A meta ads outage what to do playbook: how to confirm an incident on the status page, when pausing makes it worse, and how to keep a broken day out of your test reads.
Thursday, 7:40 a.m. You open Ads Manager and the whole account looks shot: spend ticking up, conversions sitting at zero, CPA columns red from the first row to the last. Every campaign at once, same hour, same shape.
I have watched buyers gut an account in that moment. Pause everything, kill the worst three ad sets, duplicate "fresh" replacements. Two hours later Meta confirms a reporting incident, the purchases were never missing, only invisible, and the account spends the following week relearning what it already knew. The outage itself cost nothing. The reaction cost plenty.
Outages are a normal part of this job: Meta delivery incidents, Ads Manager reporting blackouts, pixel and Conversions API failures, and the quieter kind, your own shop going down while the ads keep buying clicks for a dead page. Here is the playbook I run, in three phases: detect, respond, clean up.
The takeaways
Look at the spread before you look at any single number. A real performance problem is local: one creative fatigues, one audience saturates, one campaign's CPM drifts up over days. An outage is synchronized: every campaign breaks in the same hour, with the same shape, often across every account you can see.
That spread check takes thirty seconds and settles most cases. Then confirm it officially. Meta Status (metastatus.com) is Meta's own incident page and covers Ads Manager and the Marketing API; Google publishes the Google Ads Status Dashboard for its side. If you work with other buyers, ask them: a platform incident shows up in everyone's accounts at once, and a colleague's "mine too" is faster than any dashboard.
One caveat from experience: status pages lag. I have seen symptoms run for hours before an incident was acknowledged. So treat the absence of a posted incident as weak evidence. If every campaign you run broke at 3 a.m. simultaneously, the infrastructure hypothesis stands until something disproves it.
Not every incident costs you money, and knowing which kind you are in decides the whole response.
In most cases, no. If the failure sits in measurement, in reporting or tracking, then delivery still works and customers still see working ads. Pausing converts a measurement problem into a delivery problem. The exception is a dead destination: when the landing page or checkout is down, pause at once, because every click from that moment on buys nothing.
The reason restraint pays is what a pause does to delivery. Meta's documentation lists the changes that put an ad set back into the learning phase: edits to targeting, creative, or optimization event, large budget changes, and pausing for seven days or longer. Panic responses tend to involve exactly those edits, new budgets, swapped creatives, duplicated ad sets. Even a short pause has a price: when the ad set re-enters the auction, the pacing system recalculates and delivery wobbles for hours.
My decision rule fits in one question: is the failure between the customer and the purchase, or between the purchase and my dashboard? Funnel broken: pause. Dashboard broken: sit on your hands and write the annotation instead.
The middle case is a multi-day tracking outage. The delivery system is optimizing blind, so I cap budgets rather than pause: keep the auction presence, limit the downside, restore full spend when events flow again.
Annotate the day the moment the incident is confirmed, in whatever system you trust yourself to read later, then exclude it from every analysis that touches it. Three weeks from now, comparing two creatives, you will not remember that one Wednesday in June was a pixel incident. The note has to outlive your memory.
The exclusion matters in three places. Creative comparisons: a zero-conversion tracking day punishes whichever ad happened to carry more spend that day, and the gap it creates looks exactly like a performance gap. Test reads: if a test window contains an outage day, read the test without it, or extend the window; I have made the case for spend and day floors before any verdict in when to kill a Facebook ad, and a confounded day inside the window breaks the same logic. Automated rules: a CPA or ROAS rule that evaluates an outage day sees infinite CPA and fires with full confidence. Suspend rules during a confirmed incident, or run rules that keep evidence: mine log every evaluation and keep kills reversible for 24 hours, which has let me un-kill an ad that died on a day the data lied.
The same discipline applies to fatigue calls. A CTR drop on an outage day is the market's fault, and any fatigue read needs a market-wide comparison before you blame the creative. That argument, with the refresh thresholds, is in the ad fatigue piece.
Outages will keep happening, and none of them will announce themselves. The playbook is dull on purpose: check the spread, check the status page, classify the failure, pause only for a dead destination, annotate the day, exclude it everywhere. The buyers who lose money in outages are mostly the ones who acted fast on bad data. If you want to see how I wired the alerting and the rule safeguards behind this, the automation page walks through the full setup.
This is the thinking behind Adscalr.
See the product →