We have been A/B testing two versions of a template for some time now and have consistently seen a higher open rate in one version that we simply cannot explain. Hoping you all might have some ideas for things to look into.

To give some background, the two versions both use the same code base as a starting point, they both have the same text, the same links, the same subject line/preheader/from name, the same images with slight differences in cropping, and the file sizes are within 2 kb of each other (around 35k). The main differences come down to basic styling such as background color and font sizes, columns stacked vs side by side, but nothing drastic.

Any ideas for things to check? My assumption was that if the inbox view and sender information were identical that we could expect roughly identical open rates. The open rates are different by about 4% consistently across multiple tests using large lists.