Facebook ads performance dropped suddenly?

Tuesday: 41 purchases at a 19 € CPA, in line with the last three weeks. Wednesday: 12 purchases at 64 €. Thursday opens even uglier. Same ads, same funnel, same offer. You changed nothing, and the account is burning budget like the pixel forgot what a customer looks like.

Every media buyer gets this week eventually. I have had it more times than I would like, on my own accounts and on client accounts, and the pattern is always the same: the drop itself is rarely the expensive part. The expensive part is the panic response in the first 48 hours, because editing everything at once usually converts a two-day dip into a two-week slump.

So here is the fixed order I check things in now. Four layers: platform, account, creative, noise. Work top to bottom and you usually find the cause before lunch.

Short answer: When Facebook ads performance drops suddenly, the cause sits in one of four layers, checked in fixed order: platform (a Meta incident or auction spike), account (a learning reset or broken tracking), creative (a delivery shift, rarely fatigue), or plain variance. Read delivery metrics before conversions, and resist panic edits that extend the slump.

The takeaways

• Read delivery metrics before conversion metrics. If impressions, CPM and CTR look normal while purchases fell off a cliff across every campaign at once, the problem is tracking or post-click, and your ads are innocent.
• A bad day is often just arithmetic. A campaign averaging 10 conversions a day will hand you a 5-conversion day about twice a month from plain Poisson variance, with nothing wrong at all.
• Panic edits extend the slump. Meta's Business Help Center puts learning-phase exit at roughly 50 optimization events within 7 days of the last significant edit, and every significant edit restarts that clock.

Is it Meta, or is it you?

Start one level above your account. Check the Meta Status page (metastatus.com) for an active incident in Ads Manager or ads delivery, and ask one or two buyers in your network whether their accounts dipped at the same hour. A drop you share with strangers in other verticals is a platform problem, and no edit inside your account will fix it.

Then split delivery from conversion. Open Ads Manager and look only at impressions, CPM, CTR and outbound clicks (the Meta ads metrics worth reading first, in roughly that order). If those four look like last week while purchases collapsed everywhere simultaneously, the auction is fine and your ads are fine; something between the click and the reported conversion broke, which is the next layer down.

One more platform-shaped cause: auction pressure. If CPM jumped 30 to 50 percent overnight, you may be in a seasonal bidding war (Black Friday ramp-up, Prime Day spillover, election weeks). Your results drop because every impression got more expensive, and the cure is patience or a deliberate bid decision; a new creative cannot lower the auction price.

Did something inside the account reset the campaign?

This layer catches most real cases. Three checks, fifteen minutes.

First, the change history. Open the activity log in Ads Manager and look at the 72 hours before the drop. Meta's Business Help Center lists what counts as a significant edit: changes to targeting, creative, or optimization event, a substantial budget or bid change, or a 7-day pause. Any of these puts the ad set back into learning, and re-learning delivery is routinely worse for days. The edit is often not yours: a colleague, an agency login, or an automated rule fires quietly and the campaign relearns from scratch.

Second, tracking. Fire a test event in Events Manager and watch it arrive. A Shopify theme update, a checkout change, a consent-banner tweak, or a broken pixel-to-Conversions-API deduplication will zero out reported purchases while real orders keep landing in your shop backend. If the backend shows normal revenue and Ads Manager shows none, you have a measurement outage, and it needs fixing today.

Third, the boring stuff: billing failures, a hit spending limit, an ad account flagged for review. Two minutes to rule out, embarrassing to find on day four.

Is the creative actually the problem?

For a sudden drop, the creative is the least likely culprit, and that surprises people. Creative fatigue is real, but it is a slow bleed: frequency creeps up and CTR sags over two or three weeks. I wrote about how that decay looks and when to refresh separately; a 24-hour cliff does not match its shape.

What a sudden creative-shaped drop usually means is a delivery shift, a change in which ads get the money. Check spend distribution inside the campaign. In a CBO campaign, a new ad entering the mix can pull 60 percent of budget overnight, and if it is weaker than the ad it displaced, the campaign average tanks while every individual ad performs exactly as before. That same concentration is what makes Meta starve all but one ad in a test, read in reverse. Also check for a disapproval-and-re-review on your top spender, and skim the comment section of the biggest ad: a thread of angry comments above the CTA can cut conversion on the same creative within a day.

One mechanism I now consider non-negotiable, and which we built into Adscalr's fatigue detection: compare the ad's CTR slope against the whole market's movement for the same days. A bad market day looks identical to a dying creative if you only stare at your own dashboard.

Or did nothing happen at all?

Here is the diagnosis nobody wants: on small daily samples, a scary drop is often plain variance. If your campaign averages 10 conversions a day, Poisson arithmetic gives a standard swing of about 3, so a 5-conversion day arrives roughly twice a month with no cause whatsoever. Your CPA doubles on paper that day. Two such days back to back feel like a trend, and statistically they are still nothing.

The fix is to change the comparison window. Never judge yesterday against the day before; judge the trailing 7 days against the 7 before that. If the weekly view is flat, you watched noise, and you nearly paid for it with a learning reset. The same small-sample trap crowns fake winners on the way up, which I covered in won, or just lucky?; the drop version is the same coin, landing tails.

What not to do in the first 48 hours

The first 48 hours are for diagnosis, with one exception: broken tracking gets fixed immediately. Everything else can wait.

• Do not edit targeting, creative, or the optimization event. Each one is a significant edit per Meta's documentation, and each restarts the learning phase you are trying to protect.
• Do not duplicate the campaign for a "fresh start". The duplicate begins learning from zero and usually spends a week being worse than your wounded original.
• Do not slash the budget in one move. I keep budget changes near 20 percent a day, because in my experience larger cuts behave like a restart even when the interface does not warn you.
• Do not rewrite the creative on day two. If fatigue were the cause, you would have seen it coming for weeks.

Sitting still while CPA runs hot feels wrong. It is still cheaper than a self-inflicted reset on a campaign that was about to recover on its own.

The 45-minute checklist, in order

1. Meta Status page plus a peer check in another vertical (5 minutes).
2. Delivery metrics vs conversion metrics, account-wide (10 minutes).
3. Test event in Events Manager, compare shop backend revenue to reported purchases (10 minutes).
4. Change history: significant edits, automated rules, colleague logins in the last 72 hours (10 minutes).
5. Spend distribution across ads, disapprovals, comment sections on the top spender (5 minutes).
6. Trailing 7 days vs the previous 7. If flat, close the laptop (5 minutes).

If all six come back clean, you are looking at variance, and the correct move is the hardest one: none.

Most of this checklist is the same read I ended up automating, because doing it by hand at 7 a.m. with a nervous client on the phone is miserable. Scoring ads on a blend of six metrics, pulling wild daily numbers toward what the format normally does, and checking the market moved before blaming your creative is what the ad-intelligence side of Adscalr does on every account, every day. But the checklist works fine on paper too. The point is the order: platform, account, creative, noise. Panic has no slot in it.