He doesn't mention it, but that's easy to determine. Simply see what the average contribution of people who contributed during the A/B testing was per image/button combination. If there not all about the same then there may be some effect (negative or positive) that the added marginal people brought. If they're about the same, I think its a decent assumption to believe that it will be the same going forward.

And of course, you can simply validate it by seeing if it is the same going forward.