Validity of A/B Split Tests

Posted by Marios Alexandrou, SEO Resume.

You should always be testing PPC ads unless you've completely run out of ideas. The simplest form of testing is with A/B split tests where you run ads head-to-head with, ideally, only one thing different between them. During the tests, you need to let the ads run long enough to make the results statistically relevant before you discard one of the ads for being a poorer performer.

I don't know the exact formulas used to determine if enough data has been collected, but I do know one basic rule of thumb. That rules is that the larger the difference between the click-through rate (CTR), the fewer clicks you need to establish a high level of confidence. So if two ads running head-to-head only differ by a 0.1% CTR, then you're going to need to collect more data than if the CTR differed by 5.0%.

For simplicity, you can put in the results of any given test in to a validation tool provided by Vertster and let them do the heavy lifting.

Related SEO/SEM Articles: