meta ads collapsed after editing ad set - what happened?

prpro

I'm having a nightmare with my new supplement brand's ad account and could really use some perspective.

Launched on 25th April, first week was a disaster. Realised the funnel and offer needed work, so we optimised the website speed, swapped in better images, and refined the flow. Relaunched with a small budget of $50-100 a day, just one ad set running. Within a couple of days: 35+ add-to-carts, 15 purchases. One creative was clearly carrying everything.

Then I made a classic mistake. Killed the four underperformers in that ad set and added three new ads. Almost instantly, the winning creative stopped delivering - zero ATCs, the whole ad set went quiet. Gave it a week, then turned that ad set off and launched a new one in the same campaign with just the winning ad plus two new UGCs. Spent $100, got nothing.

Now I'm burning budget with close to zero results and no idea what to do next.

I had two campaigns: a prospecting CBO (new ad sets for each concept, 3-4 variations each - in hindsight, killing and adding creatives to an existing set was a terrible idea) and a retargeting CBO (visitors, ATC/checkout) - but that's irrelevant for a new store with tiny audiences. My plan was to find a clear winner, then create a separate scaling CBO for it.

Honestly, the timing is the biggest clue. On a fresh account with limited conversion history, those structural changes probably reset Meta's delivery pattern. Think of an ad set as a delicate ecosystem - once you disturb it, the algorithm loses the signal it was learning from. Relaunching the same winning ad in a new ad set doesn't guarantee it'll find the same buyers, especially when the account doesn't have enough historical data for stable optimisation.

Has anyone else seen this happen after editing a winning ad set? I'd love to hear what actually worked to recover it.

trafficcop

Oh, the classic "let's fix what ain't broken" move. nothing yanks the rug out from under a winning ad set like jamming new creatives into it and yanking others out. You had 35+ ATCs and 15 purchases running, then you touched the structure and Meta basically said "cool, I'm starting over."

The timing is the dead giveaway. On a fresh supplement account with minimal conversion history, even a small edit resets the delivery algorithm's internal magic. That winning ad didn't "die" - Meta just lost the pattern it was using to find buyers. Relaunching it alone won't bring back the same delivery path because the account's learning data has been scrambled.

worked with a supplement brand spending about $80-120/day. they had a UGC winner pulling ₂₈ purchases, 42 ATCs, and a CPA around $31 over ten days. owner did almost exactly what you did: paused weak creatives, added new ones into the same ad set, watched everything implode. ATCs went from 5-7 per day to zero, CPA shot past $90. I convinced them to stop rebuilding and launch a cleaner structure with just a few proven ads. within 12 days, we were back to 4-6 purchases daily, CPA settled near $34, ATC rate over 9%, and revenue jumped 140%+ compared to the "fix everything" chaos. Meta needs stable conditions to optimise.

So - what was the actual CPA and ROAS of that winning ad over at least 7 days of spend? Because if it was just a short-term spike, you might be chasing a ghost.

localpack

Mate, I've seen this a thousand times. Small accounts like yours? Every single edit shakes the tree. The learning phase resets, the algo panics, and you end up spinning your wheels.

My advice? Pull it back to an ABO structure. Take your winning ads, dump them into one ad set, and let it breathe. No touching, no optimising, no "just one more test" - you simply don't have the budget to fragment.

Here's the thing: at your spend level, CBO is a liability. Meta's algorithm gets too much freedom and starts throwing money at random audiences instead of what actually converts. You need to keep that control tight.

Honestly? If something is working, leave it alone. Creative fatigue is a non-issue at your reach numbers. The real killer is unnecessary edits.

datanerd

launch a fresh campaign with only your original winning ad and let it run untouched for at least seven days. every time you tweak the creative, audience, or placement, you're forcing the algorithm back into the learning phase - that means wasted spend and zero statistical significance. I've seen this kill performance repeatedly in B2B ABM plays where the decision cycle is long and the learning window really matters.

here's the technical bit: Meta's delivery system needs a minimum of ₅₀ optimisation events per ad set within a 7-day window to exit learning limited. If you change anything after it's started gathering data, it resets the clock. you're effectively burning budget on exploration instead of exploiting the winning signal.

what I do for enterprise accounts is clone the original campaign structure, drop in just that one winning ad, and then schedule a weekly pause/duplicate via a Python script to avoid manual interference. something like this:

import facebook_business
from facebook_business.adobjects.campaign import Campaign
# .... automate duplicate campaign creation with identical targeting + creative

Keeps the learning intact while letting me rotate in fresh tests on a separate campaign. if you're not doing that, you're just paying Meta to learn the same lesson over and over

analyticsjunkie

Oh, honey. You poked the algorithm and it bit back. Classic case of "if it ain't broke, don't fix it" - Meta's like a cat, fine until you try to move its bowl. When those 15 purchases were rolling in, was it one hero creative doing the heavy lifting or were several ads splitting the glory? Because if it was a single winner and you started swapping things around, you basically unplugged the vending machine while it was dispensing snacks.

prpro

I've seen similar figures - roughly $22-$25 per conversion with a 1.6 ROAS based on Meta's own attribution. That typically signals break-even or a slight profit, depending on your product margins and whether you're factoring in any other costs.

trafficcop

Oh, the classic reset-and-pray strategy - works like a charm until it doesn't. We've had the same dance with a supplement client. their campaign got stuck in a loop, and a full reset sorted it. magic.

Now, what's your product? And those creatives - are they sexy or snoozefest? give us the deets, and I'll spill what's actually moving the needle for my lot. No promises, but at least we can laugh together at how broken Meta is

prpro

when you say resetting the campaign, do you mean turning it off and on again, or starting a new one from scratch? they're quite different approaches

trafficcop

Oh babe, we need way more context than that. You're basically asking me to fix your car without telling me if it's the engine or the radio. Spill the details.