Understanding A/B Testing

If any of you have ever been to South Florida then you know that the panhandlers at major intersections are as much a fixture as the palm trees.

Yesterday as I pulled up to a red light I saw a man standing at the intersection with a large sign asking for assistance spelling out all the troubles he was having. Lying next to his belongings was a large cup and a smaller sign that simply read, “Need Help.”

When I rolled down the window and asked him why he used two different signs his answer was direct enough for me to skip the change cup and reach for some paper.

“The other sign (the one that simply read need help) works better in the morning ‘cause people are too tired to read.”

That explains A/B testing in its most basic form.  Find what works and when it works best and use it to be productive.

But A/B testing isn’t always that simple. You have to put a bit more effort into it.

When creating a test, knowing some of the best practices to apply can help you learn as much as possible from the results. To get the best possible results try these tips:

Business of Software - Derek Sivers

(Courtesy of Flickr User: betsyweber)

Know what you are testing

In our example above, the test was essentially which sign would bring in more money.

Calls to action are frequently tested. The location, wording, color and size are common variables that people use to see what gets the desired results.

Content is another subject of testing that is frequently measured. How long should it be, how well does it rank in the SERPs, the placement of images, should it be video or audio, etc.

The important thing is to establish what you are going to measure and build your test around it.

Make it consistent

If you are testing the placement of a Buy it Now button then everything except the placement of the button needs to remain consistent.  Varying the other elements of the button leads to skewed results.

 A good rule of thumb to follow is to test one variable per page only.

Understand that not all tests will yield results

It is quite common for an A/B test to show little to no significant differences between the variables. This simply means something else needs to be modified and tested until you find the right mix. And often times, it is the smallest change that produces the greatest results.

If a test of image placement isn’t yielding any results, try testing varied headlines or even content length. Lack of results doesn’t mean that you have failed; it means you haven’t found the right solution, yet.

Start small

A/B testing can be a pretty exciting task, especially for the more creative types so it’s easy to get caught up in project creep.

Fight off the urge to jump right into making drastic changes until you really understand the nuances of what to test and how to create one the right way. Once you build up some experience you can start moving to more complex experiments.

Test Infographics timeline NIT

(Courtesy of Flickr User: ciron810)

Test, test again, test some more

The key to successful testing is to always be doing it. If you have found the golden ticket for your landing page move on to another page on your site or even a different section of that particular page to see if you can get even better results.

The important thing to remember is that there is always something you can change for the better. In today’s climate of dynamic web sites there is always a steady stream of content, pages and activity to test.

  • http://www.netchunks.com Shiva @ Blogging Tips

    I think A/B testing is really necessary if you have a sales page/squeeze page or a product to sell because it is really important to know what type of page brings in the best conversions. Thanks for this informative post about A/B testing.

  • James

    Any software recommendations to do this? I’m in charge of our company site and been looking at different solutions. I see Google Analytics does it but I’m concerned it might use the data to influence our search engine positions.

  • http://www.spinxwebdesign.com/website-development-company-los-angeles/ Web Development Company

    Now Google is also providing this is type of testing for website.

  • Ruben

    Thanks for appreciating my post! :)