@jordanpaid Exactly, the best way to learn is by doing: testing new products, launching new offers, working in the ad manager, and putting your testing ideas into practice.
I'm increasingly separating creative discovery from creative scaling.
For a long time, I expected the same campaign to do everything:
Find new creatives.
Validate them.
Allocate spend.
Find winners.
Scale those winners.
It works, but when creative volume gets high, distribution becomes a problem.
So I'm testing a different approach:
ABO = controlled creative discovery
BidCap/CostCap = scaling validated creatives
I use the testing environment to make sure creatives actually get a chance to collect data.
Then the creatives that prove themselves earn their way into the scaling campaign.
Instead of asking one campaign to solve every problem, I'm giving each campaign a specific job.
So far, I like the amount of control this gives me over the testing process.
One thing I've been thinking about a lot when testing creatives at high volume:
"It didn't spend" doesn't automatically mean "it's a bad creative."
When you have dozens of creatives competing inside the same CBO/BidCap campaign, Meta has to make distribution decisions very early.
Some creatives get spend. Others barely get touched.
The problem is that we often look at the second group and immediately call them losers.
But how can I confidently say a creative doesn't work if it never received enough distribution to actually prove anything?
There's a big difference between:
Bad performance
and
Not enough data to evaluate performance.
That's one of the reasons I'm starting to retest creatives that received little or inconsistent spend instead of automatically killing them.
No spend isn't necessarily a failed creative.
Sometimes it's simply an untested one.
People obsess over finding new products.
I think building a creative testing machine is just as important.
A product that looks "saturated" with one angle can find an entirely different audience with another.
New hooks.
New angles.
New creatives.
New pockets of demand.
Before killing a product, ask yourself:
Did the product actually fail, or did you just stop testing creatives too early?
From what I've seen, every batch of 10-20 new creatives can produce 2-3 that have real scaling potential.
Not every batch will hit those numbers.
But that's exactly the point:
You don't control which creative wins.
You control how many good attempts you give the market.
@benradack I agree with you. I always look for a decent number of add to carts early on.
You don't need a ton of them, but I want to see them coming in at a low cost over the first 4-5 days.
Usually, by day 2, you should already start seeing the first purchases come through.
And no, this isn't another "Andromeda changed everything" take.
It's much simpler than that.
More creatives = more chances to discover new hooks, angles and ways to sell the exact same product.
You might have a winning product with a losing creative.
And kill it before you ever find out.
One of the biggest mistakes I see in product testing:
Testing 5 creatives and deciding whether the product is a winner or a loser.
In my experience, that's nowhere near enough.
If I'm testing a new product, I want at least 10 new creatives going live every single day.
Volume matters.
But there was a catch.
The longer I let the campaign explore, the more my ROAS declined.
It made me question something:
Is there a point where too much exploration becomes less profitable than aggressively pruning creatives?
Curious if anyone else running large BidCap campaigns has seen the same pattern.
The exploration phase lasted much longer than I expected.
Even after weeks, Meta kept finding new winning creatives.
The less I interfered, the more the system kept exploring on its own.
I eventually had 380+ active creatives.
Here’s what surprised me:
I barely had to pause underperforming ad sets.
They simply stopped spending, while the budget naturally shifted to creatives that hadn’t been explored yet.
I stopped killing creatives manually.
I wanted to see how far Meta’s BidCap algorithm could optimize on its own.
It ended up being one of the most valuable experiments I’ve run.