How much traffic would you send to a web landing page as the absolute minimum in order to make a decision on which test is best? I want to start split testing landing pages. — Carter Maya, Dallas, TX.

Split testing landing pages is where you share traffic between two variations of the same page. Each page tests for specific copywriting changes or layouts with the desired result. This is a great way to see what works for your business website.

The more traffic you can get, the better your test; however, too much traffic and you may lose the benefit of such testing by missing visitors actions. Two questions I frequently get:

• How much traffic do you need to have a valid test?
• How can you make traffic testing more valid?
• How long should I run a test, and does it make a difference?

The traffic you need depends on how accurate you want to be. I'm doing A/B splits on landing pages and usually look for 1,500 unique visitors per side and 100 leads to stop a test.

For sales, I look for which page hits break even first but often has a great back end, so I use what I learn to add third-weighted pages. Still a minimum of 1,500 unique visitors (not page views.)

Statistically, you need between 1,500 and 2,500 unique individuals to achieve 80% accuracy in your results. However, traffic quality, source, and many other factors contribute to the results you get.

Because traffic quality matters, I usually test with pay-per-click traffic or a fixed set of marketing campaigns. This way, I've driven most visitors to the site with my lists.

If you see a strange drop-off in your testing, review traffic logs, looking for changes in the source. You may need to extend a test if you have a dramatic change or shorten the test if the quality looks strong.

Very often, instead of just stopping the test, I'll shift traffic based on the source because some traffic converts better on one page than the other. This requires some technical support but better tunes your message into those who want to take action.

Ultimately, the only thing that tells you the success of your testing is how those leads or sales do as customers. This is why source tracking is important. I've found some landing pages that work better in the summer than in the winter, and this information is useful, too.

When I work with clients, I need them to have a high visitor count of at least 500 unique visitors a day. Anything less than that makes testing take forever. While everyone wants a rule of thumb in this area, your results will vary.

© 2008 B2B Website Profits, All rights reserved.