Design principles
The product is built around inspectable campaign logic. A candidate, consultant, or field lead should be able to understand why a voter is in a segment, how a score is being used, and whether the final universe is appropriate for the outreach plan.
That means VoterFile prefers visible filters and understandable ranking signals over opaque black-box scoring claims.
- Visible filter inputs and segment counts
- Transparent turnout-oriented ranking signals
- Probabilistic demographic planning signals, not identity claims
- Conservative handling when data is incomplete or uncertain
Turnout and prioritization signals
The implemented turnout method is designed to rank campaign effort, not predict support. Where separately approved and activated, it emphasizes participation history, then applies smaller adjustments from features such as age range, permanent vote-by-mail status, registration status, and registration recency.
Campaigns should treat the score as a planning aid for canvassing, mail, phones, and volunteer work, not as a guarantee of future behavior.
Modeled outreach signals
VoterFile has implemented BISG-style methods for broad outreach-planning probabilities, but does not generally activate them. Only a separately approved modeled-output lane may use surname and geography to produce these probabilities, which are never statements of personal identity.
That signal is useful for bilingual planning, community outreach, and segment review, but it should be used with care and local context.
Read the model cards before relying on a score
The turnout and BISG model cards publish the current inputs, formula or transformation, fallbacks, intended use, validation status, and known limitations. They also say plainly where VoterFile has not completed independent validation.
Those model cards are the controlling public descriptions for modeled signals. Shorter product pages should not make a stronger performance or identity claim than the model card supports.
- Turnout propensity v1: deterministic ranking heuristic, not a calibrated probability
- BISG v2: surname-and-ZIP probability estimate, not self-reported identity
- Both models: planning aids only, never candidate-support scores
List construction and exports
Segments are built from campaign geography, voter-file fields, turnout-oriented scoring, and related workflow filters. The product shows counts before export so the team can compare alternatives before money or volunteer time is committed.
Exports remain portable by design. VoterFile focuses on the voter-data layer and leaves downstream execution to the rest of the campaign stack.
Complete California access review before any file upload.
Submit the campaign, requester, issuing source, viewer list, geography, and intended use. VoterFile does not open upload, import, browse, export, or voter-level modeled output until the California activation checklist is complete.
Questions
Does VoterFile produce support scores?
No. VoterFile focuses on turnout, segmentation, and outreach planning signals. It does not claim to know whether a voter supports a candidate or measure.
Can campaigns review the logic before using a segment?
Yes. The product is designed to keep the filters, counts, and score context visible enough for campaign review before export.