Adform Bid Landscape

Adform bid landscape

The CPM-versus-win-rate curve for an RTB line item: what share of auctions you win at each CPM, plus reachable cookies and requests. Use this to answer "what CPM do I need to win X%". Read-only.

Connection & tooling

Runs on the Adform GraphQL MCP. Use graphql_execute(query, variables) to run and graphql_validate to check a query first. Always validate before executing. The queries below are illustrative examples — if a field or input shape isn't shown, or a query fails validation, discover the current schema with graphql_search (lighter, preferred) and fall back to graphql_introspect one call at a time with ≥2s between calls for complex input types, enum values, or union/interface resolution.

---

Building the input

rtbLineItemForecasting(lineItem: ForecastingRtbLineItemInput!) needs a complete line-item shape. Build it by reading the live line item first, then mirroring its targeting, inventory, pricing, periods, and environments into the input.

Method

1. Read the live line item using rtbLineItem(id:) (the id is the RTB setup ID from rtbLineItems(), not the placement ID) to get current targeting, inventory, and pricing
2. Mirror the complete line item structure into the ForecastingRtbLineItemInput payload
3. Run rtbLineItemForecasting with the mirrored input
4. Pair cpms and winRates arrays by index to read the CPM-vs-win-rate curve

Inline execute examples

Read the live line item

{
  rtbLineItem(id: "99999") {
    id name environments orderId campaignId
    inventories { id buyingType deals { id bidPrice } }
    targetings { id name bidMultiplier }
  }
}

Run the forecast

{
  rtbLineItemForecasting(lineItem: {
    name: "Forecast"
    campaignId: "3993873"
    # mirror the full targeting and inventory from the live line item
  }) {
    cpms
    winRates
    reachableCookies
    reachableRequests
  }
}

cpms and winRates are parallel arrays — pair index N of cpms with index N of winRates to read the curve.

---

Presenting

Lead with the actionable answer: the CPM needed to hit the trader's target win rate (interpolate the curve), then show the full curve as context, and the reachable cookies and requests for inventory sizing. State that forecasts are estimates and cross-check with adform-past-traffic if volume looks unexpected.

---

Supplementary: bid reason context via mcpStats

After sizing the CPM needed to win the target share of auctions, cross-check with a mcpStats bid reason query. This shows whether current losses are primarily price-driven (the bid landscape addresses this) or driven by other factors (creative audit, targeting, budget) that a CPM increase alone will not fix. Validated query:

{
  mcpStats {
    totalRowCount
    totals
    columns {
      dimensions { bidReason { name } }
      metrics {
        rtbBids
        lostBids
        bidReasonCount
        impressions
        rtbWinRate
      }
    }
    rows(
      filter: {
        date: { from: "2026-06-01", to: "2026-06-30" }
        advertiser: { ids: ["2133936"] }
      }
      paging: { offset: 0, limit: 50 }
      sort: [{ column: 0, direction: desc }]
    )
  }
}

If the dominant bid reason is price-floor-related, the bid landscape CPM recommendation is the primary fix. If non-price reasons dominate, flag them to the trader alongside the CPM recommendation so they are not surprised that raising the bid alone does not resolve the win rate shortfall.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: adform
• Source: adform/agentic-skills
• License: MIT
• Homepage: https://site.adform.com

Install and usage instructions live in the source repository linked above.