I’ve been running 4 Meta Ads campaigns at around $500/day each. Let’s call them A, B, C, and D.
For a long time, all 4 campaigns were surprisingly stable — average CPA stayed around $40/web purchase, and performance between campaigns was almost identical.
Then things suddenly changed.
About a week ago, campaign C and D started performing badly. CPA increased hard, and some days they even had 0 purchases.
What confused me was that overall account performance was still relatively stable. A and B continued spending and converting normally, so total blended CPA didn’t look terrible.
At that point, I made a decision that I think many media buyers would make:
I turned off C and D.
At the same time, I duplicated the two “healthy” campaigns A and B into new campaigns A’ and B’, because during that period A and B were showing very strong numbers — around $25 CPA.
My thinking was simple:
* Keep the winners
* Kill the weak campaigns
* Duplicate what’s working
* Scale stability
Sounds logical, right?
But after doing that, the account performance completely collapsed.
For the past 3 consecutive days, blended CPA has jumped to nearly $100.
And this is the part that reminded me again how unstable — and sometimes deceptive — Meta Ads can be.
The mistake is that I judged campaigns individually instead of looking at the account system as a whole.
Looking back now, several things probably happened:
### 1. A and B may have been benefiting from account balance
Sometimes weak campaigns actually help stabilize delivery distribution across the account.
C and D looked bad individually, but they may still have contributed to audience exploration, auction diversity, or conversion learning signals.
Turning them off changed the account structure too aggressively.
### 2. Duplicating “winning” campaigns often resets hidden momentum
A campaign showing $25 CPA does not mean a cloned version will inherit the same efficiency.
Meta optimization is heavily tied to accumulated data, delivery history, engagement quality, conversion timing, and audience mapping.
A’ and B’ were technically “new campaigns,” even if settings were identical.
### 3. Meta performance is often cyclical, not linear
One of the biggest traps in Meta Ads is reacting too quickly to short-term volatility.
A few bad days from C and D may simply have been variance.
But by turning them off during a temporary fluctuation, I may have interrupted a system that was actually balanced overall.
### 4. Blended account CPA matters more than campaign-level emotion
This is probably the biggest lesson.
Even though C and D looked terrible emotionally (“0 purchases today”), the account itself was still stable.
The moment I started optimizing emotionally instead of systemically, performance deteriorated.
Now I’m debating whether to turn C and D back on.
Not because they were amazing campaigns — but because Meta sometimes rewards stability more than constant intervention.
The longer I run Meta Ads, the more I realize:
Meta is less about finding perfect campaigns…
and more about protecting stable systems from unnecessary changes.
Sometimes the hardest optimization is simply not panicking.
Generally speaking, running ads is somewhat similar to playing the stock market. :))