The problem we kept meeting
Every EU project we’ve worked on had the same deliverable hiding somewhere in the work plan: a catalogue of good practices. Every consortium built one. Almost none survived the grant that paid for it. The evidence — real programmes, real evaluations, real lessons — ends up scattered across final reports and websites that stop being renewed the year the funding ends.
The waste is double. The evidence is lost to the next project, and the field keeps re-answering questions that documented practice has already answered.
What we built
Evidoria is an evidence observatory: a single, permanent platform that sources, structures, evaluates and geo-maps best practices across policy domains. Today it holds over 4,500 practices from 216 countries and territories, organised in six catalogues — from digital inclusion (the MEDICI corpus) and gender equality to AI in education, AI in the public sector, city innovation, and public employment services & nature-based solutions.
Three design decisions matter more than the size.
Evaluated, not listed. Every practice is scored 0–100 against a transparent, versioned rubric, with evidence strength classified separately — an RCT and a promising pilot are not the same thing, and the platform says so. Where the evidence is thin, the record says that too.
One knowledge map. Practices are classified in a single ontology across catalogues, so you can ask cross-cutting questions: what works in digital inclusion for older adults? What has been replicated across contexts? Where are the gaps?
Grounded AI. The copilot answers from the corpus and cites the records it used. When the corpus cannot support an answer, it abstains. We think that posture — rigour over fluency — is what institutional users actually need from AI.
What it’s for
For governments and cities: place profiles and benchmarks — what a country or region’s documented practice looks like against the corpus. For researchers: a citable, structured corpus with an open API. For consortiums: a permanent, white-label home for a project’s catalogue — configured in days, alive after the grant. That last one is C-NAPSE’s work: we run Evidoria, and we build observatories on it as project deliverables.
What happens next
Every week, Evidoria publishes an Evidence Signal — a pattern, gap or replication the corpus surfaces as it grows. One of the first is telling: governments are deploying AI faster than they evaluate it — 51 new practices in 30 days, two on strong evidence.
Explore the platform at evidoria.eu. If your project needs its catalogue to outlive its grant, talk to us.