What is a control page or control group?
The premise of Experimentation is a controlled comparison. Whilst this could compare two or more brand-new elements, pages or journeys, traditional AB or ABn Testing compares the current version of the element, page or journey to others. This current or default version is what we refer to as a control.
A control page refers to split testing, wherein you are redirecting one page to another. The control page would be the original page, with a Challenger or Variant page being the one/s you’re redirecting to.
When you divide your traffic (e.g. 50/50 for an AB test), you’ll have a group of people that see the original page, and a group that see the variant. The group that see the original page are referred to as the control group. Control group, as a term, is commonly found in medicine for single or double-blind trials. The idea is that these people don’t know whether or not they’re part of the Experimental group.
Why is a control group or control page needed?
When looking at the results of a test, you aim to make a comparison about how much better or worse the new version is compared to what you currently have. To effectively make this comparison, you need to measure your default version as well as the experiment/variant, and thus be able to compare conversion rates between the two.
For example:
- Default page has a conversion rate of 10%
- New page has a conversion rate of 20%
- The difference in conversion rate is 10%, which results in a 100% lift (increase) in conversion rate
These comparisons cannot be made without a control group or control page.
Can I avoid having a control?
It is always suggested that you have a control group/page. If site traffic is a concern, you could consider running a control page/group to a lower percentage share (e.g. 10% instead of 50% for an AB test).
Whilst doing so is not recommended, having an indication of how users are performing on the site is important, and even a suggestion is better than having nothing. Using a technique like this comes down to your risk tolerance – the higher it is, the more traffic you can sacrifice.
How does Webtrends Optimize handle a control page/group?
Webtrends Optimize sticks to a straightforward control group of 1 per Test. Other tools you use may suggest having 1 control group per variation, which is completely unnecessary and a waste of traffic.
In our UIs, you’ll find a means of controlling the traffic distribution between the control page and variations, if you decide to run with an uneven distribution like the one detailed above.
In Webtrends Optimize, if you want to serve content to people with no gauge at all of the control group, you can run a Target which has no control group. This is more often used for personalising content, where you’d like to make changes for specific users.
