Split Testing 101

Split testing is an absolute must for all online businesses. If you're not doing it at the moment - you really should be thinking about it - and if you are doing it, the likelihood is that you should be doing it even more.

Without split testing, you are leaving massive amounts of money on the table every single day. I say this with complete certainty. It's not that you "might" be leaving table, you are "definitely" leaving money on the table. It sounds like a bold claim, but it's absolutely true.

Whether it's a single page sales-letter site, a mini site, an AdSense site, a blog, or a mammoth content site, there are always ways to tweak things and increase the percentage conversion rate if it's a product that is being sold, the sign-up rate if it's a membership site, or the number of click-throughs if it's a site that earns revenues from advertising.

So, what exactly is split testing?

Split testing is simply the process of testing out different versions of something to see which one performs best. It could be a direct mailing piece, a billboard advertisement, a TV commercial, a web page, an e-mail, or anything else where you want to find out the differing results based on one or more changes to the original.

From an Internet Marketer's perspective it could be two different web pages you are testing to see which design works best. It could be testing one headline on a sales letter against another, or finding out which of two or more banner adverts brings in the most orders.

If possible, split testing works better with a fairly large number of visitors. If you get very low traffic, your split testing numbers may be inaccurate, but split testing even with low traffic is usually better than not doing it at all, it can just often take longer to get a statistically valid result.

You can do split testing in various different ways.

If you're running pay-per-click adverts through Google AdWords, you'll find there's a split testing system built in! If you're split testing things like sales page copy, a form on a squeeze page, or the CTR (click through rate) of various placements of ads on your pages, then you'll need to develop your own methods.

There are many scripts and online services available that can help you deliver different versions of a web page to visitors, and then measure conversions from each version. If you don't want to use a script or online service (some of which can be quite expensive), you can do the same thing yourself by rotating pages on your own.

You'll probably want to leave a page up for about a week or longer and record conversions. Then add a different page a keep a record of the conversions from it. Just keep in mind, conversions may be different, this can simply be because of the certain time of day or the particular day of the week, so if you were to take a page down in the middle of the day, or in the middle of the week, your results may be inaccurate.

The importance of split testing can't be over-emphasized. Something as small as changing a word in your headline, or even changing the color of your order button could potentially increase your conversions by several percent.

There has been some stories of one single word doubling conversion rates of a product. Sometimes changing something seemingly very trivial, such as the color of the words 'Buy Now' on an order button can drastically increase conversions and sometimes more fundamental changes, such as using a different font for the body copy on your web site, can make an amazing difference.

It's worth testing almost all the elements on a website from page backgrounds through to different header graphics and different navigation menus. Some of your changes might consist of just minor tweaks whereas others might mean a major change in the overall style of your site. Pretty much anything and everything is worth testing to see if has an impact on conversions.

Don't always presumne you are correct, even the expert's can get it wrong.

A major offline advertising agency was working on an ad campaign for a major product. They developed several different versions of the ad, and then asked each of their ad executives to give their guess as to which one would perform better, and why.

After each executive had picked which they thought would be the winner they ran the complete series and gauged the results. The results proved a huge surprise. Of the dozen or so executives they asked, not a single one guessed the ad that actually performed best!

Even these high-paid ad executives with years of experience were unable to judge which ad would perform best. This is why split testing is so crucial. No matter how well you think you know your market, there will always be things you can't judge on your own.

Over time split testing will ensure that you choose the highest converting type of your site. This can prove to be a long process but it is definitely one that will bring you massive rewards.

Imagine you had a web site that was converting at the rate of 1%. If you did a series of ten tests one after the other that managed to increase the response by 20% each time, what do you think the final response rate would be.

It would be a staggering 6.2%.

This demonstrates the incredible power of split testing. Just by running a series of small tests that each had a minor impact, the response was multiplied by over 600%.

So, get testing, then test some more, and continue testing until you can't think of anything else that could possibly improve the response rate.


Copyright (c) 2008 Paul Smithson