E-commerce: How to Use Analytics to Plug Holes in the Conversion Funnel
3. Determine how long you need for each test to be valid, then act. Gniwisch noted that he prefers to respond quickly to Web tests, and adjusts his site on the fly. “We think you have to get between 100 and 150 orders for a valid test,” he said. For instance, Ice.com tested two landing pages off of an MSN ad in less than a day. “We had the control, which was information heavy, and the test, which was more artsy and image heavy,” he said. “Whichever test won out in the morning stayed up all day long.”
Gniwisch pointed out that tests like that one aren’t about driving more traffic to his Web site, but rather pushing the traffic that does come to Ice.com deeper into the site. “If I have a 1 percent conversion rate on my site, and I have the opportunity to bring more people on the site, sure I’ll get more sales,” he said. But he believes that’s the wrong way to look at the problem. “If you plug the funnel and fix the places where people are exiting, you can make more money with the people who already are coming to your site,” he said. Using that attitude as a guide, he was able to increase conversion from 1.2 percent in 2005 to 3 percent in 2006, all without spending more on ads to drive more traffic to the site.
He recommended focusing on the following metrics as a starting point for further analysis on your site:
Number of visitors;
step-by-step conversion rates; and
sales per visit
And what can go wrong if you don’t use analytics? Gniwisch concluded that you can make “a particular mistake, then make it over and over again, because you don’t know it’s wrong!”