In November 2007, Accenture conducted a workshop with ABN AMRO to introduce the Accenture Digital Optimization engine and determine the experiment design, hypotheses and creative elements to be tested.
ADO is based on the science of choice modeling, which won the Nobel Prize in Economics in 2000. It uses a “full factorial” design approach to test all possible combinations of Web page elements and can achieve robust and granular test results more quickly than other multivariate testing approaches.
ADO can also detect the inter-relationships or interactions between page elements; for example, the combination of an image and a call-to-action button. Positive interactions can produce improvements in conversion rates that are greater than any uplift achieved by only considering each element in isolation.
The ADO multivariate test on the bank’s Internet Savings landing page went live in May 2008 and ran for four weeks. When the test went live, ADO started allocating one of a possible 648 page variations to each unique visitor that arrived to the Internet Savings home page (visitors who returned to the page during the test would see the same version of the page each time).
In total, seven page elements or “attributes” were tested during the experiment to determine the optimal variation for each. However, to determine the overall optimal page, ADO considered how each attribute performed in combination with the other attributes to detect any possible interactions. The test uncovered multiple significant interactions among the attributes and revealed, among other insights, the critical importance of showing a prominent call-to-action “above the fold,” meaning visitors to the page could see it immediately.
The goal of the test was to find a single optimal page that provided the highest rate of conversion. While such tests do not normally discover the most effective page variation until the end of the experiment, after less than a week ABN AMRO noticed sharp spikes in conversions that could only be attributable to visitors seeing new, more effective versions of the landing page.
After the experiment was complete, analysis of response rates revealed that ABN AMRO’s original page had the weakest performance of all the 648 variations tested. Conversion rates therefore improved significantly from the moment that customers saw other versions of the page.
The most effective page of all was one the bank would never have thought to implement without multivariate testing. In fact, when Accenture presented our final report, the bank’s lead Web designer was surprised at the counterintuitive results—the winning page broke orthodox design rules.
The experiment revealed that constructing Web pages according to assumptions, opinions, rules of graphic design or results of simple A/B tests does not guarantee customer response. Only through Accenture Digital Optimization could ABN AMRO have tested so many page variations simultaneously and identified the page that maximized customer conversion rates.