how a call runs
what it reaches
Meta Marketing API
server to server · straight to the platform’s API
access granted once · OAuth · scoped to this account
no browser session · no password sharing · no scraping
what you get
read-only by default · Every question is answered without touching the account. Pausing an ad set or moving a budget is a separate, deliberate step with its own guards, its own record and its own way back.
two fixed reads for unattended work · A scheduled run can only do the two reads it was built for, with nothing to widen. There is no path from a scheduled report to an unplanned change.
economical by construction · Every read asks for the minimum the answer needs. Nothing is pulled that the question did not ask for, so one curious question never slows everything else running on the account.
| tool | type of access | what it does | why it exists |
|---|---|---|---|
| campaigns.list | reads only | lists campaigns and ad sets with status and budgets | the account at a glance |
| insights.pull | reads only | pulls spend and results by day or ad set | the performance record |
| creative.performance | reads only | reads results per creative | finds the ad that carries the campaign |
| audiences.list | reads only | lists every audience the account can target | reach before spend |
| adset.pause | changes or removes | stops delivery once you approve | stops spend when a problem is found, with your approval on record |
| budget.set | changes or removes | moves a budget once you approve | moves money deliberately, with the before and after on record |
covers
91 operations across 22 families of the Marketing API: accounts, campaigns, ad sets, ads, creatives, insights, audiences, catalogs and product sets, pixels and conversions, targeting, budget schedules and Instagram. No percentage is given, because Meta publishes no closed list of methods to count against. Not covered: the Ads Library archive, pixel deletion and resumable video upload.
how it operates · alone, or with the rest of the fleet
alone
Meta-side spend · results · change history in one readout
01
you ask · “Is the retargeting ad set still earning its keep?”
02
it picks · the campaign-results read and the audience-status read · reads only
03
comes back · spend, results, and the health of the audience behind the set
04
lands · a plain answer in the session · the account exactly as it was
with Google Ads
Both paid platforms are read into one shared vocabulary. Spend, results and change history sit side by side, so a week can be compared across the two platforms without translating each console's own terms.
01
you ask · “How did the two paid platforms do this week, side by side?”
02
it picks · the campaign-results read on Meta and the campaign-results read on Google Ads, one date window · reads only
03
comes back · spend, results and the week’s change history from both, in one vocabulary
04
lands · one table for the week · each platform’s own terms translated once, not in your head
with GA4
Meta reports what it believes it drove. GA4 records what actually landed on your site. Read together, a Meta claim is set beside the visits and purchases it should have produced, and the two are allowed to disagree.
01
you ask · “Meta says it drove these purchases. Did they land?”
02
it picks · the campaign-results read on Meta and the session and purchase reads in GA4, matching dates · reads only
03
comes back · what Meta reports beside what the site recorded after the click
04
lands · a gap, if there is one, stated as a gap · Meta’s credit and the site’s record are allowed to disagree