What Is A|B Split Testing?
A|B Split Testing is when customers receive a message, email, or webpage view, and a different version is randomly served to different visitors or customers. Typically A|B split testing consists of two nearly-identical versions of the given source material, but with each version having one element that is different.
The two “versions” are randomly served, and tracking scripts are then used to monitor the resulting actions to determine which version is most effective in generating the desired reaction.
What Are The Benefits Of A|B Split Testing?
A|B split testing allows owners, consultants, and marketers to better understand customer habits and to figure out what best motivates them toward the desired action. A|B split testing can be used to test different web layouts, sales copy, price points, graphics, or just about any other factor that might influence the visitor or customer.
How Can I Do A|B Split Testing?
A|B split testing can be conducted a number of different ways. Webmasters will often use web scripts to randomly serve different versions of web pages. Alternately, there are third-party services available that webmasters can use to conduct split testing on a large scale.
How To Know If Testing Is Valid
In order to validate the results of A|B split testing, the sampling must be large enough to provide legitimate statistical information. If the sampling is too small, the data may be inaccurately skewed for reasons not associated with the factor being tested.
Only one factor should be tested during A|B split testing. In other words, do not try to test price points and web layout at the same time. It is important to only change one thing between the two versions of the test copy.
Keys To An Effective A|B Split Test
- Volume
The sampling of visitors tested must be significant enough that the statistical sample is not unfairly skewed. The larger the sampling tested, the more accurate the results will be.
- Single Testing
A|B testing allows you to test two items (such as 2 different versions of a landing page, etc) against each other in real-time. Test only a single issue, element, or item at a time. Attempting to test multiple items will only skew or invalidate the results.
- Randomness
The two versions of the element being tested should be evenly and randomly displayed to website visitors.
- Avoid Assumptions
What is intuitive to software marketers may not be what performs best. The only way to know how something performs is to thoroughly test it. Do not assume that just because something appears to be a good idea, that it is in fact a good idea. Support marketing decisions with split testing, so that you have measurable results. Statistics speak volumes.
- Things Change
Testing should be repeated from time to time, simply because what appeals to customers may change over time. Split testing is an ongoing process, not a one-shot deal.
Testing alone is not enough — marketers must evaluate and scrutinize their test results, refine the items they are testing, and continue to test.
Things to A|B Split Test
A|B split testing can be done for any number of different things. Webmasters, marketers, or publishers may want to test the effect of changes in the following areas…
- Headlines
Test two different headlines to see which is most effective in drawing attention to an item.
- Price Points
A|B split testing can be a very effective way to find the “sweet spot” for pricing. Pricing too high or too low can mean you are not maximizing profits. By testing different price points, you can learn what most customers are willing to pay for your software.
- Color Schemes
Color studies indicate that the use of certain colors can effect a person’s response to various things. For example, changing the color of a purchase button may increase the number of button clicks, and ultimately the number of closed sales.
- Link Properties
Change the link properties on a web page to see what format generates the most clicks. Try bolding or underlining links, and determine if there is a significant statistical difference in clicks.
- Advertising
Test different ad models, sizes, and banners on a web page to determine which is most effective in generating the desired response from prospective customers.
- Web Copy
Test different versions of web copy or copy formats to determine what sales message sells the most software.
- Email Campaigns
Setup an email campaigns with two different versions, and then split test the email using a sampling of subscribers. This will help you determine which version appeals to the most customers. Be sure to evaluate open rates, click-throughs, and conversion rates in order to determine how effective the various versions are.
- Positioning
The location of content on a website can make a difference in how visitors respond to it. Serve the same content in a different position on a web page to determine what locations are most effective.
- Size Matters
You can adjust font sizes and enlarge purchase buttons to determine if the size of the content has any impact on sales.
Statistics speak volumes. The fact is, marketers can test a multitude of factors such as copy, images and placement, length of copy, pricing, etc. While some details may seem mundane, without fully testing you could be missing out on a lot of sales.
