April 29, 2008, 12:00 AM

TrendMicro gives multivariate testing a passing grade

The online software retailer reports improvements in revenue per visitor, sales and conversion rate after testing changes in multiple elements of its shopping cart page. The company’s e-commerce technology provider, Digital River, carried out the tests.

The European unit of online software retailer TrendMicro Inc. tested changes in several elements of its shopping cart last summer and again early this year, and each trial produced a new combination of elements resulting in significant improvements in revenue per visitor, sales and conversion rates, says Keith Reed, TrendMicro’s online marketing manager for the Europe, Middle East and Africa.

Reed was familiar with A/B testing, which pits one version of an element against another, but it was the first time he tried varying several elements at once and then combining the most successful options to achieve an optimal design. “It was small changes, so we were constantly refining the test,” Reed says. “Rather than just an A/B test it’s the whole alphabet. And once one thing has a statistically valid success rate we roll it out.”

The first multivariate test last summer tried different combinations and modifications of the security messaging, “continue shopping” button and discount field. The winning combination, which eliminated the discount field altogether, produced a 9.7% lift in revenue per visitor, a 6.8% improvement in conversion rate and 10.2% increase in sales.

A second trial this winter varied the header color, button design, placement of user acceptance agreement and credit card logos. The most successful combination produced a 10.5% increase in revenue per visitor, 16.2% improvement in conversion rate and a 9.9% sales lift-but only for weekend visitors. Weekday results were no different from the previous shopping cart design.

“We have a very cyclical business with sales peaking Tuesday and dropping to a low on Sunday, then going back up,” Reed says. “I have no idea why.” Nor could he explain why changing the shopping cart design would improve results on weekends but not during the week.

But one result was consistent throughout the week: leaving off the logos of credit card companies was a bad idea. “In certain markets it was very unsuccessful,” Reed says. “People were dropping off because we didn’t show the Visa logo.”

comments powered by Disqus




From The IR Blog


Bart Mroz / E-Commerce

How smaller retailers can utilize data as effectively as Amazon

Smaller companies have more constraints, but once they set priorities can still benefit greatly from ...


Philip Masiello / E-Commerce

3 reasons retailers fall short in email and social marketing

Reason one: They’re constantly trying to sell their customer, rather than to help and engage ...

Research Guides