Explore the graph
Traverse the Signal Graph to see what's possible, what's relevant, and what's actually there.
The Signal Graph holds hundreds of thousands of raw signals, and roughly a third of them refresh every day. There's no catalog to browse and no canned segment list — /watt:explore traverses the actual graph, so what you see is what's possible, what's relevant, and what's actually there. Not a pitch; the graph itself.
This is the first step of signal engineering. You have a hypothesis — somewhere in those signals is an edge for your business — and the first move is to see what you can find. Explore is where you test it: see what's indexed, refine the angle, and only then build.
The mechanics are simple: describe who you're trying to reach in plain English. Watt searches the graph live, shows you the signals behind the idea — each with its size and freshness — and asks one concrete question at a time. Your answer steers the next search.
Explore is read-only. It discovers and describes; it never builds, counts, or exports anything. When you want the actual people, everything you found carries straight into Build an audience.
Start exploring
Type the command, or just ask — any question about what's out there for a group of people routes here.
A few asks that land here:
/watt:explore what signals are there on trail running?/watt:explore is Watt good for reaching first-time homebuyers?/watt:explore what's adjacent to running gear that I could use as a proxy?/watt:explore how do I know which of these signals are good?/watt:explore are these signals good for a prospecting campaign?
The first time Watt touches the graph in a session, a browser window pops for sign-in — Watt mentions it before it happens.
What it looks like
Every turn has the same shape: Watt searches the graph, shows you what came back, and ends on one concrete question. Your answer steers the next search. A few beats in:
You: what's out there for people getting into trail running?
Watt: Searched the trail-running side — 12 signals, strongest ~800K and fresh. (a card of candidates: In-market: Trail Running Shoes · ~640K · fresh — Interest: Ultramarathons · ~210K — Visits specialty running stores · ~1.1M · fresh — +9 more) Which feels like your people — the gear buyers, the racers, or the broader interest?
You: gear buyers — and everyone has to be in the Pacific Northwest
Watt: Keeping the gear angle. The Northwest works as a must-have — everyone in your set has it — so I'll hold it as a gate. Go deeper on gear, or look at an adjacent angle like race registrations?
That's the whole interface. You never fill out a form or learn a query language — you answer the question in front of you, in your own words, and the territory opens up from there.
Ways to steer it
There's no syntax to learn — tell Watt what to change and it changes. A few moves you'll use constantly:
| You say | What happens |
|---|---|
| "go deeper on the gear angle" | A thorough sweep of every signal behind that one concept, strongest first. |
| "keep 1 and 3" | They join your signal stack — tracked, shown back to you, ready to build from. |
| "everyone has to be a homeowner" | A must-have — held as a gate on everything else in the stack. |
| "not wedding planners" | An exclusion — confirmed with you before it's applied, then never resurfaced. |
| "how do these stack up?" | A read of each kept signal: relevance to your idea, freshness, how rare and specific, how broad. |
| "what else is out there?" | Adjacent concepts, contrast angles, and whole domains you haven't touched yet. |
| "that's enough" | A summary of the territory — what's strong, what's thin, what's unexplored — and your stack, ready to run. |
What a good ask looks like
The graph indexes what people do — behaviors, intent, purchases, life events — so asks built on a behavior light it up, and asks built on a label alone don't:
| Weak | Stronger |
|---|---|
| "women 25–34" | "people comparing meal-kit services" — a demographic can gate an idea, but a behavior defines one |
| "affluent households" | "people shopping for a second property" — name the behavior behind the label |
| "the outdoors segment" | "people who bought backcountry ski gear this season" — the graph holds raw signals, not vendor segments |
You don't have to get this right up front — explore exists to sharpen a vague idea against what actually exists. Start anywhere; the searches will show you where the substance is. And when there's nothing tight for an angle, Watt says so and offers the closest real thing.
Your signal stack
Every signal you keep joins your signal stack — the reusable shape of your audience idea, with must-haves and exclusions marked. The stack survives the session, and you can run it over and over: take it into Build an audience to generate the list, refine it after a read, run it again. You never copy anything by hand.
What explore won't do
Explore stops at understanding, on purpose. Knowing the boundary saves you chasing a dead end:
| You ask | What happens |
|---|---|
| "How many people match all of this together?" | Combined reach is generated on the fly and measured when you build — never signal sizes added up. That's Build an audience. |
| "Get me the list" | Take your signal stack into Build an audience to generate the list of members. |
| Mixing people and businesses in one signal | A signal describes one entity type at a time — a person or a business — so you explore each side on its own. Crossing between them (these people → their employers) happens in a build. |
| An audience outside the US, or about minors | The graph is US-only and adults-only. Ideas about kids pivot to parents and guardians of that age range. |