Today, eCommerce is a fast moving game where standing still is not an option. We’ve talked in the past about fostering a culture of change and how important it is to build a team that can deal with a healthy dose of chaos. We work regularly with clients that have ignored change as well as those that have been on the fore front of it. More often than not the later benefits from higher valuations as such change elevated their business.
But as the web matures [and hopefully your company grows], change becomes more complex, more costly and often with greater risks. Today, it’s all about conversion. It’s great that traffic grew by 20% but what does that mean? What is the business impact of that? Was it more sales, less sales even, or perhaps a change in average order size. Years back when we were helping to sell a website buyers only wanted to know about unique users and page views. Today, the dive into analytics is much greater.
With that in mind, we wanted to chat about testing, specifically A-B Split Testing. You get the idea to launch a new site or make changes. But how do you know the impact of such changes. If you have a successful site now, going on gut feel is full of risk. The consumer is a tricky beast and the world is full of reports about consumer behavior. But so many variables play into it we think such generalizations are at best a guideline of where to begin to make change, then you need to test that change without damaging the business.
A-B Split testing is simply launching two versions of something, such as a landing page and randomly directing half the traffic to each. The user is unaware of this and the versions should be shown in parallel to mitigate variables such as time of day etc… While it’s called A-B, certainly you could test more versions at the same time but obviously you want to make sure your sample size will be enough. Split tests really only work at testing one variable at a time, such as tabs versus buttons. If you throw multiple variables at, it’s difficult to determine how the interaction of variables determined the winner. Set up your tests much like a flow chart, testing one variable at a time and then another to arrive at the desired goal.
Defining the goal is important. Remember page views are great but ultimately in eCommerce its orders that keep the lights on…. There are a number of solution providers out there today to help set up testing environments as well as churn out the analytics to determine the best options. However you approach such a project, be sure to understand your goals first and design the tests from there. Do you want more sales, more leads, lower cart abandonment or fewer customer service calls. With goals in hand you can then design the split tests to maximize the ROI.