Hi Reader

Last time I told you how our “lazy” variations crushed the bold new ads in the account.

What I didn’t say?

​

If we hadn’t figured that out, the whole account would’ve gone off a cliff.

I didnt realize how close to a crazy drop we were.

It’s easy to think ‘haha lazy’ - and move on.

​

and hard to appreciate how hard it is to build a rock-solid testing system.

​

This is why I see most teams(even well intentioned) bleed out.

Because this is what their tests can look like:

• No structure
• 20 new ads a week
• Just vibes, Capcut, Photoshop

Then they pray one hits. And when it doesn’t?

They blame the algo, or the audience, or Mercury being in Gatorade.

As I said before, there’s a better way.

​

…and since some of you replied to the last email asking how we actually test - and what our actual framework looks like….

Here goes… how we break down creative testing:

🎯 1. Concept Extensions

Your top ad is working.

Instead of guessing, you extend the concept:

• New hooks
• Visual swaps
• Fresh intros

Same spine. New wrapper.

It’s safe. Predictable. Repeatable.

…Big potential upside. Not a lot of downside as it’s proven.

🎚 2. Remixes

This is where you take your best concepts

…and run them through a blender.

• Change the structure
• Flip the pacing
• Move the CTA
• Make it UGC instead of static.

Slightly higher risk. Higher reward.

🧪 3. Net New

This is your swing-for-the-fences zone.

Big ideas.

New formats.

Blue ocean hypotheses.

Most will bomb.

One might change everything.

That’s the structure.

That’s how we test.

And yes, we still move fast.

We still generate TONS of ads.

…and we make sure to make all 3 types.

​

So we know what we’re learning.

We know what we’re risking.

​

And we know how to build from signal.

That’s how the top 1% operate.

​

They test with purpose.

They balance risk like fund managers.

And that is why they win. Again and again.

How do we run this at scale?

Simple: we don’t do it by hand.

We use an army of AI agents that do the heavy lifting:

🧠 Brute Force AI – Analyzes top ads frame-by-frame to decode what’s working

🔍 ReviewMiner – Scans thousands of reviews and comments to surface golden insights

💬 Takuan – Talks to your Meta ad data like a live analyst, minus the dashboards

🖼 Frida AI – Cranks out image variations to bottle the magic of top performers

All built to give our clients an unfair advantage.

No fluff.

No overkill.

Just a better system.

​

If you’re spending $50k+/month and want to turn your creative testing into a machine: not a mess…

Hit reply with “LANES.”

I’ll walk you through how we structure it—and how you can too.

Later,

Shamanth

PS - ICYMI - here is the Intelligent Artifice TikTok -> https://www.tiktok.com/@intelligent_artifice0​

PPS - Some goodies from our socials:
​
​I ran semantic analysis of 𝟭𝟬𝟬+ 𝘃𝗶𝗱𝗲𝗼 𝗰𝗿𝗲𝗮𝘁𝗶𝘃𝗲𝘀 for the healthy screentime app 𝗕𝗲 𝗣𝗿𝗲𝘀𝗲𝗻𝘁 to uncover what’s hitting hard, what’s fading, and what lifestyle apps can shamelessly steal.​
​
​This is why your stats 101 based testing strategy is KILLING performance (do this instead).​
​
​The $170 million ad that looked HORRIBLE. ​
​
​Sometimes the only way to answer “Should we do it?” … is to just do it (this is the $1k stress test we use every time).​