I know how it feels. You’ve spent hours putting your site together, carefully following all the advice you’ve been reading. If Oscars were awarded for website design, creativity and wow factor you’d have a string of them. So why isn’t your webpage converting?

Sometimes it’s obvious why things don’t work. If the page has been built for list building purposes, but there’s no optin form on it, it’s just not going to be successful. But other times the reason’s not so obvious. Fear not. It is possible to determine what’s holding your site back. Consequently you’re able to focus on the areas that need to be addressed to transform your webpage into the winner you always knew it could be. You simply need to start practising split testing.

Split what?

What is Split Testing?

Split testing is the comparison of the results of a control sample against one or more variations. It sounds boring, dry and technical, which I suspect is why many people don’t bother doing it. That’s a shame because not only can it transform your results, but it’s also bloody easy!

It helps you to produce more effective websites which optimize your goals. This obviously means you not only need to know what your websites goals are, but they also need to be measurable. Thus while split testing is easy, you do need to put a bit of thought into it:-)

Typically, when you start testing, a website or squeeze page would be tested, but the same technique can be applied to other things such as emails and banner ads. The most commonly tested elements include,

• Text
• Colour
• Fonts
• Layout
• Headlines
• Sub Headlines
• Opening Paragraphs
• Call to Action
• Offers
• Hyperlinks

Essentially you create your webpage, which is then referred to as the control page. Copy it and make a slight change to each copy. Load the pages into your split testing software program, and the software will rotate the display of your pages while your test is running. So, for example, if you produced three pages, and assuming you want to display your pages in equal proportions, your first visitor will be shown page A, your second page B, your third page C, your fourth page A and so on until the test ends.

For your results to be meaningful you need to have a certain amount of traffic. The more traffic you have, the more confidence you can have that if one variation proves more successful during the test, it will also prove to be more successful over the long term. If we go back to the “boring speak” this means your result has to be statistically significant. However, let’s be honest, we don’t want boring speak, we just want plain and simple, so we can implement, draw conclusions, take action and move on.

In his book, Landing Page Optimization, Tim Ash puts forward a general rule of thumb for how large the difference between variations should be for it to be meaningful, based on different traffic volumes. He argues that for a sample of 100 impressions, the difference should be at least 20%,

• For 1,000 impressions the difference should be a least 6.3%
• For 10,000 impressions the difference should be a least 2.0%
• For 100,000 impressions the difference should be a least 0.063%

However, as I’ve already mentioned, this is actually a statistical calculation, so as you may have guessed, there are websites that can do this for us. Just pop over to a site such as SplitTester, type in your results, and it will let you know how confident you can be that a change will generate long term benefits. Fantastic!

Hopefully you can now see that split testing is simply marketing the same thing in different ways. Once you’ve found that one variation of an element outperforms another, you could either test the best performer against another variation or start to test another element. Thus if you were testing a sales page and the “Buy Now” button was performing better than the “Add To Cart” button, you might want to try “Buy Now” against “Add To Basket”. Alternatively you might be happy with the results of your test, and move on to testing the color of the button. Or perhaps you might want to test the headline. The options are limitless, so you have to be careful that the cost of testing doesn’t exceed the benefit you’re likely to get from it. But hopefully you can see that by testing you can significantly improve your websites performance, and that those who don’t test are potentially leaving money on the table.

Testing Methodologies

There are two different types of test that can be performed – AB tests and multivariate tests (also known as multivariable testing). AB testing involves any number of pages being shown in a given test, but each page can only have one difference between itself and the control page. Essentially it’s what’s been discussed above.

Multivariate testing only shows one page, but numerous elements of the page can be changed. Thus, on one page you could test different,

• Headlines
• Optin Forms
• Buy now Buttons
• Images
• Layouts

The advantage of multivariate testing is that it is often quicker to determine what is successful as all elements are being tested at the same time. The disadvantage is that because more elements are being tested, more traffic is required before the results become meaningful.

Testing Software

Split testing software splits the traffic to a page so that all the versions being tested receive either an equal amount of the traffic, or the amount specified when the test is created. The results are tracked and presented to the user in a report format.

Until recently Google Website Optimizer was a free tool allowing you to conduct both AB and multivariate tests. Unfortunately it’s no longer available. AB testing is now part of Google Analytics and the video below briefly shows you how it works.

Obviously Google aren’t going to mention the downside to their product, so these guys have highlighted a few of the shortcomings. But be warned, they’ve got an alternative product they’d love to sell you! Take a read and draw your own conclusions.

I’ve yet to hear about Googles intentions for multivariate testing and haven’t been able to find a free alternative. If you know of one, please drop me a line in the comments.

As with pretty much everything else, paid tools are available for both AB and multivariate testing. I haven’t used any myself as I’ve only done AB testing via Google, so I’m not in a position to recommend anything. However, if you’re looking for multivariate software this comparison chart might be a good place to start your research.

Is It Worth Testing?

It’s no good ignoring it, the fact is that some designs, text, colors, fonts etc. do a better job at getting prospective clients to take action than others. Not knowing how well your site is performing, and how well optimized it is, could be costing you big time.

Split testing takes the guess work out of your sites performance. Stories abound of companies who spend thousands on having websites created only to witness no improvement in business. Analytics are not installed, tests are not performed, and the owners are left tearing their hair out wondering why their site “doesn’t work” and frustrated by the lack of return on their investment.

Split testing allows you to identify what isn’t working and change it until you find something that does. The results on your profitability can be huge. Some success stories have seen improvements of 10% on conversions.

To put that in perspective, a company with a pre testing conversion rate of 1% would need 100 visitors to achieve a sale. To achieve more sales they would need to drive more traffic to their site. This would probably incur a significant cost, either in terms of buying traffic or paying for someone to generate organic traffic.

If post testing they achieved a conversion rate of 10%, the same 100 visitors would generate 10 sales. Or put another way, ten times more revenue, at no additional cost:-)

If you’d like to read some real life case studies, check out these posts:

Increase Your Conversions With These A/B Testing Case Studies

A/B Split Testing

When you split test, you don’t do what you think will lead to conversions, but what you know will produce results. With free software available, tests easy to create and the rewards potentially significant, it’s got to be a bit of a no brainer. What do you think?

Over To You: Do you test your sites performance? Perhaps you could share some of your success stories? What software do you use? Or perhaps you think its all a waste of time.

Comments welcome.