Do you want your website to bring in more customers? Cut down support costs? Reduce the number of phone calls you have to deal with? Support the sales process? Whatever you do online (and, in fact, across the whole business), A/B testing is your secret weapon.

I recently spoke at Web Expo Guildford about ‘The Joy of Text – Writing Great Web Copy’ and as I got ready for my talk, I kept coming back to the thought that A/B testing (also known as split testing) is art the heart of it.

I’m sure it’s not new to a lot of people, and of course, it’s the core of the scientific method, but I got my first insight into it from Eric Ries’s book The Lean Startup. The idea, as my friend Phil Draper puts it, is “to have a website driven by data not by intuition”.

Here’s how it works:

  • You set up experiments, either using Google Content Experiments, Optimizely etc
  • You send some of your website visitors to Page Variant A and some to Page Variant B
  • You define a goal, e.g. the number of customers who actually buy something or who go to the contact page or whatever outcome is valuable and measurable
  • You then see which version generates the best results
  • After a reasonable run, you see the winner, delete the user and try a new experiment with a new variation.

Typical changes to test might include: length of text, use and placement of images, different types of headlines, position and format of the ‘call to action’ etc.

For example, I ran an experiment on my site, Turbine, to see whether longer or shorter text was more effective at getting people to sign up for a free trial. The result wasn’t surprising – shorter text was more effective – but the scale of the improvement was astonishing. The conversion rate more than doubled. Twice as many customers! (See a full breakdown of the a/b testing results on my Bad Language blog.)

Once you start with A/B testing, it’s obvious that it applies to almost everything in business: DM and email wording, marketing campaigns, brochures, customer support etc. etc. If you can try different variants and measure the results, you can do split testing. And if you can do that, you can replace ‘gut instinct’ with valuable insights that can improve your business.