STRATFOR's Aaric S. Eisenstein on Barrier Page Optimization
A search on Mexico's problems with drug cartels may lead readers to the gateway page for Austin, Texas-based publisher STRATFOR's analysis titled "Mexico: Politics and Narco-Corruption in Michoachan." But once there, nonsubscribers have to provide e-mail addresses to progress through the site. The global intelligence STRATFOR provides its readers didn't always make visitors move past this page. So the publisher recently tested its gateway data capture page, which it calls the "barrier" page. Using 11 variables, STRATFOR tested 103,680 unique versions of its barrier page and picked the winner, which represented an 81 percent improvement in conversions.
In order to optimize its barrier page, STRATFOR hired landing page optimization firm SiteTuners of San Diego in September 2008. STRATFOR and SiteTuners coordinated the test on the highly trafficked site by showing each new visitor a page, at random, that combined the 11 variables in one of 103,680 ways. (The variables included headlines, calls to action, button formats, and the addition of "trust and credibility symbols.") STRATFOR pulled out the lower-performing pages as the test progressed, leaving one challenger to pit against the original page.
Aaric S. Eisenstein, senior vice president of publishing for STRATFOR, elaborates on the process.
Target Marketing: What made STRATFOR believe that its barrier page presented a conversion problem?
Aaric S. Eisenstein: Well, I don't know that we necessarily thought there was a problem, as much as there was an opportunity. We hadn't yet gone through a site optimization process or a testing optimization process for that page. So we wanted to see if we couldn't make that page perform even better than it was.
TM: How did STRATFOR identify the 11 variables it wanted to test?
AE: Those actually were identified in conjunction with SiteTuners' site designers ... We tested the entire page, and it was really a question of dividing the page into the relevant areas. So it kind of flowed naturally out of the architecture of the page. ... The headlines should be tested independently, or the call to action should be tested independently. The body of the page should be tested independently. As compared to, "Let's do one variable," which is the headline plus the body of the page. So it was kind of a graphical question and a functional question, looking at the different elements of the page and how they function independently of one another.