Arya

Making ads, breaking tests, fixing funnels. Building DTC engine @toniiqofficial · 2x fintech founder. Reverse-engineering DTC.

[Arya @arya_taware]

@andrewjfaris We ask the same question of creative tests. Honest answer: you rarely get a clean study.

You get direction. Turn the angle on, blended CAC moves. Turn it off, it moves back.

Repeatable beats provable.

[Arya @arya_taware]

@venturetwins We already run AI "iPhone" footage in paid ads. The giveaway isn't faces anymore. It's physics: hands, pours, liquid.

What works: prompt for native phone cues, add grain in post, then QC on an actual phone. The feed lives on a phone, not your monitor.

[Arya @arya_taware]

@andrewjfaris @TaylorHoliday We write ad copy with AI every day. The pattern: elite at variation, useless at conviction.

Feed it a real customer transcript and it writes 50 strong lines. Ask it what to say and you get the average of the internet.

The off-limits part isn't the writing. It's the deciding.

[Arya @arya_taware]

Same product. Same mechanism. Two headlines.

One said "inflammation." One said "fiber."

The fiber ad acquired customers at half the cost.

The mechanism didn't change. The vocabulary did.

You're not testing products. You're testing the words people already use for their problem.

[Arya @arya_taware]

Most people use AI to write ads. I use it to run the whole pipeline: script, voiceover, video clips, captions, static variants. One person doing the output of a 5 person creative team. The tools are not the hard part. The system is.

[Arya @arya_taware]

@venturetwins Seeing the same from the buyer side. We moved our static ad production to generated-first this year. The surprise wasn't quality, it was that less polished versions outconvert the designed ones. The low end of design wasn't undervalued. It was overbuilt.

[Arya @arya_taware]

@TaylorHoliday We see this in the creative data too. Attention and purchase intent are different systems. Our highest-engagement ad this quarter had near-zero add-to-carts: 21% CTR, nothing behind it. Engagement tells you the scroll stopped. It doesn't tell you a loop closed.

[Arya @arya_taware]

@ashvinmelwani Voted AB test tool, but the uncomfortable finding for us: the two methods gave us different winners on the same offer. Meta's delivery system picks its own audience per ad set, so you're testing offer + delivery, not offer. Tool for truth, Meta for scale.

[Arya @arya_taware]

Attention and purchase intent are different regions of the brain.

A 21% CTR ad with zero add-to-carts isn't underperforming. It's mis-firing.

You stopped the scroll. You never closed the loop.
Stop optimizing for hook rate. Optimize for the second activation.

[Arya @arya_taware]

Meta FAIR published a brain foundation model in March:
720 subjects, 1,000+ hours of fMRI, predicts 20,484 cortical vertices per second.

We layered our own Meta CPR data on top.
Now every ad gets a brain-activation score before it goes live.

Score 70+ → it ships. Below → it doesn't.

The first time DTC creative testing wasn't a guess.

[Arya @arya_taware]

In our account this quarter, static ads are outselling our videos.

Same product, same offer, less polish, higher conversion.

The mistake is treating video as the default and static as the cheap fallback. Both are formats. Test them like formats.

[Arya @arya_taware]

@johntech778 What is your methodology for judgment before shipping creatives to meta?

[Arya @arya_taware]

AI killed production cost as a variable. We spin up 100 ad variants in a day now.
The new bottleneck is evaluation: which 3 will actually convert?
Production is cheap now. Knowing what works isn't.
That's the real moat in 2026.

[Arya @arya_taware]

The most common ad-creative failure isn't a bad hook. It's a great hook with zero purchase intent.
We ran a breakup-metaphor ad: 21% CTR. Zero add-to-carts.
Beautiful. Useless.
We call it the Attention Trap — and AI tools make it dramatically easier to produce one