Once of the best use cases of AI is to monitor your ads with computer vision to figure out what creative is working. There is so much waste, but the good news is there is now an easy solve to understand what works.

Money is hiding in your creative. The CTA was appearing at second 22. The ad was 30 seconds long. That detail was hiding inside a campaign that had already run for three months and millions of dollars. And here's the thing: nobody thought to look. Not because they didn't care. Because the entire analytics industry has been built to answer the wrong question. For twenty years, the question has been: what performed? Nobody built the tool that answers: why. The cycle nobody questions Here's how it works at every brand spending serious money on paid media. You produce creative. You run it. You look at dashboards. Cost per conversion went up or down. You brief your agency. They produce more creative based on instinct, trend, and taste. You run it again. Nowhere in that cycle does anyone ask: what specifically inside the creative drove the result? The information is there. It's inside every frame of every ad you've ever run. The rhythm of the cuts, the moment the product appears, whether the talent made eye contact, when the call to action landed. The winning pattern exists inside creative that's already running. Nobody has been looking at it correctly. How you actually read creative correctly Computer vision watches every frame the way a great creative director watches one. At a scale no human can match. It sees the rhythm of cuts and transitions, when the product appears and how long it holds, whether the talent's making eye contact, when the CTA lands, what happens in the first three seconds that determines whether someone keeps watching. Then it connects what it saw to what happened in market. Not "this ad performed better." But why. What we find inside real campaigns CTA at second 22 instead of second 8. The audience had already decided to scroll before the ask. Talent not making eye contact. Lower emotional connection, lower share rate, consistently. Product appearing after 4 seconds in a 6-second ad. The audience saw the brand but not the reason to buy. 78% of spend going to food-beauty close-ups. 14% going to lifestyle creative. Lifestyle generating 3.3x the shares at lower production cost. The most validated approach, getting the most aggressively underfunded budget. In every case: the information was already there. The winning pattern existed inside creative that was already running. Nobody had looked at it correctly. What happens when you run the loop A Fortune 500 logistics company ran one full loop. Connected their Meta account, computer vision analyzed every ad, we found the variables explaining the performance gap, they refreshed based on what we found. Cost per conversion dropped 27%. A multi-brand restaurant group runs campaigns across six brands simultaneously. The performance spread between best and worst creative mapped to 10-15% transaction lift potential. At their scale, that's 2.4 million additional transactions per year. Same spend. Better creative. More revenue. Why the tenth loop is worth more than the first Here's where it stops being a campaign optimization and starts being a strategic asset. Every loop generates data specific to your brand. Your creative patterns, your audience responses, your category signals. The model gets more accurate each pass. Not in a general way. In a brand-specific way. After 12 months, you don't just have better creative. You have a proprietary intelligence that knows what works for your brand, with your audiences, in your category. That intelligence doesn't transfer to a competitor. It doesn't degrade when models improve. It compounds because it was built from real signals inside real campaigns. A competitor starting today can't buy their way into a year of your loops. That's a structural advantage. Most analytics run in a dead-end model. Data in, dashboard renders, human stares at charts, nothing automated happens next. The loop is different. Every insight becomes a brief. Every brief becomes creative. Every deployment teaches the next cycle. The question stops being "did it perform?" It becomes: what specifically made it perform, and how do we build more of that? Where this goes The future I'm building toward isn't one where brands keep running billions in creative blind. It's one where every campaign teaches the next one. Where the gap between what you make and what actually works closes a little more every loop. Where creative intelligence compounds the way great brands always have. One insight at a time. We're early. But the direction is clear. The brands that start running loops now will have 12 months of compound intelligence by next year. The ones who wait will be playing catch-up on data they don't have. The money's always been in the creative. Now we can see it.

Great day at AdPipe

Amazing to see AdPipe featured in the @WSJ alongside @ZieglerCAT

One of the highlights of my career working on something creative with @JasonIsbell