01
Ingest
Data Layer
Multi-Source Intelligence Inputs
News, public discourse, legislative and election data, policy signals, paid and earned media, and custom client sources, all ingested live and modeled together.
The Engine
Pharos combines AI, computational linguistics, behavioral science, narrative theory, and proprietary intelligence architecture to transform public discourse into decision-ready strategic intelligence.
The Limits of Conventional Intelligence
They were designed to count what already happened. They cannot explain why narratives are moving, where movement is happening, how messages land differently across audiences, or what action to take next.
Sentiment Snapshots
Surfaces what people said when asked.
Misses
Why opinion is moving. What's coming next. How sentiment varies block to block.
Social Listening
Tracks who's talking and how loudly.
Misses
Whether the volume signals real movement. What it means for the people you need to reach.
Dashboards
Aggregates what's already been measured.
Misses
Strategic recommendations. What action to take, where, with which audience, and why.
Prediction Markets
Aggregates speculation about end-state outcomes.
Misses
The structural drivers of those outcomes. The narratives moving voters before sentiment shifts.
Architecture
The Pharos engine combines five tightly integrated layers. Each layer adds a distinct dimension of intelligence. Together, they convert raw public discourse into decision-ready strategic intelligence. Signals move through them in order: ingest, interpret, score, segment, recommend, and measure, with the Activation layer carrying the last two.
01
Ingest
Data Layer
News, public discourse, legislative and election data, policy signals, paid and earned media, and custom client sources, all ingested live and modeled together.
02
Interpret
Language Layer
Large language models trained for political and policy discourse interpret what's being said, how it's framed, and what it means in narrative context.
03
Score
Behavioral Layer
Decades of behavioral science put to work: why messages resonate, where persuasion is asymmetric, and how different audiences read the same words differently.
04
Segment
Geographic Layer
Narratives don't move uniformly. Pharos scores narrative dynamics at the resolution of a Census Block Group, then rolls up to ZIP, district, DMA, and state.
05
Recommend + Measure
Activation Layer
The stack converges into outputs teams can deploy: activation audiences, strategic briefings, geographic targeting, and recommendations that connect directly to media, outreach, and operations.
Science
The Pharos engine is grounded in disciplines that have shaped how we understand language, persuasion, behavior, and population dynamics. Operationalized for real decision environments, not academic ones.
Discipline 01
Modeling how language carries meaning, frame, and intent at scale across millions of signals.
At Pharos: Powers narrative structure recognition, framing detection, and theme extraction.
Discipline 02
Decades of research on how messages persuade, how audiences interpret, and where decision-making bends.
At Pharos: Drives audience interpretation models and persuasion asymmetry scoring.
Discipline 03
The structural patterns of how stories form, propagate, and reshape collective interpretation.
At Pharos: Frames how narrative momentum is detected, tracked, and projected forward.
Discipline 04
A research-backed framework for the moral dimensions (care, fairness, loyalty, authority, sanctity) that shape how messages land.
At Pharos: Powers the Behavioral Layer's audience-resonance scoring.
Discipline 05
Statistical projection from sampled signals to population-level dynamics, calibrated against ground-truth data.
At Pharos: Translates signal patterns to neighborhood-level intelligence across 240K block groups.
Discipline 06
Engineering practices that make AI-driven intelligence traceable, defensible, and scrutable.
At Pharos: Every recommendation surfaces the underlying signals that drove it.
Modules
The engine surfaces through proprietary intelligence modules called Scapes. Each Scape applies the full architecture to a distinct decision environment, with vertical-specific data, models, and outputs.
Live
Intelligence for live races.
Statewide, Federal, Mayoral, and Local race intelligence. Narrative momentum, geographic opportunity, persuadable audience modeling, and activation guidance.
Live
Live tracking of issue dynamics.
Continuous monitoring of how issue narratives form, move, and land across geographies and audiences. Built for advocacy, donors, and policy teams.
Live
Strategic policy intelligence.
Translates legislative activity into constituent pressure, district movement, legislator alignment, and strategic outreach guidance. Sits above traditional bill-tracking tools.
Live
Ballot measure intelligence.
Calibrated for ballot initiatives and referenda. Geographic activation modeling, narrative pressure detection, and measurement of campaign deployment performance.
Live
The living voter file.
The enrichment layer for the file you already have. Every record comes back with a ranked issue list, a momentum signal with direction, and one message to lead with. Refreshed every week.
Live
Activation-ready audience segments.
Voter-level audiences built from narrative intelligence, split into GOTV and Persuade clusters with the lead issue for each neighborhood. Delivered through the ad-tech stacks your team already runs.
Featured · New
Paid + earned narrative measurement.
By PharosGraph. Measures how paid spend and earned coverage move narrative across neighborhoods, with competitive ad intelligence and in-flight optimization.
See the engine on your terrain
Bring us a race, an issue, or an initiative. We will walk the architecture end to end on your terrain, from the signals that feed it to the recommendation that comes out.
More in How It Works: Narrative Command Center and Intelligence Briefs.