Multivariate Testing

Share on facebook
Share on twitter
Share on linkedin

I listened to a great presentation during Search Marketing World recently given by Russell Sutton of Conversion Works and his thoughts and insights on Multivariate Testing. This allows you to test the effectiveness of different page designs/layouts on traffic retention.

If some of your landing pages have high bounce rates then you could well do with looking for clues as to why the rates are so high. As Google point out, visitors will leave a web page within a couple of seconds if they don’t find a compelling reason to go deeper into your site. If that compulsion is lacking on your site then it could be contributing to the high bounce rate. If you have invested heavily in your website, SEO or indeed a PPC campaign, then these high bounce rates are very costly.

Whilst there is a huge emphasis on getting traffic to your site, there is not always the same emphasis placed on: firstly retention of that valuable traffic and secondly having that traffic convert. i.e. getting to site visitor to take a desired action which can vary from getting sales to simply downloading information. There are plenty of software packages out there to help you with your multivariate testing, but in Google Webmaster Tools there is a free tool just for this.

Multivariate Testing allows you to change particular elements of your landing pages and to see which of these changes makes improvements on your goals. Russell’s blog has some really great tips on landing page design and how to try and retain more visitors. Some of these tips are very obvious (sometimes embarrassingly so!) so they are well worth a look. I came across a good example of how multivariate testing achieved some great results for Laura Ashley.

The company noted from their analytics that their ‘shopping bag’ pages have relatively high attrition rates. Visitors reaching this page had the intention to purchase, having already selected goods to buy. But for some reason(s) many of the customers gave up the process at this point – a particular antagonising statistic for the management at Laura Ashley. The set up a multivariate test and made several variations of the shopping bag page changing some elements in each page (such as images, calls to action, amount of information on the page etc).

They left the original page as the default and as a benchmark. As customers reach the shopping bag page theses variations are rotated, each customer getting a different page including the original.

After a while when the system had enough data to make the experiment statistically significant the results were available for each page variation. The page variations worked either way, some showing a marked improvement and other variations showing a disimprovement on the benchmarked default page. Interestingly the best performing page showed an uplift of 11% in the number of visitors continuing to the checkout page (which was the experiments goal).

The page below on the top produced a 3.88% uplift and the page on the bottom performed nearly three times better. To a casual observer the differences are minimal the page on the top having the option to update the shopping bag and has some extra information about the product. The page on the bottom has the “continue shopping” more prominently in green but these minimal changes improve conversation three fold!

It’s a nice growth in business from existing traffic by simply making some alterations to your existing site.