80% of the funds analysed in WeeFin’s 2026 Sustainable Finance Barometer now rely on proprietary methodologies for their ESG scoring. This figure reflects a fundamental reality: generic methodologies are no longer sufficient to address the complexity of the issues at stake, evolving client expectations and increasing regulatory requirements.
To enable financial players to design their own models in a rigorous, traceable and collaborative manner, WeeFin has developed the Methodology Lab. Gabriel Levy, Product Research Analyst at WeeFin, explains the principles that guided the development of this module and the key features it offers ESG teams.
What is the right approach to selecting the right data to work with when developing a proprietary methodology?
G.L.: This is a crucial step, because however robust it may be, a methodology will never produce the expected results if it is based on unreliable data. To make the right choice, you need to be able to compare different indicators that could serve the same purpose and decide based on objective criteria. This is precisely what the Methodology Lab enables: a data exploration module that provides all the information needed to make an informed choice: descriptive statistics, coverage of your portfolios, correlations with other scores, and many other angles of analysis.
Better still, the Methodology Lab allows you to change a model’s data source at any time, in just a few clicks. You can thus build two versions of the same model with different inputs, compare them under real-world conditions, and then validate and deploy only the one that best meets your objectives. With the Methodology Lab, you can build with complete peace of mind.
The Methodology Lab allows you to create your own models by assembling operators in a no-code environment. How was the library of available operators put together?
G.L.: These operators were not chosen at random. They are the result of work carried out hand-in-hand with our experts who build ESG methodologies on a daily basis. The library is organised around two families of operators: on the one hand, those covering the most common uses, typically the transformation of indicators through standard normalisation or their aggregation into a score, which should be sufficient to cover 90% of Methodology Lab users’ needs. On the other hand, there are more technical operators designed for advanced use cases (arithmetic formulas, conditional mapping, etc.). In building this library, our guiding principle has always been the same: to offer maximum flexibility, regardless of the user’s profile.
What guarantees are provided regarding the audit trail?
G.L.: The Methodology Lab keeps a record of all the steps used to build a new methodology. Each version includes not only a precise history, with the date and time of creation, but also an owner. This means we know who did what at every stage. This is important because it ensures full traceability, which in turn guarantees the regulatory compliance of the methodology designed.
How are the created models validated?
G.L.: We have included statistical tests in the Methodology Lab known as “expertise checks”: coverage tests across all sectors, detection of under-represented sectors, tests across various dimensions (geography, entity type such as small-cap/big-cap, etc.) and detection of outliers by segment. All these tests can, of course, be customised to suit the user’s needs.
Does the Methodology Lab allow teams to work together on developing new methodologies?
G.L.: Yes, thanks to a permissions system that facilitates collaboration without the risk of unauthorised changes. An owner thus has the rights to access and modify the model, whilst collaborators can view it but cannot modify it directly. It is possible to duplicate a model in order to modify it subsequently. It is also possible to designate one model as the entry point for another. The Methodology Lab allows users to work on separate models and then aggregate them. For example, one can design models dedicated to the environmental and social pillars in parallel, and then aggregate them into a macro-model. This is a very powerful feature.
.jpg)
