> Larger lunch tables were "driving more than a 10% difference in performances". A fact that would probably have gone undetected without such data analysis.

Isn't this poor data analysis, though? Correlation doesn't imply causation and all that. I think it's more plausible that people who sit at 12-person lunch table have more friends on the company, and that drives performance up, than the lunch table affecting performance. I mean, it could be any reason, but I highly doubt the lunch table is the cause of the higher or lower performance.