Multi Variate Testing
Testing can be done for direct optimization or for deepening the understanding of the visitor behavior. It can be done with anything on a website; content; multimedia; layout, look&feel, URLs, forms and any other aspect of a website. What is actually tested is usually not very important, as long as we can control and measure it. So it is fine to test a HTML form against a Flash form, a call-me-now button against a chat option, a dancing zebra against an attractive lady.
_tcm89-17741.jpg "MVT-image1")
Multi-variate testing handles the fact that simple testing is not independent from the environment (i.e. a test requires ceteris paribus); testing a pale-green against a lime-green button while next to it somebody else tests an attractive lady against an old mans image gives unreliable results is that is not taken into account. So testing various parts of the website as one experiment and keeping track of the possible correlations between them is what MVT does.
Testing needs a clear definition of success, otherwise nothing is ever better than anything. Typically success is defined measurable in the short term, not so much as contribution to the business. So a homepage of a car manufacturer is tested for how many visitors open the brochure, not for how many buy a car (too hard to track, too far away, too independent).
A logical way to define success is to track if a visitor moves to the next phase of the marketing funnel, so landing pages should lead to interest, etc.
Tags
:
MVT,
testing,
web analytics,
Nedstat