Getting more return from your A/B testing

Many companies struggle with the ROI of their conversion optimization program. It is common not to find enough “winners” which puts pressure on the ROI of the program. Here are 4 ways to change this.

Value

Consult carefully with each other about what the right KPIs are for the optimization program. Conversion in general is often not the right KPI. More new customers, more customer sales, more returning customers, less churn or more sales of specific products (with more margin/return) is often much more relevant for the long term. Setting up and assessing A/B testing on these may not directly create an overall improvement (due to cannibalization), however, it does create another revenue stream. Which can have very positive effects in the longer term.

Verified

Science also spends billions of dollars each year on research within your industry and for your services and products. Such valuable information is easy to find through Google Scholar and ScienceDirect. Free for students. With the advent of subscription-based DeepDyve, getting surveys has already become much more affordable. The reports found in these three platforms give direction on what you can optimize. And give an early idea of how successful your hypotheses will be.

View

Through the tag management solutions now common, you can fairly easily create a script that, at the start of a visit, asks why someone is visiting the Web site. At the end, you then ask if they succeeded in completing this task. These are simple questions with a high response rate. You feed the answers into your analytics solution as an event. This allows you to see the success rate per visitor task and the difference in behavior between successful and unsuccessful visits. This is very valuable information to determine where optimization can have an effect.

Voice

If you already know through your ‘view’ efforts where you can improve and ‘verified’ has given you direction on how to improve your KPIs, it becomes a matter of understanding what specifically is not clear on the site. 5-second tests are very helpful here. Online panelists are shown a web page or message for five seconds and answer what they saw. This is relevant input to display the message you want to test, quickly findable and clearly worded. For the current site and any variants you want to include in an A/B test.

Ton Wesseling fills in the recurring conversion column each month on behalf of Online Dialogue Emerce magazine: the magazine about “the next step in E-business.” This month he writes about: Getting more return from your A/B testing.

Click on the image to enlarge it, or download the pdf here.