- Jul 3, 2026
- 85
- 20
One thing I've changed in my Meta workflow this year is how I think about creative testing.
The old approach was pretty simple:
Find a winner → duplicate it → change the headline → change the background → make another version → keep scaling.
I'm becoming less convinced that this is the right way to operate now.
A lot of the discussion around Meta's Andromeda delivery system in 2026 is pointing in the same direction: creative diversity matters more than creative quantity. Meta's retrieval system is increasingly using the content of the ad itself as a signal for matching ads with potential users. Recent advertiser discussions are reporting the same thing: broad campaigns seem especially dependent on creatives that clearly communicate who the product is for.
So instead of producing 10 variations of one winning ad, here's the framework I've started using.
1. Test concepts first, variations second
Imagine I'm selling a productivity product.
Instead of:
Creative A: Product image + "Save Time"
Creative B: Same image + "Work Faster"
Creative C: Same image + different background
I'd rather test:
Concept A — Problem: "Still spending 3 hours doing this manually?"
Concept B — Outcome: "How we reduced this task from 3 hours to 20 minutes."
Concept C — Demonstration: Show the product actually solving the problem.
Concept D — Social proof: Customer explaining why they switched.
Concept E — Comparison: Old workflow vs new workflow.
Five ads, but five completely different reasons for someone to pay attention.
That's very different from five cosmetic variations.
2. Let each creative identify its audience
This is probably the biggest change in my thinking.
If I'm targeting ecommerce store owners, I don't want the algorithm to figure that out only from my audience settings.
The creative itself should make it obvious.
Instead of:
"Want to increase your revenue?"
I'd rather say:
"If your Shopify store is doing $20k–$100k/month and your CPA keeps increasing..."
Now the creative contains useful context about the intended buyer.
This matters especially when running broader targeting, because community reports suggest the creative itself is carrying more of the audience signal than before.
3. Don't kill the winner—expand the idea
When a creative works, I don't immediately create five clones.
First I ask:
Why did it work?
Was it:
The hook?
The problem?
The persona?
The offer?
The demonstration?
The format?
Then I keep the underlying insight and rebuild it in another format.
For example:
Winning UGC video:
"3 mistakes ecommerce stores make with X."
Next tests:
Static checklist → same problem.
Founder video → same problem, different voice.
Product demo → shows the solution.
Customer story → proof that the solution works.
Same successful idea, but genuinely different creatives.
4. Track concepts, not just individual ads
My creative sheet now needs another column:
CONCEPT
Because CTR alone doesn't tell me enough.
If five ads built around "saving time" consistently outperform five ads built around "saving money," that's much more useful than simply knowing that Ad #17 had the best CTR.
Now I've learned something about the market.
That insight can influence:
Landing pages.
Email copy.
Future creatives.
Even the offer itself.
The takeaway
I think we're moving from:
Media buying → finding the perfect targeting/settings
toward:
Media buying → feeding the delivery system better signals through genuinely different creative concepts.
Recent advertiser discussions around Andromeda are heavily focused on exactly this tension: how to provide enough creative diversity without simply flooding an ad set with meaningless variations.
For me, the important distinction is:
10 creatives ≠ 10 ideas.
I'd rather have 5 genuinely different concepts than 20 versions of the same winning ad.
For anyone spending seriously on Meta right now: are you seeing better results from increasing creative volume, or from increasing the number of genuinely different concepts?
The old approach was pretty simple:
Find a winner → duplicate it → change the headline → change the background → make another version → keep scaling.
I'm becoming less convinced that this is the right way to operate now.
A lot of the discussion around Meta's Andromeda delivery system in 2026 is pointing in the same direction: creative diversity matters more than creative quantity. Meta's retrieval system is increasingly using the content of the ad itself as a signal for matching ads with potential users. Recent advertiser discussions are reporting the same thing: broad campaigns seem especially dependent on creatives that clearly communicate who the product is for.
So instead of producing 10 variations of one winning ad, here's the framework I've started using.
1. Test concepts first, variations second
Imagine I'm selling a productivity product.
Instead of:
Creative A: Product image + "Save Time"
Creative B: Same image + "Work Faster"
Creative C: Same image + different background
I'd rather test:
Concept A — Problem: "Still spending 3 hours doing this manually?"
Concept B — Outcome: "How we reduced this task from 3 hours to 20 minutes."
Concept C — Demonstration: Show the product actually solving the problem.
Concept D — Social proof: Customer explaining why they switched.
Concept E — Comparison: Old workflow vs new workflow.
Five ads, but five completely different reasons for someone to pay attention.
That's very different from five cosmetic variations.
2. Let each creative identify its audience
This is probably the biggest change in my thinking.
If I'm targeting ecommerce store owners, I don't want the algorithm to figure that out only from my audience settings.
The creative itself should make it obvious.
Instead of:
"Want to increase your revenue?"
I'd rather say:
"If your Shopify store is doing $20k–$100k/month and your CPA keeps increasing..."
Now the creative contains useful context about the intended buyer.
This matters especially when running broader targeting, because community reports suggest the creative itself is carrying more of the audience signal than before.
3. Don't kill the winner—expand the idea
When a creative works, I don't immediately create five clones.
First I ask:
Why did it work?
Was it:
The hook?
The problem?
The persona?
The offer?
The demonstration?
The format?
Then I keep the underlying insight and rebuild it in another format.
For example:
Winning UGC video:
"3 mistakes ecommerce stores make with X."
Next tests:
Static checklist → same problem.
Founder video → same problem, different voice.
Product demo → shows the solution.
Customer story → proof that the solution works.
Same successful idea, but genuinely different creatives.
4. Track concepts, not just individual ads
My creative sheet now needs another column:
CONCEPT
Because CTR alone doesn't tell me enough.
If five ads built around "saving time" consistently outperform five ads built around "saving money," that's much more useful than simply knowing that Ad #17 had the best CTR.
Now I've learned something about the market.
That insight can influence:
Landing pages.
Email copy.
Future creatives.
Even the offer itself.
The takeaway
I think we're moving from:
Media buying → finding the perfect targeting/settings
toward:
Media buying → feeding the delivery system better signals through genuinely different creative concepts.
Recent advertiser discussions around Andromeda are heavily focused on exactly this tension: how to provide enough creative diversity without simply flooding an ad set with meaningless variations.
For me, the important distinction is:
10 creatives ≠ 10 ideas.
I'd rather have 5 genuinely different concepts than 20 versions of the same winning ad.
For anyone spending seriously on Meta right now: are you seeing better results from increasing creative volume, or from increasing the number of genuinely different concepts?
