When it comes to landing pages you have a variety of types at your disposal. Depending on the original link source or ad, a landing page can conclude with a purchase (transactional), inform the reader (reference), or can collect reader information (opt-in or squeeze page). Regardless of the type of landing page, all three have the same basic main pieces. They are the offer, the pitch, and the call to action. Although all three are used in each type of landing page, this article will be geared more toward the most popular form of landing page in the Web Hosting industry and that is the transactional landing page.

Thus far you have learned all the basics necessary to build landing pages. In fact, after reading the first two articles in this series (The Pieces and Making it Fit) you can go on to create landing pages that will increase your conversion rates. However, the third step is what separates the good from the great and the advanced users from the experts. In this article, we will look at testing and refining your landing pages.

The Purpose of Testing

Landing pages serve a purpose. Each element found on a landing page should help in achieving that purpose. For instance, if a landing page is designed to move the visitor to purchase, then each element needs to aid in the selling.

With that in mind, testing looks at the impact each element has on the landing page's mission. Through different forms of testing, elements are replaced, tweaked, and eventually perfected to provide the maximum amount of conversions possible.

Testing also shows if a landing page that is optimal for one advertising campaign is suitable for another. This is rarely the case because although multiple sites can provide the same form of traffic (an audience that wants to be web hosting for example), it does not mean the same landing page will work for each site. One site might have a technical audience, another might have a business audience, and another might have those interested in dabbling in web hosting. All three sites have those willing to buy, but each will respond to a different message.

Testing can be as simple as swapping two different landing pages and seeing which one performs better. It can be as complex as pairing multiple landing pages testing multiple variables in comparison to the advertisement that brought the traffic in, in the first place.

A /B vs. A/B/C/...Z vs. Multivariate

There are two forms of testing, single variable and multivariate (multiple variables). Each form of testing has its pros and cons and in many cases a company will probably use different tests for different pages. Single variable is often referred by how many tests you are performing at once such as A/B/C or A/B/C/D. In each of these instances only one variable is tested, the letters represent how many tests are going on. Of those, A/B is by far the most common and has several advantages over the other forms.

A/B split testing requires the least amount of traffic and actionable clicks to perform its analysis. It is also uncomplicated. A single operator need only place a JavaScript function in the header to swap the element and provide a cookie (such as A cookie and B cookie). The cookie is dropped off if the reader performs the action indicated on the page. Although you can do this test manually (having two pages swapped for every visitor), using JavaScript can help avoid future search engine problems. The testing continues until there is a clear winner.

As an example, if you wanted to test the opening sentence as your element, Landing Page A would have the original sentence, Landing Page B would have the new sentence. When a visitor comes to the page, the header gives them a cookie and changes the element to reflect the cookie. After 1,000 total visitors, the testing shows that Landing Page B has a click through percentage of 3% while Landing Page A only has 1%. From that we can ascertain that Landing Page B is the clear winner.

As more testing pages are added, more visitors are required to determine a clear winner. This is fine if the page receives several thousand visitors a day, but if the page only receives ten or twenty a day, the testing will become bogged down. For those sites that can support thousands, multivariate testing might be a better option.

Although A/B/C/D/E/F type testing might be easier to control, it can miss important data that multivariate tests uncover. Therefore, if a landing page can support that sort of testing, it is best to go with a multivariate instead.

In a multivariate test you not only see gains for each singular variable change, but you can also see how the different variables work with each other. You may find that large text for the opening sentence does poorly on its own, but when coupled with a blue color layout does extremely well. In this manner, multivariate is not just testing variables, but the interdependencies of each variable on each other. Because of this, multivariate offers a more precise and accurate depiction as to the best elements for your landing page.

Multivariate testing can become quite intricate however. For instance, if you were to use split testing to test two variables (say color and title) with three different types you would do an A/B/C test one week (for the color) and another A/B/C test the next (for the title), six landing pages in total. If you were to do the same test for multivariate you would require 9 total pages because you are not just testing the color and titles, but you would test each title with a different color scheme as well.

Why Use Third Party Software?

If you are not using third party software, the accepted method is to use JavaScript early in the landing page. The script would load your pages in alternate succession and use a cookie to log the visitor and the page version. If the visitor converts the cookie will be logged during payment processing. So from this simple method (coupled with some form of traffic analytics... which you should have running already), you can see how many people visit each landing page (which should be even, give or take) and how many convert on each page.

Although software is not technically necessary in the above scenario, it makes your life a lot easier. If your landing page has some 30 elements to test you are going to be testing for a really long time. Moreover, if you wanted to test using a multivariate method (which will increase the efficiency of the process) you are in for a real nasty ride trying to keep all the data organized and ready to use.

Not to mention the statistical math involved. If you want your data collection to be accurate you will have to calculate things such as proper sample size, confidence indexes, and the like. These are not difficult if you know the equations, and various spreadsheet software like Excel can greatly help in this, but with everything you have to do each day is it worth spending five or six hours a day? If there was no third party software around ready for use I would say it's worth it, but the fact of the matter is there are tons of software packages that can automate a great deal of the entire process.

Many software packages also allow the weighting of landing pages. So if you already have a good landing page in place and do not want to lose too much business as you test, you can set that page to take up 60% of the visiting traffic for instance, while testing only takes up the other 40%.

A few things you should definitely look for when it comes to third party software is: statistical significance (such as confidence level, margin of error, statistical validity), page element rotation, multivariate and split testing (should be allowed to have at least A/B/C/D testing), multiple campaigns, action tracking, real time reporting, rotation weighting, and repeat visit statistics are all must haves.

What to Test?

Every element is up for grabs. Headers, logo, artwork, body text, hyperlinks, opening sentence, closing sentence, calls for action, etc. can all be tested for improved conversion rates. Attention should also be paid to the layout as a whole. In this manner, color, layout of elements, size of elements, the look and feel of elements, etc. should also be tested.

Testing an element need not include huge, sweeping changes. Something as small as making a word bold or italicizing a header could end up changing your conversion. With the hundreds of elements and choices per element, there can be millions of different areas to test. So where does one start? The rule of thumb is, start big then narrow the focus.

The very first thing to test is the page as a whole. When it comes to selling a product for instance, you can start with a product page, comparison page, buy now page, etc. Once you figure out which page does the best, test the high impact elements. For instance, on a comparison page, is it best to have the buy button at the beginning or the end? Should you compare four products or three?

The testing your landing pages is a continuous process. It is one of the few processes that can directly affect your income and as such should be treated with as a top priority.