Political Signal

Forecast poll movement two weeks before the polls do.

Detect shifts in voter sentiment using high-frequency public signal. Built for campaigns, comms shops, and policy desks that cannot wait for the next wave.

How it works

The exact thing your team is trying to call.

Two-week directional change in head-to-head ballot share at the district or state level, with calibrated confidence and issue attribution.

  • District-level signal resolution, not national averages
  • Issue attribution that names the topics moving the needle
  • Candidate sentiment broken out by demographic cohort
  • Crisis flagging when negative coverage compounds
  • Daily refresh with full audit trail per data source

Signal sources

What feeds the political signal model.

  • 01News RSS across 600 regional outlets
  • 02X political timeline samples
  • 03Reddit issue-specific subreddits
  • 04Google Trends issue salience
  • 05Public poll aggregators for calibration only
Sample entities tracked:Senate-OH-2026Governor-NC-2028Mayor-LA-2026

Use cases

What teams actually do with Political Signal.

  • Greenlight reviews for political signal

    Run scoring on a portfolio of candidates and surface the ones whose political signal signals diverge most from their internal narrative.

  • Continuous monitoring

    Subscribe via webhook and receive a payload only when the score crosses a threshold you defined. No dashboards required.

  • Cross-vertical analysis

    Pull the same entity into other verticals to see whether the signals agree or disagree. Disagreement is often where the alpha is.

  • Backtesting before commitment

    Score the last 24 months of decisions against the model and compare hit rates. The calibration_bucket field tells you how much to trust the model in the bands you care about.

API

Calibrated, explainable, ready for production.

Every prediction ships with a confidence interval, the top feature drivers explaining the score, and the five nearest historical analogs the model leaned on.

Available via REST and a typed TypeScript SDK.

See full API reference
GET /v1/predictions/political-signal/{entityId}
{
  "entityId": "race_oh_sen_2026",
  "forecastDelta": 1.8,
  "ci95": [
    0.4,
    3.1
  ],
  "issueAttribution": [
    {
      "issue": "economy",
      "contribution": 0.7
    },
    {
      "issue": "immigration",
      "contribution": 0.5
    },
    {
      "issue": "abortion",
      "contribution": 0.3
    },
    {
      "issue": "crime",
      "contribution": 0.2
    },
    {
      "issue": "foreign_policy",
      "contribution": 0.1
    }
  ],
  "analogs": [
    "race_pa_sen_2022",
    "race_ga_sen_2022",
    "race_az_sen_2022",
    "race_nv_sen_2022",
    "race_wi_sen_2022"
  ]
}

Compared to

How Political Signal differs from what is already out there.

VendorHow SignalGrid is different
Morning ConsultDistrict-level resolution and issue attribution, not national crosstabs.
PollyTwo-week directional forecasts with calibrated confidence, not just static polling.
Zignal LabsModels trained for forecasting, not only narrative monitoring.

Try Political Signal on your own entities.

Start free during the public beta. We will pre-select political signal for you on signup.

Other verticals