Transparency
At Made with You, transparency starts with us. Here's how we build our ESG analysis platform in line with our values.
Our analysis methodology
How it works
Our platform automatically analyzes companies' ESG communication from their public website. Here's the process:
Content collection : we explore the site via sitemaps and extract relevant content (web pages, publicly accessible PDF reports).
AI analysis : the content is analyzed by Mistral (mistral-medium-latest) according to 5 transparency criteria: authentic messaging, business coherence, operational transparency, methodological clarity, and long-term commitment. The model is trained to detect greenwashing and evaluate substance over marketing.
Size adaptation : the analysis automatically adapts to the company's size. The LLM has accumulated broad knowledge of different organization types that enables this nuance. We don't penalize SMEs for lack of formalism if their approach is coherent and authentic. Large groups are evaluated more strictly on data and formal methodologies.
What we evaluate and what we don't
We evaluate the clarity and authenticity of what the company communicates on its site. Is the discourse substantive or marketing? Is it coherent with the business model? Are there concrete proofs or vague promises?
We don't evaluate the company's ground reality. We don't conduct operational audits, we don't go on-site, we don't verify figures in factories. Our analysis focuses on communication transparency, not actual ESG performance.
The role of AI
AI is not human. It's a knowledge concentration tool that we treat as such: a synthetic analyst capable of rapidly processing volumes of information no human could read in the same time, according to the criteria we give it.
Our approach's limits
Bot protection: some sites block automated access. In this case, the analysis relies solely on the Mistral model's preexisting knowledge (which may be incomplete or outdated). We clearly indicate this in the analysis sources.
LLM biases: like any language model, Mistral has limits. It may lack specific sector context, misinterpret technical discourse, or reflect biases present in its training data.
Public content only: if a company does good work but doesn't communicate it on its site, our analysis won't capture it. Transparency ≠ performance.
Our intent
Our goal isn't to rank companies from best to worst. It's to help quickly spot signals: who communicates substantively? Who does hollow marketing? Where are the inconsistencies?
A bad score doesn't necessarily mean the company does badly. It can also mean they communicate poorly or not at all. A good score doesn't guarantee perfection, but indicates clear and coherent communication.
We want to encourage transparency, not punish those making efforts without the means to produce exhaustive reports.
Our technology approach
AI choice
We use Mistral as our analysis engine. This choice isn't trivial: Mistral has been audited by Carbone4, a reference in environmental evaluation. This external validation gives us confidence in our tool's carbon impact.
We only send the most relevant excerpts to the AI. No massive transmission of useless information. Once the analysis is generated, it's available to everyone without mobilizing new resources: compute once, share widely.
Hosting
All our infrastructure (data, databases, emails, application) is hosted on Scaleway in Europe. We chose this partner for their concrete commitment to measuring and reporting CO2 and water consumption of their data centers.
Not everything is perfectly evaluated at their end yet, but they progress transparently. That's exactly the approach we want to support.
Data protection
No collection, no resale. We don't collect personal data and don't sell any information.
Our model is simple: provide a useful service and get paid for the value created, not through advertising.
This approach avoids conflicts of interest. Your privacy stays private.
What's next
We keep improving our approach. Every technical choice is guided by its real impact.
We can't ask others for more transparency without being transparent ourselves.