Everyone who’s anyone in the internet industry will tell you that if you want to improve your website’s conversion rate, you should be AB testing.
Just in case you hadn’t heard of it (or indeed have, but not actually had time to work out what it was), AB testing is the practice of creating two different versions of a webpage (or a whole website) and sending a representative proportion of your traffic to each one. You can then see statistically which one performs best (rather than just picking the version that your boss likes most).
AB testing is great: there are a whole host of tools out there that can help you do it easily, and it can be a really effective way of driving up the performance of your website – with you safe in the knowledge that the changes that you’re making aren’t making the site worse. There’s just one problem: statistical significance.
Statistical Significance is the mathematical test that you need to have passed to make sure that your results aren’t a fluke – that A really is better than B.
Its a little complicated, but you need to get to a statistical significance of 95% or more before you can really trust in any results you’re evaluating and the only way to do that is to get users through your site (as a rule of thumb, you need about 200 to convert to get to that point). This is fine if you have loads of traffic (it might take a matter of a day or two) but, if you’re a niche site, getting the results of just one test can take forever.
So, if your traffic volumes aren’t high enough to make AB testing practical, how can you improve your conversion rate? After all, just because your volumes are low, doesn’t mean that your revenues from online aren’t worth focusing on.
Luckily there are a load of other quick and easy tricks you can rely on:
- Splat the bugs. Your site almost certainly has some bugs in it. Individually, they’re probably not all that big, just little things that don’t work right, but collectively they add up. They create a feeling that your website isn’t all that slick, and they hinder users doing what you want. Take some time to iron our the creases, get the site working well, and you will see the benefits.
- Speed up. While you’re at it, check your site performance – use tools like YSlow to see how fast your site loads and follow the steps to speed it up. Users hate waiting – one recent study found that if your site takes more than 4 seconds to load, 40% of potential visitors will have left.
- Check your heuristics: Heuristics is the fancy term for best practice. There are a number of established guidelines that your site should probably be following if its going to win over visitors. Take a critical look at your website and see if there are some areas that could be improved.
- Do some user testing. You get familiar with your site very quickly: you know it’s layout, understand what its trying to say, and are used to it’s’ quirks. As a result, you’re blind to many of its faults. Invest in some user testing and see what visitors actually think about your site. It doesn’t have to be expensive and I’ve never yet encountered user testing results that don’t include one “aha!” moment.
- Spy on your visitors. Ethics aside for a minute, there are some clever tools available that allow you to replay the sessions that your users have carried out on your website. Products like Sessioncam or, if you’re a larger business, UserReplay allow you to see where users went, what they saw, and often what they typed. Again, its a rich source of real-world insight into how people actually see your website. For those on a real budget, click tracking tools such as Crazy Egg provide quick and cheap insights on where users are focusing within your page.
- Make BIG changes so you can see the difference. Now, I know I said that statistical significance means you need lots of users to understand how different versions of your site perform, but there is a caveat. The beauty of statistical significance is the bigger the variation in the performance of each version, the less users you need to be confident that the results are sound. If you’re making little changes to a page, then chances are the results will be pretty similar and that it will therefore take a while, but if you’re bold and try fundamentally different alternatives, you might be rewarded with a big enough gain that the results are reliable much more quickly.
Have we missed anything? Let us know how you’ve got on optimising your websites and if there are any tricks or techniques you’ve found work well.
p.s. You’ve no idea how difficult finding a link that simply explained Statistical Significance! Thanks Cennydd.