OK. So you've written your sales letter and put it up on the web. Or it could be that you've done your testing the traditional way and physically mailed it to your list.

You've had the results and they're sort of "all right". But nothing to cause you to break open the champagne.

What occurs next? Should you leave this project to die a natural death? Go with it? Or you could tear your copy apart and rewrite some or all of it to improve your response?

The best answer is that you should improve your copy. Which is where split testing comes in. If all you do is change something without split testing, then you have no way of knowing whether or not your changes worked for you or against you.

If you've been involved in mail order, you will remember that split testing literally involved splitting your list in two. And you'd mail half your list one version of your letter and the other half another version. Being sure to only change one thing at a time, otherwise you won't know which "thing" worked or didn't work.

The headline should be the first thing you test. You may have what you think is the best headline in the world but until you've tested it against another headline, you won't know for sure.

The internet allows you to test things like your headline very quickly and simply. Upload two different copies of your sales pitch (probably called something imaginative like index1 and index2). Then measure which gets more orders over time.

As soon as you've determined the best version, repeat the process and test again. Either with another headline or with another variable completely. Maybe your guarantee policy. Maybe the color of your headline. Maybe the price you're charging for your product. And remember to test a higher price as well as your first instinct of cutting the price. Sometimes higher prices get the same - or even more - orders, or they may just earn you more money.

One way to speed up the testing is to use specialist testing software. Software will allow you to test a number of variables at once. It will then use some complicated math to rapidly decide which is the best version of your split test. (You don't need to know about the calculations the software runs). You can find this split testing software here.