The example is a blue vs orange button? What value does this post hold that hasn't been said with greater rigour elsewhere? If your org is testing stupid shit your AB tests will be stupid. The only time your tests should have that level of granularity is if is incidental because you are Indeed and testing, feature flagging, releases, etc. are all rolled into a single fully integrated tool like Proctor so everything gets tested by default all the time with a deviation alerts.