Meta burning 30% of budget at 5am for nothing

24fire

Meta's algorithm has decided that 5am is the perfect time to torch my budget. Zero traffic on the site. Eleven link clicks at roughly double my average CPC. CPMs conveniently spiked too. The machine clearly thinks sleepless scrollers at 3am are my ideal customers.

What's the actual logic here? Why dump spend into dead hours when there's no signal to learn from?

A colleague suggested switching to bid caps and time scheduling instead of giving the system free rein. Probably the right call at this point. The algo optimization feels less like optimization and more like a tax on inattention.

Wonder if this ties back to the massive AI infrastructure spend they're trying to recoup. Someone has to pay for all that compute. Might as well be us.

fieldevent

Someone's gotta pay for that massive AI infrastructure bet - hundreds of billions worth - and it sure as hell isn't coming from genuine innovation. Platform hasn't shipped a truly new concept in years, just copies and pivots. Meanwhile we're out here watching CBO torch 30% of daily budget before breakfast because the algorithm "needs to learn."

The disconnect is wild. We optimize every dollar on the floor - badge scans, meeting sets, pipeline attribution - and the platform just lights money on fire to fund a moonshot that doesn't move the needle for advertisers.

copycat

That early-morning spend spike is classic CBO behavior - the algorithm sees cheap inventory overnight and gobbles it up before your actual audience wakes up.

Swapping to bid caps puts a ceiling on what you'll pay per result, and dayparting (or ad scheduling) tells the system "only show up when our people are actually online." Together they hand the steering wheel back to you instead of letting the algorithm joyride at 3 a.m.

Worth testing a 6 a.m.-10 p.m. window first, then tightening from there once you see where conversions actually cluster.

justauser

We've hit the exact same pattern. The clock rolls over and CPM more than doubles, dragging every other metric up with it. Then CPM cools off as the day goes on, but by then the CBO budget is already torched on the priciest traffic. Honestly feels less like a bug and more like it's designed to drain the account 😅

24fire

Nothing like a nice CPM spike to compound the pain. The algorithm has apparently decided that someone scrolling at 3am with diminished judgment is my ideal customer. Meta's machine learning is a strange god - it worships at the altar of engagement, and at that hour, engagement and intoxication move together. You feed it budget, it finds the cheapest attention it can, and sometimes that attention is three drinks deep. The system isn't broken. It's working exactly as designed. The question is whether the design serves your business or just its own optimization loop.

conversionninja

Same disaster over here, with a CPM pushing 50 euros no less. Watching that number climb while the sky is still dark is enough to make anyone blow a fuse 😭

buggedout

CBO just chases whatever looks cheap in the moment. Overnight hours show low CPMs because fewer advertisers are bidding, not because those impressions are worth anything to you. It's optimizing for cost efficiency it can measure, not buyer quality it can't.

The good news is you don't have to ditch CBO to fix this. Ad scheduling still works with CBO as long as you're on a bid cap, cost cap, or ROAS goal instead of pure lowest cost - which it sounds like you already are.

What's worked for me:

Set the campaign to pause during your dead window (midnight to whenever real buyers actually show up)
CBO still shifts budget freely across ad sets during the remaining hours
You just stop feeding it cheap-but-useless inventory

Has anyone tested whether pausing overnight hurts the algorithm's learning phase? That's the one thing that always makes me second-guess this approach.

24fire

Bid caps are the discipline, but CBO is the reconnaissance mission. Letting it run wide to find where the buyers actually live makes sense. Hard to gather any meaningful signal on buyer pockets, though, if the budget burns through before lunch. the algorithm needs room to breathe, not just a spark and a fizzle.

roichaser

Usually see the bulk of spend happening in that post-3 pm window too. around here, we've found our best CPA comes in the late afternoon to evening block, so that pacing pattern makes sense. Sounds like the algorithm is front-loading spend to hit daily budgets, but the real conversions are happening later. Might be worth checking your hourly performance data to confirm.