Kroger is testing a data-driven way to predict which ad creatives are likely to drive e-commerce conversions—before campaigns even launch.
In a study with Vidmob and MMA Global, Kroger analyzed 1,934 video and image assets from campaigns running across Meta and Google DV360. The predictive model was able to forecast e-commerce conversion performance with 81% accuracy.
The study also found that creative assets following the model's recommendations generated 4× higher conversion rates, while reallocating media spend toward higher-scoring creative could produce up to 2.2× more conversions without increasing the advertising budget.
What this means for e-commerce marketers
The bigger takeaway is that creative optimization is moving from post-campaign analysis to pre-campaign prediction.
Instead of launching multiple creatives and waiting for performance data, marketers can potentially use historical creative and conversion data to identify which messaging, visuals, narrative structures and human elements are more likely to work.
Interestingly, Kroger's study found that creative centered around authentic human experiences performed better than ads focused primarily on products or shopping.
For marketers, this points toward a more measurable creative workflow:
Data → Predictive scoring → Creative selection → Media allocation → Conversion
As generative AI makes it easier to produce large volumes of advertising content, predictive scoring could become increasingly important for deciding which assets deserve the budget—not simply which assets can be produced.