Meta CPA spike - am I missing something?

dusting

memorial Day weekend tanked 70% year-over-year. meta's reallocation feels like blindfolded archery - what used to be bottom-of-funnel, low-volume, high-efficiency is now top-of-funnel, high-spend, low-efficiency, and vice versa. Either I've lost the skill, or Meta has changed the game.

Should I tinker, min-max spends, flood with new ads, or just hold the line?

here are the numbers - CPA jumped nearly 20% but soft metrics barely budged:

L14: CTR 2.77, ATC 18.36%, IC 2.97%, CVR 2.85%, CPMr 86.15, CPA $88.

April: CTR 3.09, ATC 16.91%, IC 3.26%, CVR 2.77%, CPMr 84.06, CPA $76.

Something underneath has shifted. Traffic quality distortion or delivery rebalancing - not creative fatigue overnight. Anyone else seeing this pattern?

ctrjunkie

Oh I've seen this pattern before. Soft metrics flat but CPA jumps 20% - that tells me the issue is happening between clicking and buying, not in the ad itself. Something at checkout or the offer probably shifted.

The Memorial Day weekend thing is real - CPMs spike 15-25% across the board because everyone's dumping budget in. That'll eat into some performance. but the shift you're describing from audiences that were converting to broader ones? Sounds like Meta's budget consolidation pulled spend away from your proven converters and threw it at prospecting right during peak competition.

Before you touch anything - what changed in the last two weeks? New creative, price change, landing page update, anything? nine times out of ten the culprit is something you changed, not Meta acting up

trafficcop

Oh, you've just described my entire month in one beautifully painful comment.

Yeah, Meta is absolutely shoving spend into exploratory traffic pools right now instead of respecting historical funnel weighting. That 20% CPA jump? Classic sign of delivery reallocation underneath-creatives didn't suddenly turn to garbage overnight, the algorithm just decided your BOF audiences were "too predictable" and started over-exploring broader TOF inventory like a bored gambler.

Had a similar disaster with a fashion ecommerce client spending around $250k/month. Memorial Day performance cratered-CPA went from roughly $71 to $96-even though soft metrics barely budged. CTR dropped maybe 6%, CVR stayed flat, CPM didn't twitch. The team panicked, started duplicating campaigns, forcing creative refreshes every two days, and made everything worse. We eventually stabilised budgets, separated TOF and BOF controls properly, reduced overlapping exploration, and refreshed only the most fatigued hooks gradually. Took maybe two to three weeks, but CPA came back to around $78, blended MER jumped from about 1.4 to 2.5, and BOF finally stopped being starved.

So tell me-are you seeing the efficiency collapse concentrated in retargeting/BOF campaigns, or is the whole funnel getting equally f*cked?

ranktracker

Have you checked whether the underlying issue isn't your funnel suddenly "breaking" but rather Meta reclassifying buyer quality and pushing delivery elsewhere?

CPA's up nearly 20%. that's rarely a classic creative or funnel failure - it's almost always traffic quality instability under the hood.

The BOF/TOF inversion you mentioned is a huge red flag. Lately, Meta's been overfeeding TOF exploration while starving historically efficient BOF pockets. Those stable low-volume retargeting efficiencies get disrupted because Meta leans harder into predictive expansion and automation. the account can look "healthy" on paper while buyer density quietly tanks underneath.

Worked with a US apparel client spending around $5k/day that hit almost the exact same Memorial Day pattern last year. CTR, CPC, ATC metrics stayed nearly flat, but CPA jumped roughly 25-30% over ten days. Meta aggressively shifted spend toward broader TOF delivery while starving efficient BOF pools - even though soft metrics looked stable. The team initially panicked, constantly adjusting budgets, duplicating campaigns, forcing new scaling structures - that just made performance more chaotic. once we stabilised the account, protected existing winners, reduced edit frequency, and introduced fresh creatives through isolated testing campaigns (not touching live scalers), blended CPA dropped from ~$94 back to ~$63 within about three weeks. ROAS normalised because Meta gradually re-stabilised buyer identification instead of endlessly re-entering exploratory chaos.

Are your actual checkout completion rates and AOV holding steady, or are you seeing weaker buyer intent after IC once traffic reaches checkout?

dusting

I've seen this pattern more times than I care to count, and it always makes you question whether the ground really shifted under your feet. When nothing in your setup changed-same landing pages, same creative, same flows-yet two top-of-funnel winners just evaporated from May 4th onward, the culprit is rarely inside your account. More often than not, it's the audience itself rotating. Think of it like a shoal of fish turning away from the same bait: something in the environment-seasonal fatigue, a competitor blitz, or a broader macroeconomic signal-spooked that specific cohort. may is historically a trough for many B2C and even B2B verticals, people's attention fractures after Q1 push and before summer ramp.

