Close
Close
This guide © 2024 by UCL - Library Skills is licensed under CC BY-NC-SA 4.0 
To measure the overall citation performance of a group of papers, we recommend using indicators which are normalised - that is to say, they have been adjusted to compare like for like, to ensure that citation numbers are compared between broadly similar papers.
We recommend using either (or both):
These indicators show subtly different things, and so it can be useful to use them both, especially for benchmarking.
The CNCI shows the average level of citations in a group of papers compared to a global average. It can, however, be skewed by a small number of very highly cited papers. Conversely, the share of papers in the top 1%/10% show how many papers are "very highly cited" or "highly cited", but does not give any information on the citations for those below the threshold.
Data for both of them is available from InCites.
The first type of normalised citation indicator we recommend is the category-normalised citation impact, or CNCI. This can be obtained through InCites.
The CNCI is calculated by taking all papers with the same subject area, document type (article, conference paper, book chapter, etc), and publication year. The average number of citations for these papers is calculated, and then each paper is compared to the average. For example, if the main set of papers has an average of ten citations, and an individual paper has fifteen, it has a CNCI of 1.5.
For a group of papers, the CNCIs are calculated for each paper and then averaged together - so a group with equal numbers of 1.0 and 2.0 papers would have an overall CNCI of 1.5, with the global average at 1.0.
Scopus provides a similar field-weighted citation impact (FWCI) - this is calculated in a similar way although the definitions used are a little different. This can be seen on individual article pages in Scopus under the "metrics" tab.
The second type of normalised citation indicator we recommend is the share of papers in the top 1% or top 10%. This can also be obtained through InCites.
As above, this is calculated by taking all papers with the same subject area, document type (article, conference paper, book chapter, etc), and publication year. Rather than average the citation numbers, however, they are ranked and each paper assigned a percentile rank - so that the top cited paper is 100, one with an average 50, etc. The percentile rank for each individual paper can be examined manually and analysed, but InCites also provides an aggregated count for the share of papers in top 1% and number of papers in top 10%.
A group of papers with 3% of papers in the top 1% thus has about three times as many very highly-cited papers as the world average.
Scopus provides a similar percentile value for individual papers- this can be seen on individual article pages in Scopus under the "metrics" tab.
While normalised indicators are a lot better than simple citation counts or journal-level impact factors, they do have some issues that it is worth being alert to.
For example, recently published papers can have quite volatile figures - if the group has less than one citation per paper, then getting a single citation moves the paper straight from a value of zero to already being above-average. We would recommend avoiding them for the most recent calendar year and treating them with caution for fields with very low citation rates.
Similarly, InCites usually reports a wide range of document types, some of which rarely recieve citations and so can have very high normalised citation rates if they do. We recommend filtering to articles and reviews, and potentially also conference proceedings (not conference abstracts) depending on the fields you are assessing.
