Can apply to: Journal articles or other serial publications
Metric definition: A download is an event triggered by a user clicking on the download button, in contrast to simply viewing a web page.
Metric calculation: A count of downloads during a period of time.
Data sources: Publishers, subject repositories, institutional repositories, researcher profile systems, and altmetrics aggregators.
Appropriate use cases: Article downloads can be used as a leading indicator or proxy for others intent to use, rather than actual usage. This may (or may not) be reflected in eventual citations.
Limitations: Article downloads is not an accurate measure of consumption or how many people have read the item, despite the common tendency to equate downloads with usage. Downloaded files may languish unread in personal libraries (resulting in an inflated count of readership) or may be shared with a journal club or other individuals (resulting in an underestimate of readership). Tools that allow for automated crawling and downloading of content may also result in inaccurate counts. In order for web analytics tools to provide an accurate count, they need to be configured to monitor and count these events. Standard analytics tools like Google Analytics may not see downloads when people connect directly to the file through Google or Google Scholar. Platforms commonly used for institutional repositories can count these from the server side, or use plug-ins to provide an accurate count. Finally, the correlation between citations and downloads may vary by discipline and institution.
Inappropriate use cases: Article downloads should not be used as direct measures of usage, research quality, or impact.
Available metric sources: Downloads from publisher sites, subject repositories, and institutional repositories are likely to be fairly accurate, while downloads from personal websites, blogs, and other platforms may be less reliable, depending on the configuration of the site and analytics tools.
Transparency: Varies by source
Timeframe: Typically immediate, but there may be a reporting delay of a few hours up to 30 days, depending on the source.