Signals protocol

Follow a media buyer from campaign brief to signal activation across automotive, geo, and retail data — a visual walkthrough of the AdCP signals workflow.

Sam is a senior media buyer at Pinnacle Agency. His client, Nova Motors, is launching the Volta EV — their first electric vehicle. The brief: reach in-market auto buyers with high purchase propensity, near dealerships, and focus on consumers who haven’t bought a Nova vehicle before.

Without AdCP, Sam would be emailing three data providers for segment availability, waiting for IO sign-offs, then uploading CSV segment files to two separate DSP platforms — a process that takes days and breaks every time a provider updates their taxonomy. With AdCP, his agency platform handles it in minutes.

This walkthrough follows Sam from brief to live targeting.

Step 1: Describe what you need

Sam starts with what he wants to accomplish, not a segment taxonomy.

Sam’s agency platform translates the brief into a get_signals call. No need to know which providers have what — the signal agent searches across all of them:

{
  "$schema": "https://adcontextprotocol.org/schemas/v3/signals/get-signals-request.json",
  "signal_spec": "In-market EV buyers with high purchase propensity, near auto dealerships"
}

The signal agent searches catalogs from every authorized data provider — automotive, geo/mobility, retail, identity — and returns what matches.

Agency language → protocol terms

What Sam says	What the protocol calls it
Target audience	Signal (from get_signals)
Segment taxonomy	Signal catalog (adagents.json)
Data provider	Signal source (data_provider_domain)
Activate on my DSP	activate_signal with destination
Audience size	coverage_percentage in signal response
Data cost	pricing_options in signal response

Step 2: Review what comes back

The signal agent returns matches from multiple providers. Each signal has a value type, pricing, and coverage estimate:

Sam sees signals from three providers he didn’t have to find or negotiate with individually:

• Trident Auto Data — purchase_propensity (numeric, 0-1 score). CPM: $1.50. likely_ev_buyers (binary). CPM: $2.50.
• Meridian Geo — competitor_visitors (binary, people who visited competing dealerships). CPM: $2.00.
• ShopGrid — new_to_brand (binary). CPM: $3.50. category_buyer (categorical: electronics, automotive, home). CPM: $3.00.

Sam notices that purchase_propensity is numeric — his agent can set a threshold (score > 0.7) to focus budget on high-intent prospects rather than a blunt include/exclude. He also sees that category_buyer is categorical, so he can target “automotive” specifically without paying for the full ShopGrid audience. The competitor_visitors signal catches his eye — it comes from Meridian Geo, a data company founded by Kai Lindström to make location and behavioral data accessible through open protocols. Sam picks three signals: purchase_propensity for intent scoring, Meridian Geo’s competitor_visitors for conquest targeting near rival dealerships, and new_to_brand to reach households that haven’t bought a Nova before.

Step 3: Verify and select

Before activating third-party data, Pinnacle Agency requires verification. Sam’s platform fetches the data provider’s catalog directly:

https://shopgrid.example/.well-known/adagents.json

And confirms:

1. The new_to_brand signal exists in ShopGrid’s catalog
2. The signal agent is listed in authorized_agents
3. The authorization covers retail signals (via signal_tags: ["retail"])

If the authorization check had failed — say ShopGrid had revoked the agent’s access — Sam would see the signal flagged before spending a dollar on it. This independent verification means buyers don’t have to take the signal agent’s word for data provenance.

Step 4: Activate on your platforms

Sam activates his three signals on two DSPs — Nova DSP for programmatic display (broad reach, lower CPMs) and StreamHaus for premium CTV inventory (household-level targeting for the brand spot):

Each activation call specifies the destination platform and account:

{
  "$schema": "https://adcontextprotocol.org/schemas/v3/signals/activate-signal-request.json",
  "signal_agent_segment_id": "shopgrid_new_to_brand",
  "pricing_option_id": "po_shopgrid_retail_cpm",
  "destinations": [
    {
      "type": "platform",
      "platform": "streamhaus",
      "account": "agency-ctv-seat-456"
    }
  ]
}

The signal agent pushes segment membership to each platform. Sam gets back deployment IDs he can reference when building media buys.

Step 5: Build the campaign

Now the signals are live on both platforms. Sam’s agent builds media buys using create_media_buy. The activated signals are already available as targeting segments on each DSP — the media buy references the products discovered via get_products, and the DSP applies the signal-based targeting automatically.

• Display (Nova DSP): Sam selects a display product that targets purchase_propensity > 0.7 AND competitor_visitors = true. This reaches high-intent auto buyers who’ve been visiting competing dealerships.
• CTV (StreamHaus): Sam selects a CTV product targeting new_to_brand = true. This reaches households that haven’t purchased a Nova vehicle with a brand awareness spot.

The key point: signals and media buys are separate concerns. The Signals Protocol gets data onto platforms. The Media Buy Protocol gets campaigns running on those platforms. They compose together — Sam didn’t need a custom integration between his signal providers and his DSPs.

Step 6: Manage and measure

The campaign runs. Two weeks in, display CPAs are running 40% above target while CTV is pacing well. Sam reallocates: he deactivates the geo signal on Nova DSP to reduce display data costs:

{
  "$schema": "https://adcontextprotocol.org/schemas/v3/signals/activate-signal-request.json",
  "signal_agent_segment_id": "meridian_competitor_visitors",
  "action": "deactivate",
  "destinations": [
    {
      "type": "platform",
      "platform": "nova-dsp",
      "account": "agency-display-seat-123"
    }
  ]
}

Deactivation removes the segment from the platform. Billing stops. The CTV signals stay active — each activation is independent.

Every step uses a standard AdCP task. Sam didn’t need to know which data providers exist, negotiate individual contracts, or build custom integrations per DSP. Data providers like Kai Lindström’s Meridian Geo publish their catalogs once, and the signal agent handles discovery across all of them. Each platform handles targeting with its standard tools once segments arrive.

Go deeper

• Key concepts: Signal types, sources, and authorization — the building blocks behind this walkthrough
• Ecosystem: Who participates and how — data providers, retailers, publishers, CDPs, agencies, identity companies
• Publish your data: Data provider guide — how to create a signal catalog
• Protocol spec: Signals specification — formal requirements and conformance
• Get certified: The Signals specialist module teaches signal discovery and activation through interactive labs with a sandbox signal agent