What is held, from when, and what is not
EURVISTA is only as good as the record underneath it. This page says where that record comes from, how far back each part of it goes, how it is checked, and where it stops. The gaps are on this page rather than buried, because a coverage figure you cannot see is a coverage figure you cannot allow for.
Current holdings
- Legislative files
- Documents
- Recorded votes
- Individual vote casts
- Amendments
- Plenary speeches
- Parliamentary questions
- Registered organisations
- Declared lobby meetings
- Consultation submissions
- Delegated & implementing acts
- National implementing measures
Figures come from the extract manifest, the same file the platform reads. Nothing on this site is typed in by hand, so a number here cannot contradict the platform.
Where the record comes from
Every source is a public register operated by an EU institution or by the institutions jointly. EURVISTA holds no third-party licensed data: nothing on the platform is behind someone else’s commercial terms.
- European Parliament
- The Parliament’s open data API and its public document register: procedure files, committee and plenary documents, recorded votes and individual casts, tabled amendments, plenary speeches, written and oral questions with their answers, member mandates and committee roles, and declared meetings.
- Council of the EU
- The Council’s public register: documents and their cover pages, voting results, national statements and reservations, preparatory body meetings and agendas, and working party comment documents.
- European Commission
- Legislative proposals, the Have Your Say consultation platform, declared meetings of Commissioners, cabinet members and Directors-General, the comitology register of delegated and implementing acts, and Commission press material.
- EU Transparency Register
- Registered interest representatives with their declared goals, estimated costs, staff numbers and the categories they register under. Used to resolve organisation names in declared meetings to a single entity.
- EUR-Lex
- Adopted legal texts and the national implementing measures that member states notify against directives.
- Other public registers
- European Parliamentary Research Service publications, and opinions of the European Data Protection Board and the European Data Protection Supervisor.
How far back each layer goes
Different layers start at different dates, and in most cases the start date is the floor of the source register rather than a choice we made. Where it is a choice, it says so.
- Plenary votes and casts
- September 2009 onward. This is the declared floor of the Parliament’s own published roll-call data: earlier terms are not available in machine-readable form, so their absence is a property of the source, not a gap in the ingest.
- Committee votes
- Recorded committee roll calls published as voting sheets, back to the point where committees began publishing them. Coverage varies considerably by committee, and the platform reports the per-committee count rather than a single headline figure.
- Council votes
- December 2009 onward, from the Council’s published voting results.
- Plenary speeches
- July 2021 onward, which is the floor of the Parliament API’s structured speech data. Earlier debates exist in the verbatim report and are a known extension, not something we claim to hold today.
- Parliamentary questions
- January 2019 onward. This is a scope decision rather than a source limit: it covers the ninth and tenth terms, which is where the demand sits.
- Amendment texts
- 2019 onward for full verbatim texts. Older tabling documents exist in legacy formats that are not reliably machine readable, and are not claimed.
- Declared meetings
- From the start of each institution’s declaration regime. These began at different dates for the Parliament and the Commission, and obligations still differ by office.
- Commission press material
- 1975 onward.
How it is checked
One spine, joined on procedure codes
Records from six registers are joined on the procedure identifiers the institutions themselves use, not on fuzzy title matching. Where a join cannot be made with confidence, the record is held unlinked rather than attached to the wrong file.
Traced end to end on a known file
The pipeline is verified by following a single well-documented file from Commission proposal through committee, plenary, Council and into national transposition, checking each institution’s record against the others.
Verbatim, or not at all
Amendment texts, speeches, questions, answers and consultation submissions are stored as written and attributed. Scanned documents are read by optical character recognition and the output checked, rather than approximated.
Failures surface, they do not hide
Every ingest writes a dead-letter record when a source cannot be read, and a pipeline health view reports them. A run that fetched nothing is a failed run, not a quiet success.
Extraction dates on the record
Each layer carries the date it was last extracted, and pages show it. Where a layer has gone stale beyond its refresh tolerance, the page says so rather than presenting old data as current.
Figures generated, never typed
No count appears in prose anywhere on the platform or this site as a typed literal. Every one reads from the extract manifest, and an automated check fails the build if a number is hard-coded.
What is not held, and what is incomplete
Read this section before relying on the platform for anything that matters.
- Declared meetings are not all meetings
- They are a record of what was declared under each institution’s rules. Obligations differ by office and by institution, and a substantial share of attendee rows in the source data carry no register number, so they cannot be resolved to an organisation at all. The platform reports that resolution rate on the page rather than presenting the resolvable subset as the total.
- Meeting-to-file linkage is partial
- Only a minority of declared meetings name a legislative file in the source record. A per-file meeting view is therefore a floor, not a count, and the platform labels it accordingly.
- Council documents are unevenly released
- The Council releases documents at its own discretion and on its own timetable. Per-year document coverage is materially below complete for recent years, and some documents exist only as a register stub with no released text.
- Amendment outcomes are rarely recorded individually
- Most amendments are voted in blocks. An individual outcome is recorded for only a small fraction, and the platform shows a fate badge only where one genuinely exists.
- Consultation submissions include private individuals
- Have Your Say submissions carry no register number and include citizen campaigns. The platform caps campaign submissions in its counts and says when it has done so.
- No forecasts, ever
- The platform does not predict when a file will pass, whether an amendment will be adopted, or how anyone will vote. Where it reports how often something has happened, it names the cohort and the period, and that is historical frequency rather than a statement about the future.
- No influence scores
- Influence is reported as an ordinal band with its evidence spelled out and the formula versioned. A numeric score would imply a precision the underlying record does not support.
- No third-party licensed data
- Everything is drawn from public institutional registers. Nothing on the platform is subject to another company’s licence terms.
Personal data in the corpus
The corpus contains personal data: members of the European Parliament and other office-holders acting in a public capacity, named interest representatives from the Transparency Register, and individuals who submitted to public consultations. We publish a standing notice explaining what is held about whom, on what lawful basis, and how to exercise your rights, including the right to object and the right to erasure.