OpenEdition’s statistics service transitions to COUNTER 5.1

Last February, OpenEdition’s usage statistics service updated to the latest COUNTER standard: Release 5.1. The service provides every institution subscribing to the OpenEdition Freemium programme with customized statistics reports. It was rolled out thanks to the COMMONS project.

The service’s URL (available to subscribing institutions only) and user credentials remain unchanged: 

• Access to the service: https://counter5.openedition.org 
• Access to the SUSHI API: https://counter5.openedition.org/api/doc

Background

OpenEdition’s COUNTER 5 usage statistics service was set up in 2021 and certified in 2021 and 2023.
The reports presented in COUNTER specifications consist of monthly usage summaries per book, journal and type of visit. The SUSHI API lets client applications send requests to the service to retrieve COUNTER-compliant reports, so that they can incorporate them into a statistics aggregator such as ezMESURE, the French national repository of usage statistics for higher education institutions. Since its certification in 2021, OpenEdition has been added to the list of COUNTER compliant platforms: https://registry.countermetrics.org/.

Changes to OpenEdition’s COUNTER statistics service

After completely overhauling our COUNTER application in 2025, we further updated the statistics service in February 2026. It now offers: 

• COUNTER 5.1 compliant usage data
• Improved performance
• Correction of calculation errors

From COUNTER 5 to COUNTER 5.1

The COUNTER 5 standard will continue to apply to statistics from 2024. It is referred to as COUNTER 5.0 in our application to avoid ambiguity.
Usage data from January 2025 onwards complies with the new COUNTER 5.1 standard.
Changes introduced by COUNTER 5.1:

• The metrics have changed compared to the previous version of our application, especially in terms of downloading full versions of books in PDF/ePUB format.

Read our documentation for more information.

• The properties of Access Types have changed:
  • ‘OA_Gold’ is now ‘Open’
  • Journals with an embargo period: articles no longer under embargo (‘Open’); articles with a moving wall delay (‘Controlled’)
  • Freemium Open Access content: harmonized as ‘Controlled’

Read our documentation for more information.

Documentation

We have updated our online documentation: https://counter5-openedition.readthedocs.io. Feel free to check out the Friendly Guide To Changes In Release 5.1 too, and a full description of the COUNTER 5 standard in the COUNTER Code of Practice.

Statistical methods – a few insights

Usage statistics are generated by analysing OpenEdition’s server access logs. Processing these logs is a complex and sensitive operation because of the plethora of data and the limited amount of information we have on the clients (humans or bots) visiting our platforms.

We set up the processing we think is needed to generate accurate usage statistics. However, the array of data we handle and the experience we have acquired in this domain makes us wary. Despite our best efforts, we’re likely to have missed a few things and there may be statistical inaccuracies in some cases.

It’s also worth noting that the processing set up for the latest version of the application differs from that for the previous version, which may lead to different statistical outcomes.

Figures must therefore be interpreted with caution, especially if there are large discrepancies or values that don’t seem plausible. Below, we take a look at some of the technical aspects to explain what we mean.

Access logs – a source of data for statistics

Statistics are generated by analysing OpenEdition’s server access logs. Each visit to a web page generates several requests on our servers (the page itself, the images, Ajax requests, etc.) and, therefore, several lines in our access logs. It’s important to sort out which lines are meaningful (visits to the web pages of an article) and which are not (e.g. images).

OpenEdition’s platforms attract a variety of people, but also bots. These bots can be search engine crawlers (Google, Bing, etc.), AI data crawlers gathering training data (Open AI, Claude AI, etc.), or web scrapers harvesting data for malicious purposes (DOS attacks) or other unknown ends.

Some bots identify themselves as such (e.g. Google’s bot, via the user-agent string ‘Googlebot/2.1’), whereas others use web browser user-agents and often appear in logs as human-generated traffic.

Traffic adds up to several million requests every day. For example, on 21/01/2026, OpenEdition Journals’ access logs contained 9,517,332 lines.

Although we’re setting up tools to manage our access logs and curb the effects bots are having on our platforms and statistical reports (by denying entry to some bots as soon as they try to access our platforms, or identifying and excluding bots from our logs), the sheer number of visits and the limited amount of information available in our access logs makes filtering complicated, and we can’t be entirely sure of the results.

On the bright side, the statistics for subscribing institutions are based solely on visits from IP addresses registered with OpenEdition. For some institutions, this can be all the IP addresses from across their campus, and for others, just those from their library’s proxy server, for when OpenEdition is accessed via the library’s online services.

As a result, COUNTER 5 statistics generated for subscribing institutions are generally spared the impact of bot visits, as it’s highly unlikely that a bot will access OpenEdition via an IP registered by an institution.

Improving the performance of the updated statistics service

COUNTER reports are now generated much faster thanks to the improvements we have made to our application: 

• We only log visits from IP addresses associated with subscribing institutions to generate COUNTER 5 statistics. This significantly reduces the quantity of data to process, yet it also currently prevents us from providing COUNTER statistics for traffic that isn’t from subscribing institutions
• We have enhanced and optimized the schema for our database (Elasticsearch)
• We have set up a cluster for our Elasticsearch database: the database is spread across several servers, making it much more efficient

In the previous application, it sometimes took up to half an hour or so to generate a TR (Title Report) with several Attributes and covering a period of several months. Now, it takes less than thirty seconds.

By restructuring the application, we were able to streamline certain processing operations and thereby correct statistical calculation errors made with the previous version (some PR (Platform Report) values were incorrectly cached as zero values).

Improving processing log accuracy

Our access logs used to contain duplicates (identical log lines), or double-clicks, as defined in the COUNTER standard (same content viewed within a thirty-second period). The previous version of our application didn’t filter duplicates properly. 

We have remedied this in the new version to make sure that duplicates are properly filtered. In some cases, this has a significant impact on the statistics generated – the number of views is sometimes lower in the reports generated for the same period by the new application.

The impact varies depending on the case: 

• The figures are 0–20% lower for the Total_Item_Requests and Total_Item_Investigations metrics
• The Unique_Item_Investigations, Unique_Item_Requests, Unique_Title_Investigations and Unique_Title_Requests metrics are not affected

Next steps

An audit will be carried out in 2026 to obtain certification, as required by the COUNTER standard.
The code for the two applications developed for this service will be published under an open license during the course of the year.

---

---