Reproducibility

Recommendations

The data lifecycle panel of the International Committee for Future Accelerators (ICFA) has devised recommendations for the HEP community on data preservation. The best practices can be filtered by actor and class, so if you are a data analyst and want to know what best practices are in terms of software- and workflow management, you can filter it by exactly that.

Have a look and explore the recommendations under the following link:

If you rather like to read an article instead of exploring a website, the recommendations have also been published via arXiv.

In the course of 2026, the OS Office will also work on more specific guidelines on this topic and we will continuously update this website.

Software for analysis preservation

REANA: Reproducible Research Data Analysis Platform

REANA (REusable ANAlyses) is a cloud-based platform that enables researchers to describe, run, preserve, and reuse their computational analyses in a fully reproducible manner. By combining containerization technology with workflow languages, REANA captures not just the code and data of an analysis, but the complete software environment and execution logic. Researchers can use REANA to document their analysis workflows declaratively, scale computations across thousands of computing cores, and share their work with colleagues through role-based access controls. 

A powerful demonstration of REANA’s capabilities came from the ATLAS Collaboration’s 2023 pMSSM supersymmetry analysis, which evaluated tens of thousands of theoretical models across a 19-dimensional parameter space. By deploying RECAST analysis preservation workflows on REANA, the team was able to scale their detector-level analyses to over 1,000 cores. This approach enabled ATLAS physicists to systematically test thousands of theoretical models that would otherwise have been computationally prohibitive, thereby showcasing the practical impact of reproducible analysis frameworks for advancing high-energy physics research.

If you want to know more about REANA and how to use it, have a look at its documentation or this training course by the HEP Software Foundation:

Snakemake

Snakemake is a workflow management system for creating reproducible and scalable data analyses. Workflows are defined in a human-readable, Python-based language and can be scaled from a local machine to cluster and cloud environments without modifying the workflow itself. Snakemake also handles software dependencies automatically and can generate portable, interactive reports combining results with the parameters, code and software used to produce them. This makes it straightforward to share complete, verifiable analyses with collaborators.

At CERN, Snakemake is already used in experiments including ATLAS, CMS and LHCb, and can be integrated with REANA for large-scale workflow execution. In February 2026, the OSPF held a meeting dedicated to workflow analysis with a vivid discussion about the newest Snakemake developments. To connect with other researchers using workflow tools at CERN, join the analysis workflows community on Mattermost. For more information about Snakemake, have a look at the Documentation.

Resources

Trainings

To get you ready to use the tools you need to make research more reproducible, there is a variety of software trainings available which is collected on a training portal:

The HEP Software Foundation (HSF) offers a wide variety of training material for software basics specific for the field. IRIS-HEP also offers in-person training events, mostly in the US or at CERN.

Other resources

Not specific to HEP, but still a very useful guide on how to make research more reproducible is given by the Turing Way.