Gallery

Guided Miss-Aisle

Posted on 30 Mar 2009 at 15:09

Kevin Partner is on a mission to find a good guide to website development, and suggests some useful tools for online businesses.

To take an example of a split test I'm running right now, my first ad has a click-through of 7.32% and conversion rate of 3.0%, whereas my second ad has 6.48% and 3.4% respectively. That's a bit of a tricky one to interpret, since the click-through of the second ad was 11% lower while the conversion rate was 13% higher - the result is fewer sales but at a slightly higher margin. In this case, I'll base a new experiment on that second ad to see whether I can increase the click-through without sacrificing conversion.

That's how to work with the ads themselves, but the next link in the chain is the landing page that visitors are directed to. Google's Website Optimiser is one of several services that allows you to split-test various aspects of your landing page (or indeed any page on your site). For example, Website Optimiser allows you to set up an experiment in which you specify parts of your page that you want to vary. It's best to stick to one or two to begin with, otherwise you'll struggle to make sense of the results and they'll take longer to come in. You need to insert some JavaScript around those parts of your page that will change, and then use a web interface to specify their alternate contents, which could be text or pictures. In my case, I'm currently testing two headlines and two bits of content, giving a total of four combinations, one of which is the original.

Finally, you need to tell Google precisely what action represents a "conversion". When a user arrives at the page, they're served up one of the four variants randomly and Google keeps track of what action they took, and can tell you which variant is the most successful and if it's better than the original. The only downside is that you need to achieve around 200 conversions per variation (in my case, 800 in total), which even for successful businesses can take some time. This is another reason to limit the variables in any experiment. If there's a single "high confidence" winner then you should implement that variation and then create a new experiment to test against it. Just as with AdWords this is a process that's pretty well never-ending, but is essential to marketing success.

Watching your visitors

One ability that I've often wished for is to see exactly what my visitors are doing on a site, which would help me to understand what parts of a page they focus on and why some people don't complete a purchase. Now an excellent new service called ClickTale (www.clicktale.com) allows me to do just that. ClickTale's most eye-catching feature is that it tracks user activity so closely it can generate what is literally a video of their on-site activity. You can see where they paused, what they clicked and which bits of the site they bypassed. ClickTale's summary report will also tell you exactly what each visitor typed into their search engine to find the page in the first place, which is useful in fine-tuning your AdWords.

ClickTale also offers a "heatmap" of user activity, which is a screenshot of the page with false colours superimposed over it, to represent how long users, on average, spend on each part of the page. This service has already provided crucial information for me when it comes to my landing page: not surprisingly, the top of the page received the most attention but I wasn't expecting to find that the bottom of the page got almost exactly as much - I'll remember that for my next Website Optimiser experiment. ClickTale includes a usable free membership, and while it's relatively expensive at around £75 per month, it's such a powerful tool that it should pay for itself.