> ## Documentation Index
> Fetch the complete documentation index at: https://docs.adcontextprotocol.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Example campaign discovery requests

> Worked examples showing how buyer agents decompose natural-language campaign requests into briefs, filters, targeting overlays, and future targeting support.

The examples below start with what a human says, then show the structured
`request_proposals` request an agent should send. The goal is not to make the brief
verbose. Preserve campaign meaning in prose and move exact values into typed
fields.

## Local service launch

### Human request

> Help Acme Home Services drive appointment calls from homeowners in Colorado
> during October. We have \$8,000 and only want display inventory priced in USD.

### AdCP request

```json theme={null}
{
  "idempotency_key": "550e8400-e29b-41d4-a716-446655441101",
  "brand": { "domain": "acmehome.example" },
  "brief": "Drive appointment calls from homeowners who may need urgent or planned home repairs. Favor trusted local and home-improvement contexts.",
  "criteria": {
    "offer_filters": {
      "channels": ["display"],
      "pricing_currencies": ["USD"],
      "start_date": "2027-10-01",
      "end_date": "2027-10-31",
      "budget_range": { "max": 8000, "currency": "USD" }
    },
    "targeting_overlay": {
      "geo_regions": ["US-CO"]
    }
  }
}
```

The seller interprets “homeowners” and contextual fit. It does not infer the
state, channel, currency, dates, or budget.

## Multi-country B2B campaign

### Human request

> Pinnacle Cloud wants qualified leads from technical leaders at growing
> businesses in the US, Canada, UK, and Germany. We need display or online
> video, USD or EUR pricing, and completed-view reporting where video is used.

### AdCP request

```json theme={null}
{
  "idempotency_key": "550e8400-e29b-41d4-a716-446655441102",
  "brand": { "domain": "pinnaclecloud.example" },
  "brief": "Generate qualified leads from technical leaders modernizing multi-cloud infrastructure. Favor trusted technology, security, and business contexts.",
  "criteria": {
    "offer_filters": {
      "channels": ["display", "olv"],
      "pricing_currencies": ["USD", "EUR"],
      "required_metrics": ["completed_views"]
    },
    "targeting_overlay": {
      "geo_countries": ["US", "CA", "GB", "DE"]
    }
  }
}
```

The brief carries job-function meaning because AdCP does not impose one global
taxonomy for “technical leader.” Geography and product requirements are exact.

## Known country, cities selected later

### Human request

> Nova Meals is launching nationally to parents ages 25–44. After we compare
> forecasts, our planners will choose separate DMAs and placements for each
> package. We do not want one product returned per city.

### AdCP request

```json theme={null}
{
  "idempotency_key": "550e8400-e29b-41d4-a716-446655441103",
  "brand": { "domain": "novameals.example" },
  "brief": "Launch a premium family meal service to time-pressed parents. Favor trusted family, food, and lifestyle contexts.",
  "criteria": {
    "targeting_overlay": {
      "geo_countries": ["US"],
      "demographics": {
        "age": { "min": 25, "max": 44, "include_unknown": false }
      }
    },
    "required_overlay_support": {
      "geo_metros": { "systems": ["nielsen_dma"] },
      "placement_selection": true
    }
  }
}
```

`targeting_overlay` is applied now and participates in every returned forecast.
`required_overlay_support` filters to products that let packages select DMAs
and placements later. It does not create a product or package per value.

## Hard prose fallback

Structured-first authoring is a SHOULD, not permission for sellers to ignore
prose. If a buyer sends:

```json theme={null}
{
  "idempotency_key": "550e8400-e29b-41d4-a716-446655441104",
  "brand": { "domain": "acmeoutdoor.example" },
  "brief": "Hard requirements: US delivery only and ages 25–44. Favor outdoor enthusiasts."
}
```

the seller still applies the hard constraints and, because they affect product
eligibility, pricing, and forecasting, must confirm the structured interpretation
in `targeting_resolution.brief_targeting`. The buyer should not infer
confirmation from a missing resolution.

Legacy callers can express the same decomposition through `get_products` with
`buying_mode: "brief"`, top-level `filters`, `targeting_overlay`, and
`required_overlay_support`. The split tasks are preferred because they make
the listing, proposal, and refinement lifecycle explicit.

## Practice

Decompose this request before revealing the answer:

> Acme Audio has \$75,000 for a six-week US podcast launch. Only USD pricing.
> We need completion reporting, want music and technology enthusiasts, and will
> choose individual shows after reviewing the publisher's recommendations.

<Accordion title="Show one valid decomposition">
  * Brief: launch context plus music and technology enthusiast intent.
  * Offer filters: podcast channel, USD currency, six-week dates, \$75,000 budget, and
    required completion metric.
  * Targeting overlay: US.
  * Required overlay support: placement or collection selection later, depending
    on how the seller models public shows.
</Accordion>

## Review every response

Before purchase:

1. Confirm offer filters excluded ineligible offers.
2. Confirm forecast and price reflect the structured overlay.
3. Verify `overlay_support` covers every later-selectable requirement.
4. Review `targeting_resolution.modifications` and any `brief_targeting`.
5. Accept a configured product only when its complete targeting resolution is
   acceptable.