That works if
- Fit:your audience is obvious
- Fit:your category is well-covered in platform taxonomies
- Fit:your signal is dense
Watt Audiences — for performance teams and agencies
Watt Audiences — natural-language audience generation across 145,000+ signals on people. Installed as a Claude skill in your agency's environment. Generate, refine, ship.
CAC down up to 80%. Conversion up to 10×.
From customers running this today.
Watch it work
A real, niche audience, built in Claude in under two minutes.
Illustrative — actual compositions vary by brand.
Curious what this looks like for your shop?
The squeeze
The data industry has been selling broken signals for years. Performance marketers running volume have felt it every day, and most won't say it out loud.
The platforms want you targeting broadly. They take your seed data, expand it, and optimize toward a stable ROAS band — usually 2–3×. Enough to keep budgets flowing. Not enough to meaningfully improve efficiency.
That works if
It breaks if
The result: you end up cycling campaigns constantly, trying to recover performance from a system designed to flatten your inputs.
Sources: Paddle, How is CAC Changing Over Time (2024). AdExchanger, More Performance, Less Transparency: Inside Meta's Advantage+ Shopping Black Box (Dec 2022). Shopify Q1 2025 earnings. Meta Advantage+ product documentation.
The alpha now sits in the signal layer underneath.
What you actually get
Watt Audiences is built for AI-native marketers. We install a Claude skill that composes audiences the way your best operator would, if they had access to 145,000+ real-world behavioral signals.
Inside the Claude org your team already uses, it:
No UI to learn. No seat licenses. You own the logic.
In production today
Their clients sell physical products where the last-mile sale is geographic, so they built their audiences around 25–50 mile radii drawn off specific points on a map. They scaled from 200 campaigns per client to 1,000+ in the same headcount.
Their team adjusts the ratio of signals continuously — a tighter household-income gate one week, a wider category affinity the next — and the audiences recompose against the latest behavioral data overnight. The product isn't replacing their expertise. It's the surface their expertise composes on.
Campaigns per client
Radius granularity
Audience refresh
Customer outcomes
$100M beauty brand customer cut acquisition cost by 80% through autonomous ICP analysis and lookalike agents built on Watt.
“We build and sell data products for a living, so we aren't easily impressed. Most 'AI-ready' data is a warehouse with an MCP bolted on. Watt is the first thing I've seen where an agent can actually reason across the signals and compose things nobody pre-built.”
National men's health customer increased top-of-funnel conversion rate 24× through autonomous audience refinement agents running on the Signal Graph.
“Every marketer claims they target the right people. Ask them what signals they're actually using. They don't know. Watt does.”
Self-qualify
This works best for agencies and operators who already know what they're doing — we make your power users 10× more effective, not your beginners passable. You're a strong fit if:
What this is built for
Three real-world patterns from agencies running this today. Each one is the same shape of problem under different vertical clothing.
Niche products · thin signal
Meta's prebuilt interests don't know what a bioplastics buyer looks like. Watt surfaces:
Regulated industries · sparse menus
When the platform's interest taxonomy has nothing useful and demographic categories are off-limits, Watt surfaces behavioral signal that does the same job — legally:
Hyper-local DTC
Stores, dispensaries, service businesses — anything where the customer has to physically get there. Watt builds audiences as compositions of:
If your audience is hard to find on Meta's dropdown menus, that's exactly the audience Watt is built for.
The engagement
A 3–5 day forward-deployed engagement. 30–60 minutes of your team's time total. We handle the rest.
01
Interview
30–60 minutes with your team. What makes your campaigns good, which signals you weight, the verticals you serve, the platforms you push to.
02
Build
We build the Claude skill(s) tailored to your agency's workflow — your team's Signal Engineer composes the prompts, the signal taxonomies, the channel outputs. 3–5 days end to end.
03
Install
You install it in your Claude organization. Your team starts prompting it daily. We're available for tuning while you ramp.
04
Own it
The skill, the prompts, the audiences — yours. No platform login, no consulting retainer, no long-term lock-in.
CAC down up to 80%. Conversion up to 10×. From customers running this in production today.
For consumer brands and performance agencies ready to compete on signal nobody else has. Let's start.
