Skip to main content
A recommendation is a prompt to investigate. Its usefulness depends on the data behind it, the result you optimize for, and the change it proposes.

​
Review a finding

  1. Open Intelligence for the selected Meta account.
  2. Read the finding and inspect the entity, period, severity, and supporting evidence shown with it.
  3. Compare the recommendation with your account’s actual objective and target. Verify what Results represents for the affected campaign or ad set.
  4. Open the related campaign or ad to check current configuration, creative, and delivery. A finding based on an older period may not reflect a change made since then.
  5. Choose the supported action when appropriate and review its preview or confirmation. You can also keep watching when the evidence is insufficient.

​
Separate evidence from an explanation

A drop in results, rising cost, or repeated creative exposure is an observation. It may have several causes. Check delivery changes, budget changes, tracking, reporting completeness, and the age of the data before treating a single explanation as proven. A low-volume row can have an unstable cost-per-result value. A recent change can also make a before/after comparison hard to interpret. Look for a sufficient and comparable period rather than waiting for an arbitrary number of days to guarantee a feature will work.

​
Explore further

Supported Intelligence views include Diagnose, Trends, History, Explore, Creatives, Creative analytics, and Learning. Some are reached through the relevant task or deep link. In History, distinguish the Activity record from a Snapshots comparison. Before/after performance can show what followed a decision. It does not by itself prove that the decision caused the change or establish a guaranteed saving.

​
If nothing useful appears

Check account selection, targets, when data was last fetched, data coverage, and whether the selected entities had meaningful activity. An empty result can mean insufficient evidence; it is different from an error. Resolve a connection or refresh error before acting on retained data. Next: Understand metrics, check freshness, or set approval behavior.