In our definition of marketing, we make it clear that marketing research is where all marketing decisions should begin. In fact, we consider research as being so critical that in our Marketing Research tutorial we state research is the “foundation of marketing.” Unfortunately, many marketers, especially those in smaller firms, fail to devote enough effort to research. The main reasons they cite for not doing research primarily center on three issues: too costly, too time consuming, or too complicated.
However, there is one business segment that seems to be embracing research – online retailers. It is true that most companies with a web presence are doing some research, such as using a web analytics program to track where visitors go on a website. Yet, online sellers are going further by implementing experimentation techniques as a way to find out answers to such questions as: What is the ideal location for product placement on a website? What is the most effective wording to include with product information? What is the most attractive website design elements (e.g.colors) for displaying a product?
For online sellers, the preferred method for conducting research is through a technique called A/B testing. With A/B testing, two different versions of information are shown to different website visitors. For instance, if a retailer is testing to find out which picture of a product attracts more interest; half the visitors to a certain webpage will see the image of a product in one way (e.g., placed on the left) while the other half will see the image of the product in another way (e.g., placed in the middle). The option producing the best results (e.g., higher sales, higher click-thru rates, etc.) may then be chosen as the webpage layout to be presented to all visitors.
Additionally, this type of experimentation can extend well beyond just changing one variable in two ways. Using multivariate testing, research can be conducted where several different changes are tested at the same time (e.g., different font size, different background color, different customer support statements, etc.). Unlike A/B testing, where 50% of visitors are shown different content, with multivariate testing the breakdown will depend on how many variations are being tested. For instance, ten variations may mean only 10% of visitors are exposed to a certain testing page.
Good information on using experiments in online retailing can be found in this story from Internet Retailer. It provides insight on how several online sellers use A/B and multivariate testing. It also presents information on different firms offering these research services and the different rates they charge.