I like the look into the process of A/B testing, but what is not obvious that for large amounts of data multiple columns is going to be more effective?
Any suggestions?
Sorry. That was a typo. Do you think that your data shows a clear number of items were it makes sense to use multiple columns. It looks like to me that the multiple column layout is more effective, but I can't see a clear cut-off point.
I don't think the data is clear enough, agreed. Unless there is a better way to compare, we might actually build both and let the user decide. So long as our targets for speed and accuracy are achieved.
Having tried it on my phone, I think that you may have failed to account for the orientation of the device. The single column was better on my phone, as it was portrait. On my desktop the multi column version was better presumably as it was landscape.