A/B testing and combination testing can be used to dramatically increase conversions on an e-commerce site. After going through this process recently on a high-turnover e-commerce site I reveal some of the things I've learned.
Opinion: why is the Macbook Air a full 300 pounds (600 dollars) more expensive than in the US?
Introduction to A/B testing
If you've ever used a Google pay per click campaign linked with a Google Analytics account you'll have some idea of what A/B testing and combination testing are. Google allows you to run various versions of your advert and see which ones result in better click through rates. You can follow this through to your site and have different landing pages etc, Google analytics can track this and tell you which landing page combinations convert best.
In this article I'm going to talk specifically about A/B testing in general as it applies to websites rather than the more complex combination testing methods.
What's my methodology?
The A/B testing methodology is fairly simple:
Take one website, make an identical version and send a percentage (say 50/50) of your visitors to each version.
Pick something on the site that you think might be a bar to conversion and change it in one version of the site (the B version).
Run both sites in parallel for a week or two and then check your conversion stats for both.
If version B converts better than the original version A then roll that change out across both sites, start again with something else.
If version B does NOT convert better then roll back the change and pick something else to try.
In terms of stats, you are most interested in the number of visits, the number of product views, the number of baskets created and the number of baskets converted to confirmed orders.
What should I test?
So now we've got a methodology what should we actually test? I would start small initially, especially if your site is already performing well. If your site is not performing or you're about to go to launch then you've got nothing to loose so aim big.
Key areas of any e-commerce site include navigation (both primary and secondary), product page, category page, basket page, checkout process, oh and don't forget the homepage.
In each area there are number of aspects you can test. For example, on the product page you can test general layout, size and number of product images, the position, size and colour of the price line and 'add to cart' buttons, the amount of product description and features you show, etc, etc.
Draw up a list of things to test, you can combine several as long as they're not all major changes and they're in the same area. Its pretty crucial with A/B testing that you test one version for a set time before making more changes. If you need to make a change to one version mid test then you should make it to both if applicable.
Once the test time period is up you should gather your stats, decide which was the more successful version and then consolidate your site before you begin again.
How should I track the results
Once you've got some results you should do some stats analysis. Obviously if you are splitting your traffic 50/50 then conversions is king, but try and look deeper into the figures to see if something more subtle is going on. Obviously you need to be stamping every session, basket and order with a site version to accurately track what going on.
Look out for bottlenecks, e.g. if version B is giving you many more product views and baskets but you're not seeing an increase in baskets to orders then there is probably something badly wrong in your checkout process which is where you should look next.
Conclusion
Most of A/B testing is common sense and if done properly it can give you a nice boost to conversions without burning large amount of development time. If you're feeling brave you can also look at combination testing which enables you to track multiple versions of the site in parallel. I would use A/B testing for testing large changes and combination testing for fine tuning conversions, testing the success of click campaign versions and testing new site functionality before a full role out.
As always, if you've got anything to add or you would like to talk to me directly about any of this then please drop me a line!
