Adform Segment Governance

Adform segment governance

Two governance jobs over first-party segments: enumerate segments with sizes and flag stale or unused ones; inspect segment builder rules and flag logic weaknesses. Read-only.

Connection & tooling

Runs on the Adform GraphQL MCP. Use graphql_execute to run queries. Use graphql_introspect on BuyerAudienceListItem and BuyerAudience to discover available fields including size and composition. Keep calls sequential (~1–2s apart).

---

Step 1 — List buyer audiences for an advertiser

{
  buyerAudienceListItems(
    filters: [{ fieldName: "buyerAudience.advertisers.advertiserIds", operation: intersects, values: ["71883"] }]
    pagination: { offset: 0, limit: 100 }
  ) {
    buyerAudiences {
      id name status
      composition { uidTotalCount }
      stats { campaigns impressions lastImpressionDate }
      createdAt
    }
    totalCount
  }
}

Step 2 — Get audience detail including builder rule

{
  buyerAudience(id: "12345") {
    audience {
      id name status
      ttl frequency idFusion lookalike
      rule { expression recalculable }
      createdAt updatedAt
    }
  }
}

---

Identifying stale and unused segments

• Zero uidTotalCount: audience has no reach — either the rule is too narrow or data has

expired

• lastImpressionDate older than 30 days (or null): segment has not contributed to recent

delivery — check whether it is referenced in any active line item's targeting

• status inactive: segment is disabled

To check whether a segment is referenced by active line items, search for its ID in line item targeting. Cross-reference using adform-line-items.

Identifying rule weaknesses

Inspect rule.expression for:

• No recency window — a segment that includes any historical event with no time limit grows

stale quickly

• Single broad event trigger with no additional AND conditions
• Missing OR/AND structure that would be needed to build a meaningful audience

---

Presenting

Two tables. Usage: segment name, status, reach (uidTotalCount), last impression date, referenced in active targeting (yes or no). Rule audit: segment name, rule summary, issue found. Lead with high-reach segments that are unused, and any segment with an invalid or logic-weak rule that may be wasting data spend.

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.