NewForm

To read the full mathematical analysis, visit the post here (https://lnkd.in/gPfV5FPj) We often get asked, "So you guys are making AI-generated TikToks, right?" No. Paradoxically, you end up spending more money to get to the same outcome. And you can prove this mathematically. On short-form video platforms, the power law reigns supreme. The best ads get outsized spend and results, thanks to TikTok's mastery of content distribution. This platform shift, coupled with Apple's iOS 14.5 App Tracking Transparency, means AIGC is lighting money on fire. ▶ First, a distinction between paid vs organic: With AI-generated content (AIGC) tools at your disposal, it seems logical to create numerous assets, test them organically, and “boostˮ the winners. But this is doing it entirely wrong. Organic content's efficacy lies in the ability to get distribution. For paid content, it's about returning more dollars than you put in. Usually, the most viral organic videos have terrible return on ad spend. When you're paying for distribution, you can be a lot more salesy. When you're at the mercy of the algorithm, you're forced to make low converting content. On organic—AIGC drastically reduces content costs and the marginal cost of distributing a piece of content is near zero. So, every asset becomes de facto EV positive. If an AIGC video costs $1 to create, and the expected value is 1,000 views, then you're already making money. ▶ But for paid content every new asset costs you money: Here, volume alone doesn't guarantee success. Each new asset needs (paid) testing, and there's no guaranteed positive EV. There is also the elephant in the room: the opportunity cost of spending the incremental dollar on testing vs your top proven ad. If you define some efficacy metric, E, and plot a distribution of all content, you get something pretty close to a normal distribution. From our experience, 1 in 30 assets drive over half the results in an account and are considered “true winners.ˮ These "true winners" are defined as the top ~3% which is roughly two standard deviations above mean. The deal with normal distributions is that they are extremely sensitive to means at the fringes. So, a slightly lower mean efficacy of content means you need to test a lot more videos to discover a “true winner.ˮ While 1 in 30 UGC ads may be a winner, with a lower base efficacy for AIGC you see only 1 in 100 or 1 in 200 ads become winners. This means your cost per creative winner skyrockets.

Hamza Alsamraee at Airbridge's MGS 2024 speaking on the death of traditional targeting and the birth of self targeted content for user acquisition. Always amazing to be back in Seoul seeing familiar faces and meeting new ones in the ecosystem. Many thanks to the AB180 - 에이비일팔공 team as always for a great event. Looking forward to MGS 2025!

[$5k referral bonus] We're expanding here at NewForm and hiring a performance marketing lead to manage and scale our clients' iOS campaigns If you/anyone you know has experience scaling iOS campaigns on Meta and Tiktok, we'd love to chat. We'll send $5k as a referral bonus if we end up hiring the candidate Linkedin, pls do your thing!