Theoretically, in a closed system, to find out the best among two or more variables, each variable can be tested against one another and the best of all variables will be the winner. In the context of website, which is a closed system, the elements in a page can be considered as the variables and it is possible to do a similar test.
To effectively test a page it’s important to note the elements that need to be changed on it. Taking an example of two elements to be tested, two versions of the page are created each having a distinct changed element. This is where the split comes into action. This may be a content split, where the display of one page is varied dynamically by altering the content of the page, or split path, where traffic is split between two different URL’s. Whichever split testing method is chosen, virtually visitors land on any of these two versions of pages randomly or sequentially. Moving on, test can be done with more than two elements. So if there are four elements to be tested, there will be four versions of a page that will be visible to four different groups of visitor traffic.
To successfully arrive at a conclusion, it is prudent to let it run long enough so that there is sufficient data. With a shorter time span, the results obtained may be deceiving because the test may be missing out the significantly different traffic patterns during the weekend or quarterly results and such aberrant periods. Results such as conversion or profits resulting out of different versions of a page are observed only when there are sufficient data. This helps in finally determining a probable winner of the two or more different page versions.