Research metrics appear everywhere in academic life: on journal homepages, in grant applications, on researcher profiles and in institutional reports. Most of them are calculated from the same raw material, citations and publications, yet they answer quite different questions. A number designed to describe a journal says little about a single article, and a number designed to describe a researcher's citation record says nothing about whether their work was discussed in the news.
Much of the confusion around metrics comes from treating them as interchangeable measures of "impact" or "quality". They are not. Each has a defined level of analysis, a data source, a calculation method and known blind spots. Knowing those four things is most of what metrics literacy requires.
This guide groups the most common indicators by what they describe: journals, articles, authors and online attention. For a closer look at the Journal Impact Factor and how journal rankings shift, see our analysis Is the Impact Factor losing its power.
Journal-level metrics
Journal-level metrics summarise the citations received by everything a journal published over a period. They are useful for comparing journals within the same field, and they are the wrong tool for judging individual papers or people.
Journal Impact Factor (JIF)
Published by Clarivate in the Journal Citation Reports, the JIF is a ratio of citations received in a given year to the citable items a journal published in the two preceding years. It draws on the Web of Science Core Collection. The JIF is a metric: having one is not the same as being indexed in a particular collection.
CiteScore
CiteScore is Elsevier's journal metric, calculated from Scopus data. It uses a four-year window: citations received over four years to documents published in those same four years, divided by the number of those documents. Both the numerator and the denominator are limited to five peer-reviewed document types: articles, reviews, conference papers, book chapters and data papers. Elsevier states that CiteScore values can be viewed free of charge on Scopus.
SCImago Journal Rank (SJR)
SJR, published by the SCImago Research Group using Scopus data, does not count every citation equally. It applies an iterative, prestige-weighted method in which a citation from a highly ranked journal contributes more than one from a lower-ranked journal, over a three-year window. Journal self-citation is capped so that it cannot dominate the score.
Source Normalised Impact per Paper (SNIP)
SNIP, calculated by the Centre for Science and Technology Studies (CWTS) at Leiden University from Scopus data, adjusts for differences in citation practices between fields. Citations from publications with long reference lists count for less, so a journal in a field where authors cite heavily is not automatically advantaged over one in a field where they cite sparingly. Of the four journal metrics here, SNIP is the one explicitly designed for comparison across disciplines.
What to remember: JIF, CiteScore, SJR and SNIP come from two different databases with different coverage, use different windows and count different document types. A journal can rank well on one and less well on another without anything being wrong. Always note which metric, which year and which subject category is being cited.
Article-level metrics
Article-level metrics describe an individual piece of work. They are closer to the question most evaluators actually care about, but they need context too.
Citation counts
The simplest measure is the number of times an article has been cited. It depends heavily on the database used, on the age of the article and on the field. A five-year-old paper in a fast-moving biomedical area and a five-year-old paper in mathematics will accumulate citations at very different rates, and neither count says on its own whether the work is sound.
Field-normalised indicators
Field-normalised indicators compare an article's citations with what would be expected for similar work. The best-known example is Field-Weighted Citation Impact (FWCI), available in Elsevier's SciVal and based on Scopus data. According to Elsevier, FWCI compares the citations an output receives with the average for publications of the same year, type and discipline, counting citations over the publication year and the following three years. An FWCI of 1.00 means the work was cited exactly as expected; 2.00 means twice the expected rate.
Elsevier itself warns against relying on FWCI for small sets of publications. Because it is an average, one or two highly cited papers can inflate the figure for a group of ten, and the value can fluctuate over time. Pairing it with a median or with simple citation counts gives a more stable picture.
Author-level metrics
The h-index
Proposed by the physicist Jorge Hirsch in 2005, the h-index is the largest number h such that a researcher has h papers each cited at least h times. A researcher with an h-index of 15 has 15 papers with at least 15 citations each.
Its appeal is that it combines productivity and citation impact in one number. Its limits follow directly from its design:
- It favours long careers. The index can only rise over time, so early-career researchers are disadvantaged regardless of the quality of their work.
- It is field-dependent. Fields with higher citation rates produce higher h-indices, so comparisons across disciplines are misleading.
- It depends on the database. An h-index calculated in Google Scholar, Scopus and Web of Science will often differ, because each covers different publications.
- It ignores contribution. A paper counts the same whether the researcher led it or was one of hundreds of co-authors.
- It is insensitive to the most important work. Once a paper passes the threshold, further citations do not change the index.
An h-index is best read as a rough description of a citation record, alongside the record itself, not as a ranking of people.
Alternative metrics
Alternative metrics, often called altmetrics, track attention to research outside formal citation. Altmetric, one of the main providers, records mentions in news outlets, policy documents, patents, Wikipedia, blogs and social media platforms. Its Altmetric Attention Score is a weighted approximation of the volume and type of that attention, with, for example, a news story counting for more than a social media post.
Altmetric states plainly that the score is not a measure of the quality of the research or the researcher. A paper can attract a great deal of attention because it is controversial or flawed. The value of attention data lies in the underlying mentions: seeing which policy document cited a study, or which audience discussed it, can reveal reach that citations miss, especially for applied research.
Using metrics responsibly
A few principles cover most situations:
- Match the metric to the level. Journal metrics describe journals; article metrics describe articles; author metrics describe citation records.
- Name the source and year. Scopus, Web of Science and Google Scholar produce different numbers for the same work.
- Compare like with like. Use field-normalised indicators, or compare only within a field.
- Read beyond the number. Look at the actual citations, mentions and work behind any figure.
- Use several indicators together and treat them as context for expert judgement, not a substitute for it.
For our view on why journal metrics in particular should be kept out of individual evaluation, see The Impact Factor describes journals. Stop using it to judge researchers.
Conclusion
Research metrics are tools, and like any tools they work well only for the job they were made for. The Journal Impact Factor, CiteScore, SJR and SNIP describe journals in different ways. Citation counts and FWCI describe articles. The h-index summarises a citation record. Altmetric-type data shows where research travelled beyond academia. Used together, with their data sources and limits in view, they can add useful evidence to a judgement. Used alone, or at the wrong level, any one of them can mislead.
