What you need to know
- Audit one authorized Meta account at a time and keep campaigns with different objectives separate.
- Distinguish saved targeting from observed delivery. A configured audience is not a report of who saw an ad.
- A good finding includes the object ID, evidence, uncertainty, and a specific next step.
Prepare an audit brief you can verify
Start with the business question. “Why did qualified leads become more expensive?” is more actionable than “fix Facebook Ads.” Record the workspace, ad account ID, objective, reporting window, and definition of a result. Decide whether you are auditing Facebook placements alone or the account’s Facebook and Instagram delivery together.
A Meta account can contain campaigns optimized for different outcomes. Compare lead generation with lead generation and traffic with traffic before deciding where to move budget. If the retrieved report does not distinguish the outcomes you need, treat that as missing evidence instead of combining unlike results.
Use this six-part checklist
| Check | Evidence to collect | Question to answer |
|---|---|---|
| Connection | Selected account and permission-readiness result | Are we looking at the intended client? |
| Delivery | Campaign, ad-set, and ad status with available metrics | Is the object paused, restricted, or actually delivering? |
| Measurement | Result definition, date window, attribution settings if available | Are the periods and outcomes comparable? |
| Audience | Saved targeting plus an actual delivery breakdown | Does delivery support the audience hypothesis? |
| Creative | Ad details, creative configuration, and ad-level performance | Is the message aligned with the destination and offer? |
| Decision | Exact IDs, proposed changes, and validation result | Can another person review the action before it happens? |
Work down the list before editing a budget. A disconnected account is an access problem; a paused ad is a status problem; a low conversion count may be a measurement or traffic problem. The same recommendation will not fix all three.
Separate delivery problems from performance problems
When an ad has no spend, retrieve its campaign, ad set, and ad details. Check whether a parent object is paused and inspect the available status or error information. Do not interpret zero delivery as proof that the creative is bad. If the API does not return the reason you need, check the native Ads Manager diagnostics.
For an account with spend but weak outcomes, inspect where the change happened. Did impressions fall, did clicks become more expensive, or did results per click change? Explain the evidence chain without presenting correlation as a proven cause. A report can identify where to investigate; it does not observe every step of a customer’s journey.
Review creative and targeting as separate hypotheses
Ask for the ad’s creative configuration and the ad set’s saved targeting. Then ask for the available delivery breakdown. These answer different questions: what the campaign was configured to do, what message it used, and where delivery occurred. A saved audience description alone cannot establish the demographics of people who saw the ad.
Avoid a universal “creative fatigue” threshold. Rising frequency alongside weaker response may justify a creative test, but audience size, campaign purpose, placements, and time window matter. Review the image or video itself in the native interface if the assistant only receives metadata. Do not say a visual has been inspected when no image was actually retrieved.
Reconcile the numbers before changing the plan
Match the account, dates, time zone, filters, and result columns between the assistant’s output and Ads Manager. Check attribution settings where available. CRM-qualified leads and platform-attributed leads answer different questions, so report both with clear labels rather than silently replacing one with the other.
Use the same denominator for comparisons. Cost per result should be calculated from total spend divided by the relevant total results, not by averaging ad-level ratios. If the result count is zero, leave the ratio undefined. If revenue is missing, say ROAS cannot be calculated from this report; do not substitute clicks or leads for revenue.
Finish with a proposal, not an unreviewed change
Request a short list of actions ranked by the strength of their evidence and their business impact. Include the exact account and object IDs, the old and proposed values, the rationale, and the metric that will determine whether the change helped. Keep uncertain ideas in a separate test backlog.
PlugMind supports Meta campaign, ad-set, and ad workflows, but each operation still needs the required permissions and fields. For new ads, validate the exact proposal before approving creation. New advertising objects are created paused. Enabling them is a separate decision that can start spending.
Frequently asked questions
Is this checklist for Facebook or Instagram Ads?
It applies to Meta advertising workflows. State whether the audit includes both platforms or only particular placements, and request that breakdown when available.
Can the assistant verify my Meta Pixel is installed correctly?
A campaign report alone cannot prove that. A tracking implementation test needs separate evidence from the website and the relevant event diagnostics.
Why can I read reports but not update an ad?
Reading and writing can require different permissions, roles, assets, and input fields. Check the exact tool result and live permission readiness rather than repeating the write.
Sources and verification
Platform references are listed below. PlugMind tool descriptions were checked against its implementation. Example prompts are suggested workflows, not claims of results from a live advertising account.
