Policy Trace EU↗
A pipeline that traces which submissions actually shaped adopted EU legislation and measures how balanced that influence was.
Every EU directive arrives at the end of a long paper trail. Before the final text, there are Commission proposals, public consultations, position papers from industry, submissions from civil society, expert input - often thousands of pages.
Some of that language survives into the law. Most of it doesn't. Which is which? For almost any given article, almost nobody can say.
The trail is public but nobody follows it
The remarkable thing is that the evidence already exists. Commission proposals, consultation submissions, and adopted texts are all on the record. In principle, you could line them up and see whose wording made it through.
In practice, no one does this systematically. Influence over EU law is constantly asserted by different actors, lobbyists claiming wins, by NGOs claiming losses, by journalists inferring both and almost never measured against the actual text. PolicyTraceEU exists to close that gap.
What it does
PolicyTraceEU is a retrieval-augmented pipeline that traces textual influence through the EU legislative process, matching the language of adopted law back to the submissions that shaped it. It runs at the level of individual articles, across two channels:
- a verbatim channel, which catches near-exact reuse of language from a submission into the final text; and
- a semantic channel, which catches the same idea carried through in reworded form.
Point it at a directive and it will show you, article by article, which Commission proposals, industry submissions, and civil society submissions the adopted language actually tracks.
Why the hard part is not the matching
The value of PolicyTraceEU is in the guardrails that stop it from mistaking coincidence for influence:
- Temporal precedence: a submission only counts as influence if it genuinely predates the text it supposedly shaped. We map sequence before correlation.
- Common-source filtering: if an industry body and an NGO both quote the same pre-existing regulation, that shared language is not one influencing the other. The pipeline factors it out.
- Draft-as-baseline provenance: influence is measured as change from the Commission's proposal, not from a blank page. This isolates influence on the law as it evolved, rather than crediting submissions for simply echoing the starting draft.
A calibration harness keeps the matching tuned, so the alignment scores can have a consistent interpretation across texts.
Measuring pluralism beyond influence
Tracing whose words are used in the implemented law is only half the question. The other half is whether the influence was balanced. The PolicyTraceEU pipeline produces a Pluralism Score, a measure of how evenly the shaping voices are distributed across industry, civil society, and other stakeholders. A law can be built from many submissions and still be dominated by a single interest; the Pluralism Score is designed to make that visible.
Where it is
We built the current prototype covering 11 EU directives with article-level comparison across Commission proposals, industry submissions, civil society submissions, and adopted law.
PolicyTraceEU is Synika's founding line applied to democratic accountability – law and code together, turning a public record no one has time to read into evidence anyone can check. If you work on EU policy, transparency, or lobbying research, we would welcome a conversation.