Search marketers can use split A/B tests to help choose between specific ad copy and keywords, landing page designs and information-capture forms, among other factors. But the results are only as good as the test--run a skewed trial and you may come to a conclusion that isn't true, or worse, actually miss out on a higher-converting tactic. So Resolution Media's Jeff Campbell lays out the 10 steps for split A/B testing success, starting with choosing just two elements to compare.
Pick the two and then determine what the success metric is, how long it should/will take to actually see results, as well as the desired confidence level. Next, choose how long to run the test, factoring in effects like seasonality and traffic irregularities, and then choose the sample size. Now run the test, analyze the results--being sure you've met your own latency, sample size, and confidence level barometers--and choose a winner.