Social Rocket
Kroger teamed up with VidMob and MMA Global to apply an AI model to real campaign data in order to better assess the conversion impact of ad creatives. According to the analysis published on 30 September 2026, the model correctly predicted the e-commerce conversion performance of the creatives studied in 81 percent of cases.
The figure is not a general "81 percent guarantee" for AI creative scoring. It applies to this specific study using Kroger data. What is more interesting is how closely creative characteristics and actual purchase data are now being linked.
1.934 creatives from Meta and DV360
According to MMA Global, the analysis covered 1.934 image and video assets from a year of running Kroger campaigns. Delivery ran across Meta and Google Display & Video 360, among others.
VidMob used AI to break the ad assets down into characteristics such as messaging, branding, narrative structure and visible human interaction. These signals were then linked not only to clicks or view rates, but to actual e-commerce conversions.
Marketing Dive also reports that creatives that more closely matched the recommendations derived from the model achieved, on average, a four times higher conversion rate. According to the study, a hypothetical reallocation of budget to higher-rated assets could have enabled more than twice as many conversions with the same media budget.
These are modelled or observed study results – not a promise that the same uplift can be reproduced in every campaign.
Why creative scoring is becoming more important now
The practical benefit lies in speed. Generative AI, creator content and automated variant production mean that advertising teams can test more and more assets. At the same time, it is becoming harder to run every variant long enough to collect reliable performance data.
Scoring before or early in a campaign can therefore help sort out weak variants faster and allocate test budgets more precisely.
This fits a broader trend: with Meta Ads, AI creatives are growing while attribution is being reshuffled at the same time. Our overview of AI advertising and automated ad spend also shows how strongly decisions about delivery and optimisation are now being taken over by models.
For Kroger, creative is only part of the data advantage
For years, Kroger has been building its advertising business on first-party purchase data. Most recently, the retailer reported strong growth for Kroger Precision Marketing; at the same time, retail data is being connected more closely with platforms such as TikTok, YouTube and CTV. You can find more in our analysis of Kroger's retail media business.
In parallel, Kroger is experimenting with AI directly in the shopping funnel. The company is among the retailers bringing agentic shopping assistants into product search and the shopping basket.
What marketers should take away from the study
The most exciting point is not the 81 percent figure alone. What matters is that creative intelligence is moving closer to real business outcomes.
For advertising teams, this suggests three sensible guidelines:
- Creative scoring should be trained on conversion or revenue signals, not just attention.
- Models must be validated on your own campaign and audience data.
- AI scoring does not replace experiments, but it can prioritise which variants get media budget at all.
The more ad material is generated automatically, the more important this second layer of automation becomes: not just producing more creative, but identifying faster which creative is likely to contribute economically.