Let Your Algorithms Do the Shopping
Behavioral Insider: Can you explain how this 'virtual layer' works in your shopping cart optimization application?
Pete Olson: It gives us a flexibility to recreate the user experience, to identify behavioral components based on clicks, and then make changes to that virtual layer that drive click-through and conversion activities for our customers. In a shopping cart or a lead-generation form you have a data model in the core of the client's engine. But on top of that, you have a graphical layer. What we can do through our software is create a much more pleasing experience, a different set of experiences. For example, it might alter [the experience] from a single page to a two page or a three page. So in that sense, it's really advanced A/B testing, where you are looking at it from a usability standpoint, but you are not fundamentally changing that data model.
BI: Walk me through an example.
So to this shopper what they are going to see in one experience may be a visual layer that is focused around privacy and privacy messaging sprinkled throughout that funnel. A different visitor might see an experience that is much more focused around marketing messaging like shipping and handling and return policies.
We are building an anonymous profile about an individual and then we are reconciling that information with the actual click-through and past activities the visitor takes. With 2CheckOut we start with seven different virtual experiences. And through using the advanced A/B modeling we are going to run each of those seven experiences. Based on the outcomes of these seven different experiences that drives us into what we call a secondary 'wave.'
From a testing principle what you are looking at is going beyond that basic testing concept but you are focusing on the behaviors and focusing on executing click-throughs through a series of checkout steps. 2Checkout has roughly seven unique steps within their checkout funnel and those steps include personal information, shipping information, billing information, a confirmation step and alternative payments, a gift wrapping page, and then ultimately what we call the thank-you step. Amadesa has an opportunity to rewrite that experience of each of those pages.
Rita Brogley: Each of them is running independent A/B tests. We look at the results as the wave progresses and throw out versions that are not performing well and we continue to fine-tune until ultimately we get a single version of a cart we believe performs best. In the future we will be personalizing it to the individual user.
BI: And how is this different from typical A/B or multivariant testing?
Olson: This virtual layer essentially allows us to rewrite the entire page content and all of the dynamic content within that page. We are really looking at the motivational components of the page itself and rearranging the order in which questions are asked. We might create a two page, a three page or more simulated experience to those shoppers. What we find typically is a lot of the market saying put it all on one page. And we find in nine out of ten pages that that is exactly the opposite of what should be occurring,
Brogley: We have seen in the market a lot of fine-tuning going on. The carts are fundamentally flawed in some way because people are tied to a form that is a certain number of pages. What they are doing is changing colors and buttons and a few things here and there. And we believe that you have to really make more dramatic changes to achieve the greatest impact, to get the kind of results we are getting in terms of conversions. With 2CheckOut, conversions rates were over 28%, and on an annualized basis we increased their revenue by $10.6 million. We haven't been able to do that just by fine-tuning.
BI: What are some of the key changes that made the difference?
Olson: Take the debit form page in the payment page. We found by converting that to a single page they had in two distinct steps, we were able to improve conversion by 2.6% just on that alone. We talked about rearranging the way in which the information is being asked. So we took the country field and we had the user populate that first. Based on the response we are able to introduce U.S. and city and state custom field and were able to break those out.
BI: Now, shopping cart optimization is just one piece of the personalization suite Amadesa offers. Are many clients ready to move on to robust personalization?
Olson: We look at personalization along a continuum. For the majority of our customers, they are at a comfort level that starts with A/B testing. And it's almost random A/B testing. Once they get past that and graduate, we find they go into the segment based marketing, segment based testing: first vs. repeat visitors or visitors coming from a geo-location, or they have some sort of activity they have taken on the site and we marked them as a particular segment. That is where we are starting to see a lot of interest. In terms of absolute number of customers running personalization vs. other applications within the continuum, probably a little bit less than 20%.
BI: It seems as if personalization is an idea that appeals to people in the abstract but rarely gets implemented. The technology seems to be far ahead of the impulse. Is that fair?
Brogley: We're seeing the same thing. To some extent, everyone has a comfort level. Our world is very comfortable with analytics and getting more comfortable with the concept of testing. Concepts of targeting specific behaviors and trying to improve the metrics of a Web site based on those behaviors is still coming to light. One of the things we find, though, is that with many of our customers we will begin working with them on pure A/B testing or a shopping cart optimization. It is very measureable and specific. We build their trust. And then it is much easier to move them up the personalization continuum. As a company, if we went out and tried to convert them right away to dynamic personalization they would be very reluctant because there is s still a lot of confusion and it is less tangible. Cart optimization gets to people with something they can measure and show results.