Sometimes, you realize that certain variables, usually those that are not easily manipulated, have a powerful impact on the success metric. For instance, maybe cellular users of your web site are likely to click less on anything, in contrast with desktop customers. Randomization could lead to set A containing slightly