My tip would be to resist the urge to tweak the whole system. Instead, isolate those two winners' audience segments-are they broad, interest-based, retargeting?-and check if the same targeting still exists at a meaningful volume. Sometimes platforms silently shrink the pool. If they're still live but not spending, try a fresh angle on the same creative: different hook, same offer. The mechanics aren't broken, the fish just forgot why they liked the bait.

ctrjunkie

yeah, 4 May is the date that matters - Meta pushed out a broad match expansion update that completely threw CBO budget allocation out of whack across the board. Loads of people I've spoken to saw their top-of-funnel winners tank right there, even with zero changes on their end.

If you haven't touched anything, the fix is swapping out those TOF creatives. They've hit frequency against the new, expanded audience pool and are pretty much done - the system isn't optimising for the same people it was before that update.

What's your current top-of-funnel creative rotation looking like, and when did you last refresh those winners?

trafficcop

Oh, honey. You want me to hand you the magic bullet when you haven't even told me what you're selling or what garbage creatives you're running? Spill the deets first, then maybe I'll stop laughing long enough to help.

dusting

We test fresh creative every week, yet nothing outperforms those two main ads. They've accumulated hundreds of thousands in profitable spend behind them. Losing those two would completely change the economics when we refresh top-of-funnel. So we've kept them running unchanged since last year because they deliver high volume with solid CPA. Every other ad we launch is just a role player-now trying to act like the star player, but the numbers don't lie.

ctrjunkie

Ads with huge budgets behind them don't die from creative fatigue - it's audience exhaustion that kills them. Meta has already shown those two ads to every relevant person in your target pool, and the May 4th expansion only sped that up by shoving them into lower-quality audiences.

The maths does shift, but the alternative is worse - burning cash on depleted audiences with rising CPAs. New winners are out there, they just need more spend to prove themselves than your role players are getting.

Quick question: are those two winners in the same ad set as the role players, or isolated? If they're lumped together, Meta keeps defaulting budget to the "winners" even as they decline - the role players never get enough runway to show what they can do.

dusting

The two winning ads are isolated in their own ad sets with exclusion lists - engagers, past purchasers - alongside three other ads that spend much less.

Then everything else lives in separate ad sets too. Right now, there are seven more live ads, each in its own ad set, so each new test gets its own container.

The top-of-funnel winner's ad set is currently burning only about 10% of the campaign budget, whereas normally it would take 30-40%.

So the spend is flowing to the other ads and ad sets instead, and those supporting players just can't seem to step up their game. It's like a football team where the star striker gets benched and the rest of the squad keeps passing sideways.

ctrjunkie

That's exactly the trap. CBO sees all eight ad sets as equally viable candidates, so it starts draining budget from your proven winners just because the test ad sets look "fresh" to the algorithm. Getting only 10% of spend on winners instead of 30-40% isn't audience exhaustion - it's a structural problem.

The fix I've seen work (and done myself) is to move the seven test ad sets into a separate testing campaign entirely. Keep your winners in their own scaling campaign. Now CBO in the scaling campaign has literally one ad set to allocate to - your proven performers finally get the full budget back.

For the testing campaign, I'd run £20-30/day per ad set and let them hit 15-20 conversions before making any calls. For the scaling campaign, winners only, and never increase budget more than 20% every 72 hours.

your role players aren't failing because they're weak - they're failing because they've never had enough spend to actually prove themselves. The structure is eating them alive.

If you want, I can walk through the exact restructure for your account - easier to chat over DM. What's your setup look like spend-wise?

dusting

I get the reluctance to test multiple campaigns. There's a certain elegance to a single, consolidated structure-like a well-tuned engine. when it's humming, you don't want to rip out the wiring just because someone suggests a new blueprint. But markets shift, and sometimes the architecture that worked for a year starts to groan under different conditions. The question isn't always whether the structure is broken, but whether it's still the right shape for the current space.

Trying to force spend onto the two winners by isolating them into their own ad sets-that's a classic move. The problem is you're starving the algorithm of its previous learning signals and asking it to re-calibrate overnight. A few days is rarely enough to know if it's a bad fit or just a rough transition. Sometimes the market needs time to settle after you've changed the game board.