Methodology
Every number displayed on this site is computed from a transparent formula and traceable to the underlying data. This page lists every methodology currently in use across the platform, with the version of the formula, what is included, what is excluded, and known limitations.
Methodologies are versioned. The score you see today carries the version of the formula used to compute it. When a methodology is improved, the new version replaces the old, and affected metrics are recalculated.
Key figures, defined
These terms mean different things and are never used interchangeably on this site.
Example: France holds 81 seats; 82 MEPs served during the term because one seat changed hands. We show this as “81 seats · 82 this term”, never “82 of 81”.
cosignaturesCounts how often two MEPs appear as authors on the same parliamentary document or question.
edge_count(A,B) = count(distinct documents where A and B both have participation role AUTHOR)
delegation-cohesionHow unified a country's delegation is across all roll-call votes.
cohesion = average over votes of (majority_position_count / expressed_positions_count)
group-loyaltyHow often a MEP votes with the majority position of their European political group.
group_loyalty = votes_matching_group_majority / eligible_expressed_votes
attendanceShare of eligible roll-call votes where the MEP cast a position (For, Against or Abstain).
attendance = votes_cast / total_eligible_votes
similarityHow often two MEPs cast the same expressed position across votes where both were present.
similarity = same_position_votes / comparable_expressed_votes
delegation-loyaltyHow often a MEP votes with the majority position of their national delegation (MEPs from the same country).
delegation_loyalty = votes_matching_delegation_majority / eligible_expressed_votes
topic-rulesAssigns one primary and up to 3 secondary topics per vote using keyword rules, concept mappings, and country detection.
confidence = min(0.98, 0.45 + min(0.4, total_weight × 0.18) + min(0.15, evidence_count × 0.05))
importancePrioritisation indicator (0–100) combining 8 transparent factors per vote.
score = vote_type + legislative + topic + group_division + country_division + close_vote + turnout + rebel_votes; score = min(100, score)
Why this matters
Public-interest data tools must be auditable. Every score on this platform can be traced back to its formula, the input data, and the version of the methodology used. Where applicable, scores carry a "Why this number?" link that opens the audit detail for that specific value.
Data sources are the European Parliament's official roll-call records. The platform's code is open source under MIT; data is published under ODC-By 1.0; analytical content under CC-BY 4.0.