Methodology

Why not all votes are equal

The Assemblée held 19 114 public votes between 2017 and 2026. Most carry no positional information at all: median turnout is 106 voters out of 577, and most concern a single amendment dealt with in a few minutes.

The site therefore keeps 2 623 votes, under two complementary rules. First, a vote is kept regardless of turnout whenever it decides the adoption of a whole text, a standalone article, a procedural motion or a resolution: the vote of the members present is real even when the outcome is known in advance and others choose to abstain or stay away. Second, for every other vote — the vast majority, individual amendments — only those with at least 350 voters are kept: below that, the vote tells us almost nothing about anyone's position. In both cases, a vote where the chamber was nearly empty (fewer than 30 voters) is discarded: no one was there for the vote to mean anything.

Absence is not a position — abstention is one

Not voting can mean being absent, having delegated one's vote, or simply not being concerned. The site therefore shows no attendance rate: with such a low median turnout, that percentage would be massively misleading.

Abstention is different: it is a vote in its own right, cast by someone present who explicitly chooses to vote neither for nor against. The site therefore distinguishes it from absence and shows it as one of the three possible outcomes of a meaningful vote, on equal footing with for and against.

Tabling a lot is not proposing a lot

Tabling amendments en masse is a well-known obstruction tactic. A single text once drew over 2,400 amendments from one group alone. That is why the number of amendments tabled is never shown without the number adopted — the two rankings are often reversed.

Separately, 44 % of amendments are never discussed. The adoption rate shown is calculated only against amendments that were actually examined.

On group and party pages, the same principle applies at collective scale: the share of a group's amendments concentrated on its most-amended text is published alongside its adoption rate, never alone. A tabling pattern heavily concentrated on few texts with a low adoption rate is the arithmetic signature of an obstruction tactic; one spread across many texts with more effect is another. The site computes both numbers and publishes them side by side — it never labels a group on the reader's behalf.

A parliamentary group and a party are two different things

The reference data distinguishes the group (the formation that sits in the chamber) from the party. A person can belong to one without sitting under the other, and groups reshuffle over the course of a term. The site preserves this distinction everywhere, including in its colour coding.

A vote mostly tells you about a group

Group voting cohesion ranges from 96 % to nearly 100 %. Knowing how someone voted therefore mostly tells you where they sit. That is why tabled amendments — which carry an individual signal — are shown alongside votes.

What is not done yet

Identity resolution across sources (Assemblée, Sénat, HATVP) is not finished: the same person can appear twice, and some links are incomplete. No merge is ever done automatically — each one requires a logged, reversible human validation.

Coverage spans the Assemblée nationale (mandates, votes, amendments) and, since 2026-08-30, the 348 currently sitting senators — mandate and political group in every case; since 2026-08-31, also votes on meaningful votes and group cohesion (2,157 public Sénat votes since 2017, same method as for the Assemblée, see above), floor amendments (96,563 amendments recorded since July 2017, covering 340 of the 348 senators — committee-stage amendments remain out of scope), and a link to the HATVP disclosure for 348 out of 348 senators. That last link is found by matching names against the HATVP's public directory, unlike the direct link the Assemblée itself declares for its deputies: each page cites the source actually used, never the same one by default. When the same person is already documented on the Assemblée side, this site never automatically merges the two sources (non-negotiable rule): their existing page is enriched with their Sénat data, never duplicated. The national register of elected officials is archived but not yet used on the pages. Since 2026-09-01, coverage also spans the 81 currently sitting French Members of the European Parliament — mandate, EU political group, declared national party and X account when given (52 of 81); no votes or plenary speeches yet, as the European Parliament's API does not expose them in a directly usable way.