Split testing is a technique whereby two alternatives are tested against each other to see which ones produces a higher value. For example, if you are optimizing a landing page, you might randomly put up 1 or 2 images to see which has a better conversion rate.
However, testing only 2 choices against each other and making a black and white decision about dropping one of the choices are not optimal from a mathematical point of view. There exist powerful techniques that enable the testing of many choices against each other and that also don't rule out a choice forever because of what may be random fluctuations.