Orphex MCP

Orphex Meta Delivery & Audience Review

A Meta delivery assessment separating actual diagnostics, audience/creative hypotheses, nonadditive reach, and supported next tests.

Campaign optimizationMITv2.0.0

When to use

When Meta ad-set delivery or audience context needs investigation before account changes.

Bring the right data

Use your own export, or start with the template. The fictional example shows the expected shape.

Download CSV templateDownload example data
View the input columns
period Required
Matched complete reporting interval
ad_set_id Required
Stable ad-set scope
delivery_status Required
Actual supplied delivery status, not inferred
optimization_event Required
Actual optimization event
currency Required
ISO currency
spend Required
Observed spend
impressions Required
Observed impressions
reach Optional
Reported unique/estimated reach within this row scope
link_clicks Optional
Named link clicks; not all engagement clicks
results Required
Reported named optimized outcomes
Add your business context

Supply your objectives, conversion definitions, currency, constraints, and approved brand facts once, then reuse the profile across reviews. Leave unknown values explicit.

Open profile source
View the versioned resources

See an example

Fictional data · an illustrative review, not a customer result.

Try asking

Review delivery decline and audience saturation hypotheses.

View the example result

Fictional example output

Observed Purchase CPA increased from $20 to $40 (+100%). CPM rose from $10 to $12.50 (+25%); link CTR fell from 2.00% to 1.00%. Within-week frequency rose from 2.00 to 4.00, while reach fell from 50,000 to 20,000.

The supplied status changed to Learning limited; its cause is not established by these counts. The creative edit coincides with the change but is not proof that it reset learning or caused the drop. Frequency 4 is not a universal saturation threshold. Weekly reach is nonadditive, so neither 70,000 unique people nor a pooled frequency can be asserted.

Prioritize the actual delivery diagnostics, current event/attribution settings, change history, and matched placement/audience breakdown. Audience expansion and removing exclusions require checking which settings are actual controls in this campaign. Do not switch to an easier event just to clear status or promise a budget increase will resolve it. No targeting, event, or budget change applied.

Before you start

Review the required context
  • Supplied data with the documented task-specific columns, stable scope, and refresh/maturity context
  • Business definitions and constraints relevant to the decision; see the reusable business-context reference

The method

Align evidence

Record account timezone, currency, objective, performance goal, optimization event, attribution setting/date basis, conversion maturity, entity scope, budget owner, actual status/reason, schedule, audience controls/exclusions, placement settings, and material edits. Keep lead, landing-view, link-click, and purchase outcomes distinct. Verify current account/API field definitions and availability before comparing exports; names and supported controls can differ by campaign type.

View SKILL.md

Orphex Meta Delivery & Audience Review

Review the actual Meta campaign/ad-set delivery context and business outcome. Interpret platform status as a diagnostic signal, not a causal explanation or a reason to change the optimization goal automatically.

Align evidence

Record account timezone, currency, objective, performance goal, optimization event, attribution setting/date basis, conversion maturity, entity scope, budget owner, actual status/reason, schedule, audience controls/exclusions, placement settings, and material edits. Keep lead, landing-view, link-click, and purchase outcomes distinct. Verify current account/API field definitions and availability before comparing exports; names and supported controls can differ by campaign type.

Calculate outcome CPA, link CTR, CPM, and within-scope frequency only when their exact inputs are supplied. Reach and unique audience measures are nonadditive across ad sets, days, or overlapping audiences. Do not sum reach, average frequency, or infer audience overlap from repeated targeting descriptions. Observed frequency has no universal fatigue threshold and cannot prove saturation or self-competition.

Rank delivery explanations

Separate eligibility/policy or instrumentation problems, supplied learning diagnostics, budget/pacing, bid/cost controls, audience availability, placement mix, creative age/edits, and conversion-path evidence. Do not invent a fixed event-volume requirement or assert that every edit resets learning. Status, event counts, and coincident edits alone do not identify the reason; use the actual diagnostic and current official account documentation. A recent incomplete conversion window can explain apparent deterioration but requires demonstrated lag.

Distinguish optimization suggestions/signals from hard audience constraints for the actual setup. Do not recommend broad expansion, blanket exclusions, switching to an easier goal, or overlapping clone campaigns without a business reason and valid supported controls. Prefer one bounded test tied to the diagnosed constraint with a mature review window and a guardrail.

Deliver observations, supported constraints, unresolved hypotheses, next diagnostic/test, and change scope. Preserve purchase/lead-quality goals and keep all account edits within explicit authorization.

Official references

Portable inputs and examples

State whether the result is complete, partial, or blocked for the requested decision. Link material findings to actual supplied rows/sources and separate observed metrics, hypotheses, and estimates. Lead with a short business conclusion, then evidence, uncertainty, and the next measurable check. A data export or installed skill does not authorize account changes.

Source:View on GitHub Open SKILL.md