Faversham Web Site A/B Testing & Web Site Multivariate Testing

Want to increase your conversion rate? Want to know what layout or content is best to mak eyou more money?

Web Site A/B Testing

Much has been written about A/B testing as a method of increasing conversion. But how useful is it?

A/B testing is a slightly scientific way of optimising the effectiveness of a website. It involves showing two groups of visitors to your site a different version of your page, to see which converts better.

When would we do this? Perhaps when you've broken away from a tried-and-tested page layout and want to try a brave new design. Using Aurora Optimiser, we can divide traffic between the old (A) and the new (B). We then observe their site visits for an agreed period and decide which page holds most conversion potential.

But an even more compelling reason is simply to eliminate the guesswork. How many times have you sat in a meeting in which everyone has an opinion about a web page, and yet nothing gets done? A/B testing makes everything so much easier, replacing subjectivity with science.

So, is everything in the A/B testing garden rosy? Well, not quite. A/B testing is slightly crude. It's hard to test every page element and we might have to do a few tests to find the winning element.

We'll also need to run tests over a decent period of time, to have any validity. It could take thousands of visits and hundreds of orders before your statistics have any real meaning. This can obviously take a long time if visitor numbers are low. We'll also have to factor in calendar variables, which can influence the outcomes.

Also, our A/B testing might show that our new page is a winner, but it's sometimes hard to see which page element has made the big difference. For example, we might have a new headline that's working wonders, but our new scrolling page layout could be a disaster. Our page could show better conversion, but we won't know to ditch the dodgy scrolling. This is where multivariate testing comes in. 

Web Site Multivariate Testing

We don't mind a bit of heavy-duty content development so we can test a handful of versions rather than just two. We then make a choice from the bigger group. Alternatively, we can put page elements to the test in various combinations. Welcome to multivariate testing.

With multivariate testing, we can choose which parts of your design or copy affect sales and then analyse them. Do it right and multivariate testing is certain to show an improvement on the previous iteration and will also provide clues into what converts. We can then put it to good use on your other pages.

How does it work? We could use multivariate testing to analyse three elements, say a new image, the body copy and the navigation. Together, they have to create an effect so compelling that visitors simply can't say no. This is particularly vital and an economic necessity if you've paid for those visitors through pay-per-click or banner adverts.

To measure our different versions, we'll create at least two alternatives of the image, copy and navigation and test each combination in each version. With three elements each with two alternatives, we'll need eight combinations (2 x 2 x 2).

We then divide your traffic randomly to find the perfect combination. The downside is that testing lots of elements or alternatives needs huge numbers of visitors to gather worthwhile data.

On the upside, we'll always be able to improve on current conversions.

Recent Work

View Portfolio

View Portfolio

Related News and Articles