maps output. The second is relying exclusively on one database (Scopus or Web of Science), losing coverage — combining two databases is more robust practice. The third is interpreting network centrality as quality — an author with many citations may be widely criticized, not widely respected. The fourth is ignoring linguistic and geographic bias of databases — Scopus and Web of Science underrepresent science produced in non-English languages and in systems like SciELO. The fifth is presenting elegant visualizations (maps, networks) without explaining parameters — VOSviewer and CiteSpace produce visually impressive outputs that hide methodological decisions (citation thresholds, clustering algorithms) that significantly affect the result.