Kroger can forecast e-commerce purchase conversion from ad creative with 81% accuracy before a campaign launches, according to a study conducted with data firm Vidmob and trade association MMA Global and shared with Marketing Dive.

The study combined Kroger's live campaign data, Vidmob's creative data platform and MMA Global's industry research standards to score creative on messaging, narrative structure, branding and on-screen human interaction. Creative aligned with the model's recommendations delivered an average four-times improvement in conversion rates. Reallocating media spend toward higher-scoring assets could potentially generate more than twice as many conversions from the same ad budget.

"[There are] two types of creative optimization. One is actually making changes, moving pixels around, changing the nature of the creative. But the other is just like moving dollars behind better creative," said Alex Collmer, founder and executive chairman at Vidmob.

The research used artificial intelligence to analyze more than 1,900 video and image assets from Kroger campaigns running across Meta and Google's DV360, identifying creative attributes tied to e-commerce conversion.

Kroger's test builds on a similar effort by Kellanova, also run with Vidmob and MMA Global, which connected creative decisions to view-through rates as a proxy for sales. The new study goes further, demonstrating the connection to actual sales. Collmer argues that shift changes the optimization workflow itself.

"Historically, people made [ads] and then, effectively, spent money blindly behind them, and then after the fact, looked to make optimizations once they saw performance," Collmer said. "The ability to get ahead of that and basically not spend behind assets that aren't going to work is a pretty big deal."

The study also attacks a long-standing budgeting convention: creative spend classified as non-working and media spend classified as working. "For too long, the industry has existed in this world where creative investments are labeled non-working and media investments are working. That's just absurd," Collmer said. "Every marketer you talk to says creative is 50% to 70% of results. How on earth can the thing that's the far majority of results be the waste?"

Testing at scale

Predictive creative scoring gains urgency as content volume explodes. Generative AI adoption and the rise of creators as a media channel are driving what Collmer calls "a 10-times or 100-times expansion of the number of assets." Without an up-front signal, marketers risk spending more money behind content before knowing what it will do.

Vidmob moved to commercialize the approach in June with Vidmob360, a tool that lets clients connect creative performance data to AI tools of their choice, including Claude, ChatGPT or Copilot. The company also offers a headless AI recommendation system that returns machine-readable creative adjustment guidance.

As advertisers adopt the same AI tools, Collmer contends creative performance data becomes a differentiator. "Because they're all adopting the same tools, then using those tools won't actually be an advantage," he said. "It's not going to help Kroger or their competitors win market share if they're using the same tools."

The same dynamics apply to creators, whose income depends on content performance. "Creators are super hungry for data because their paycheck is based on how their content performs," Collmer said. "They see performance as a way to be more successful."