Kellanova, which makes your favorite snack aisle hits like Pop-Tarts and Cheez-It, is putting AI in the driver’s seat of its ad creative — specifically through a year-long test with Vidmob and MMA Global. The company scanned 443 creative assets across ten brands on Meta platforms using AI models trained to predict which combinations of elements (logo position, tone, pacing, etc.) snagged at least three seconds of attention. It appears to be working — the view-through rate on assets scored by the AI doubled, and ROI from those pieces edged up 11%.
They didn’t stop there. Kellanova spun up separate predictive models for sweet and salty products, producing pre-flight creative benchmarks with around 83% accuracy in forecasting view-through rates. The whole system, powered by large language models like Google’s Gemini, lets Kellanova greenlight content before blowing its shrinking ad budget. With a $155 million cut in marketing spend from 2023 to 2024, every second (and dollar) counts. Now, the brand’s expanding its AI toolkit beyond Meta, eyeing TikTok, Pinterest and Reddit next.
Naturally, all this has implications for the agencies on Kellanova’s roster. The company is shifting toward output-based pricing and using AI-generated creative scores to influence agency compensation. This has sparked a bigger industry conversation: does AI finally provide the metric CMOs need to justify creative spending to the C-suite? Kellanova’s not trying to play Creative Director with an algorithm — yet — but they are asking their partners to put a little more “performance” in performance marketing.
Read more at Digiday.