In a randomized experiment, chance decides who gets each treatment, so the two groups differ only by the treatment and by chance. The crew measures the difference of the group means, then asks what chance alone could do: it writes every result on a card, shuffles, deals two new groups and records the difference, many times. If the observed difference is at or beyond the shuffled differences less than 5% of the time, the difference is significant and the treatment is the best explanation. Otherwise chance alone could have made it.