Built so any builder can traverse the world's signal agentically. Your agents don't join tables. They navigate 15 trillion pre-computed relationships in plain language, at sub-second speed across petabytes of behavioral data.
Architecture
Watt's data architecture has two layers.
The base graph is a heterogeneous information network. It stores nodes (people, businesses, emails, mobile IDs) connected by edges that represent relationships. Unlike a table, it captures how things relate, not just what they are.
The Signal Graph sits above it. Every day, Watt calculates five types of relationships across the entire base graph: top predictors, top discriminators, top co-occurring, top exclusionary, and top absentee. That produces 15 trillion pre-computed relationship outcomes, recalculated daily.
This is what makes sub-second petabyte-scale queries possible. Your agent doesn't traverse raw data. It traverses a pre-computed map of how everything in the graph relates to everything else.
The Signal Graph
145,000+ signals on people, 55,000+ on businesses, and 15 trillion pre-computed relationships, tied to verified identities and traversable by any agent in plain language.
How it works
Watt ingests 145,000+ signals on people and 55,000+ on businesses from established data providers: purchase patterns, life events, employment, firmographics, identifiers. US-only, privacy-compliant, consent-flagged.
Every signal becomes a node in a heterogeneous graph linking people, companies, and the behaviors that connect them. 65 billion nodes. 250M+ verified people. 60M+ businesses.
Watt pre-computes 15 trillion relationships across the graph every day: predictors, discriminators, co-occurrences, exclusions. Your agent traverses them in plain language.
A simple Claude plugin. Thirty minutes from signup to a finished audience pushed to your channel of choice. Built by one Signal Engineer in your Claude org, with the analytical horsepower of a 200-person data team behind it.
Differentiators
Data isn't stored in flat tables. It lives in a traversable graph. Agents navigate relationships between identities, companies, and signals without upfront schema knowledge. The graph adapts to the query.
A production-hardened Model Context Protocol layer with rate limiting, audit logging, and role-based access. Drop Watt into any agent framework with a single config line. No custom glue code required.
No schema to memorize. No joins to write. Your Signal Engineer asks for what the agent needs and Watt traverses the graph. Discovery and activation share one query surface.
Data domains
Watt's graph covers signal categories spanning intent, purchase, lifestyle, financial, demographic, employment, political, content, household, and affinity for people, and tech stack, hiring signals, and industry for businesses. They're not separate silos. They're intermingled in a single traversable graph.
Name, age, household composition, education, interests, and 12,000+ demographic signals unified across sources. Part of Watt's unified signal graph.
Firmographic data, org charts, funding history, technology stack, and headcount signals for 300M+ companies. Part of Watt's unified signal graph.
Browsing intent, keyword affinity, and research patterns to surface in-market signals in real time. Part of Watt's unified signal graph.
Purchase behavior, spend patterns, category affinities, and wallet-share estimates across retail and digital. Part of Watt's unified signal graph.
Signal categories evolve as new data is onboarded. Use Watt Chat to discover what's available for your use case.
Entity Roadmap
DMV data already acquired. Household and location entities next.
Every vertical, the same Signal Graph
Data minimization and purpose limitation baked in at the architecture level.
Dedicated legal team specializing in data privacy, CCPA, and GDPR compliance.
Clear acceptable use policies with enterprise MSA available on request.
We'll map your agent architecture against the Signal Graph and show you what one Signal Engineer can ship.
