Winners falling off a cliff at week 1-2 on Meta?

ranktracker

Honestly feels like something's broken with creative lifespan on Meta the last couple months. i've had tested winners that used to run 4-5 weeks now dying around day 10-14. CTR drops from 1.8% to 0.7% basically overnight, and frequency isn't even that high yet. the last 2 months have been proper weird.

context: B2C SaaS, _8k/month spend, no in-house designer.

a few questions:

Is this just normal post-iOS audience saturation getting worse, or are the auctions actually different in 2026? feels like the algorithm is way more aggressive about pushing fatigue scores than it was last year.

for people producing more volume to compensate - are you doing real net-new concepts every week or mostly just variations on the proven angle (different hook text, swap background colour, change format)?

and is anyone actually trusting AI-generated statics for cold audiences yet, or is it still strictly for retargeting/mid-funnel testing?

Not looking for tool recs necessarily, more curious about what the cadence actually looks like for people running similar spend. CPA is holding at $46 for now but this fatigue thing has been a pain in the arse.

bouncekiller

Oh honey, yes. Scour through the threads here-it's been the same sob story for weeks. It's not your creative, not your product, the bloody platform's having a meltdown. Anyone telling you Meta's stable right now is either lying or trying to sell you a course.

ranktracker

Spot on. Andromeda's architecture is built to maximise spend, not performance. You've got to kill anything live past 72 hours and set hard audience exclusions from the start. Otherwise the algo just funnels budget into retargeting loops.

trafficcop

Oh absolutely, it's almost like firing half the people who actually understood the platform was a genius move for performance. But sure, blame the creative, blame the algorithm - it's definitely not the revolving door of talent.

datanerd

seen the same pattern across multiple accounts I've been auditing. What's shifted in the last couple of months is that the algorithm is punishing weak post‑click engagement signal much faster now. winners that used to run for four weeks are getting clipped at day 10-14 because the landing page session quality starts dragging down the ad set's overall score. The CTR cliff people are seeing is a symptom, not the root cause - the real death is session depth falling off.

$46 CPA at $8k/month with no designer is solid work, honestly. Don't overrotate on the "fatigue" excuse, that's usually a red herring when the actual issue is post‑click alignment.

cyberpunk

It's not just you. i've been tracking this across several accounts i manage, and that 10-14 day cliff has become the baseline. My take is it's less traditional frequency fatigue and more Meta's auction getting greedy - it finds your best audience segment, extracts the value fast, then you're paying premium for the leftovers. Used to take four or five weeks to hit that wall. now it's almost instant.

On your second point: variations might buy you three or four extra days if you're lucky. New hook, different background, reformat for Reels - it helps a bit but doesn't reset anything. A genuinely new concept with a different emotional angle is the only thing I've seen extend lifespan. that needs real creative thinking, not just asset churn. If you don't have in-house design on a decent retainer, that squeeze hurts.

AI-generated stills for cold audiences - i wouldn't write them off entirely, but the ceiling is low. they work for direct response where "does this communicate the offer" matters more than "does this make someone feel something." If your CPA is sitting around the mid-forties, I'd be careful experimenting too aggressively on cold. you risk eroding what's holding.

The cadence question is the real headache. Most people i know at similar spend are producing two to three net-new concepts a month just to tread water. That's a lot when you're outsourcing everything.

backlinker

More ads equals more instability - that's what everyone misses. Throwing hundreds of creatives at a pattern-recognition machine confuses the bid system, not impresses it. Expecting stable results from constant churn is like expecting a local GBP listing to rank after you've changed the NAP every week. The industry calls it 'ad fatigue', but really it's creative noise syndrome.

communityfirst

Yeah, this is exactly what I'm seeing across several accounts too - the fatigue curve has genuinely compressed. Used to be you could ride a winner for three or four weeks before touching it. Now it's more like ten days and the algorithm starts cannibalising its own delivery before frequency even gets spicy. My take is Meta is honed in on early CTR signals far more aggressively and basically soft-retiring creatives faster than before.

On the volume question - pure variations like hook swaps or colour changes are buying you maybe three or four extra days at best. The real lift comes from net-new angles, different emotional entry points, a different "enemy" in the copy itself. For AI statics, they're holding up better than I expected on cold traffic if the concept is tight, but they still fall off faster than custom photography in my experience. So i treat them as high-volume test fuel rather than long-term runners. It's a team-wide shift we're all having to adapt to.

datanerd

Two-week creative cliff on B2C SaaS accounts at $5k+ spend is real. I'm seeing 10-14 day CTR walls across the board, frequency under 8 when it starts. You're not alone.

This is really three separate things, and most people diagnose only one.

First is audience saturation. iOS privacy means the addressable cold pool shrinks every quarter. That part is just true and sucks.

Second is policy fatigue scoring. Meta's audit algorithm got aggressive about flagging health-adjacent language on B2B/B2C copy. You don't get rejected, you just get auction-priority debt. CPM climbs, CTR tanks, feels like audience fatigue when it's actually policy scoring.

Third is format fatigue, not audience fatigue. A winning static runs 10-14 days before the feed saturates to it. That's always been the wall. But now you can't reset it with a bg color swap or copy tweak. The angle is burned, not just the visual. So pure-format variation on proven hooks hits different than launching new concepts.

Quick diagnostic: pull your pixel data. If CTR drops but EMQ stays solid, it's policy signal. If EMQ tanks too, it's audience saturation. Different fixes, completely different spend strategy.

One move that works: test pure-format variation before spinning up new concepts. Same angle, different layout, different thumbnail treatment, different copy length. Lots of people skip that and go straight to cold testing on new hooks when the current hook still has runway.

datanerd

Seen the exact same pattern across my ABM plays. We're now building 10-12 day creative refreshes into the media plan - anything longer and the algorithm just stops spending. And it's not just hooks, we're swapping out core angles entirely each cycle. The old 4-5 week lifespan is dead.

backlinker

i've been seeing this pattern too, and honestly the advice to just chuck more ads into Andromeda is nonsense. Every time a brand rep or agency partner tells you "you need more ad units to feed the algorithm," check their revenue model. They're billing by the creative, not by the result.

The real issue isn't volume-it's that Google's AI is collapsing the learning phase for some verticals. If your tested winners are tanking in week 1-2, it's usually because the budget threshold crossed a point where the system treats the campaign as 'learning limited' again. or the audience overlap from other campaigns in the same account is cannibalising conversions. throwing more creatives at that problem just amplifies the noise.

I've seen better stability by consolidating into fewer, higher-signal ad groups and using asset-level reporting to kill underperforming headlines before they drag the whole group down. but that doesn't sell creative packages, does it