Paid-media teams diagnosing performance changes before acting / media buying

Media Buying Performance Diagnosis

A source-linked diagnosis with ranked hypotheses, evidence gaps, and a recommendation for no change, budget review, creative work, or an experiment.

4 stages2 specialist agentsDepends on the selected path

The problem

When paid performance worsens, teams jump to budget or creative changes before separating measurement, audience, channel, page, and market causes.

AI can compare multiple causal categories and time windows systematically while keeping uncertainty and business judgment visible.

What you do

  1. Name the affected accounts and campaigns, what changed, the KPI, and the comparison window.
  2. Provide exports or select one read-only provider route.
  3. Review the diagnosis before starting any mutation workflow.

What the agent does

  1. Validate measurement and comparison quality before explaining performance.
  2. Use stored evidence or one ready metric-source route; unavailable alternatives stay optional.
  3. Rank causal hypotheses, name counter-evidence and gaps, and recommend the smallest useful next step.
  4. Stop after diagnosis unless the operator starts a follow-up workflow.

When this workflow helps

A reusable method, adapted to the request.

  • A campaign, channel, creative set, or audience needs a performance review.
  • The agent should retrieve only relevant historical fields before diagnosing.

Before it starts

Depends on the selected path

  • Affected accounts and campaigns
  • The exact change date, KPI target, comparison window, and attribution assumptions
  • Stored snapshots, exports, or one ready read-only provider route

What proves it worked

Evidence, not a success claim.

  • Measurement-quality check
  • Evidence table and ranked hypotheses
  • Counter-evidence and data gaps
  • Recommendation with explicit no-mutation boundary

Workflow path

The reusable stages of the work.

A real run expands these stages around the request, context, selected tools, approvals, and dependencies.

4 ordered workflow stages
  1. 01

    Orient From Performance Context

    Read the review scope, metric targets, recent snapshots, related learnings, and active experiments.

    Passes to next stage
  2. 02

    Fetch Additional Metrics

    Execute only the selected read action that the concrete run plan grants for missing evidence.

    Passes to next stage
  3. 03

    Diagnose Drivers

    Compare evidence against targets and separate likely channel, audience, creative, landing page, and budget drivers.

    Passes to next stage
  4. 04

    Record Followups

    Store recommended follow-up resources, budget review notes, and learning candidates for later review without executing a mutation.

    Verified outcome

Safe stopping and recovery

Useful even when the whole path cannot run.

  • This workflow is recommendation-first and stops before budget, creative, page, or campaign mutation.
  • If evidence is insufficient, recommend no change and list the minimum missing data.
  • If a live provider is unavailable, use supplied exports or stored snapshots rather than blocking on every provider.

Documented follow-on work

The next workflow is conditional, not hidden.

Specialists inside this workflow

Clear roles for each part of the job.

Connected work

Built to use the tools you already have.

Meta AdsGoogle AdsTaboolaCustom Media Tool