Simulated floor vote
Thousands of simulated votes for each bill, with party positions and each legislator's likelihood of breaking ranks.
Platform
The platform our team uses, open to clients.
We built our own platform for legislative and government data. Clients see the same screens our team uses in its analysis.
Chamber · 513 seatsfictitious example
lean Yes · 268undecided · 41lean No · 204
Chamber
Each legislator's likely vote.
Each legislator's likely vote, the chance of approval with an uncertainty range and, in amber, the undecided members: that is where the work is.
Modules
What the platform does.
Stakeholder profiles
Record, issues, committees, constituency and degree of influence of each decision-maker.
Government scenarios
The same issue with the government for or against, and under different governments, side by side.
Agenda radar
Bills and rules in progress, with stage, rapporteur and chance of reaching a vote.
Engagement plan
Priorities, arguments, owners and next steps, with export to a report.
Forecast ledger
An append-only record: the forecast before the decision and the outcome after.
Open method
Tests and limits.
We tested the model by re-running past votes using only what was known the day before. We publish both the hits and the limits.
What the test showed
- 372 roll-call votes in Brazil's Chamber of Deputies re-run without looking ahead (2023 to 2026).
- The test found a flaw: the center follows the government much more when it votes Yes. Once fixed, the error (Brier score) fell from 0.149 to 0.123.
- A bill's official topic can be inferred from its summary with 89% accuracy.
Limits we disclose
- The chance of approval applies once the bill reaches a vote; whether it gets there is a separate question, answered by the radar.
- Voice votes have no individual record and are not part of the per-legislator forecast.
- Change-of-government scenarios are extrapolations from the past, with a larger error.
- So far, our tests cover the Brazilian Congress.
Demo