Most web development is done by trial-and-error. Trouble is, nobody bothers to take note of the trials or the errors! Split testing is a little more formal. In this type of test, we plan the test out ahead of time, with the goal being specific to the type of research being done.
For example, let's say we wanted to see if a left-side navigator or a top-side navigator was better for our site. When I say better, I mean better conversion, since this test only measures people who have already made it to your site. So in this case, we want to try out two different navigation page solutions, A and B. A, with the left-hand navigator is saved with a filename fileA.htm and B, fileB.htm. Now, the primary page that they visit will have a piece of javascript in it that will evenly distribute visitors between the two pages (perhaps switching back and forth once per hour.) The logfiles are then analyzed for each page, and, using a special kind of log report, we can measure which of the designs, A or B, provided best performance.
Ok, all of that sounds fun, but the real important part of this is that we make a note of it. We scrap the losing design, and introduce a new change to the layout... say, larger fonts on key areas, or color changes, etc. ... Now design A will compete with design C, and so on.
I think it's clear that this kind of testing takes time. But the time is calendar-time, not so much billable consultant time. As long as good notes are kept, patience is the key to this research. This type of research can be carried out with not only designs, but keyword choices, PPC Ad headlines, and other things.
Because it is time consuming, brainstorming about what to do first is important - to narrow down the plans to the problem areas. These are usually known by doing a good log report ahead of time. |
