********************************************************************************************************************* Get £50 off a website 10 point MOT. Use the Contact Us form on the top of this page for more details ******************************************************************************************************************

One thing that’s vital to the success of your internet business is conducting regular and thorough testing of your web site’s landing page. But, in order to do so, you need to be prepared to learn what might almost seem like a foreign language, such are the concepts and terminology associated with landing page testing.

And this is further complicated by the fact that, while the names of the concepts are the same, they may carry different meaning in some areas of the testing community.

What’s The Big Idea?

The goal of testing a landing page is to find out how visitors react to the page and the various elements on it. It starts with collecting and  summarising a small sample of data (known as descriptive statistics), and then predicting people’s behaviour once they land on the home page from the same traffic sources (known as inferential statistics). Ultimately, it’s all about finding the best possible landing page from all the various permutations available.

Elementary, Dear Tester!

Let’s look at the various elements involved in testing and their function.

1. Input and Output Variables

Testing consists of 2 basic factors – input variables (also referred to as “independent variables”) that can be controlled and manipulated by you, the tester; output variables (also referred to as “dependent variables”) that are used to measure and evaluate results.

2. Variable

Refers to a tuning element and can vary in coarseness or granularity from just a landing page heading to the re-design of an entire page. A single variable within a multivariate testing strategy is referred to as a factor.

Multivariate testing involves several variables so each needs to be identified separately. An example might be using a capital “V” for all the variables plus a number to identify each one uniquely.

So, say you’re testing a headline, sales copy and a button (the call-to-action), here’s how those variables might be designated:

• V1 = Page headline
• V2 = Page
• V3 = Call-to-action button text
• V4 = Call-to-action button colour

Variables don’t have to be associated with items in different locations. Take a look at the example above – the button has 2 variables associated with it, one for text and the other for colour. Nor do they have to refer to a specific location. For instance, you may want to check the effect changing the font on your landing page has, so that would affect the entire page rather than just one area.

3. Value

A value is the state of a variable at a given time, and, since, for multivariate testing, values are generally consistent across a given range, it makes it fairly easy to predict intermediate values when needed. So, if an engine produces 100 horsepower at 1000 rpm, and produces 200 horsepower at 2000 rpm, we can safely predict that at 1500 rpm, the engine produces 150 horsepower.

But, most landing page tuning involves variables with discreet values, such as our button example, in which the button may be assigned various colours, say green, blue and red. The distinction would be shown by assigning a different lower case letter for each colour to the variable name, with the letter “a” assigned to the original, baseline value, as follows:

• V4a = blue button (the original colour)
• V4b = red button
• V4c = green button

4. Branching Factor

This is the number of values a discreet variable may contain, which must be at least 2, the original value plus an alternative. In our button example, that variable has a branching factor of 3.

When it comes to traditional multivariate testing, the number of values a variable may hold is known as the level of the factor.

5. Recipe

A recipe is a unique set of values assigned to variables for a given test. Given our example above, let’s say the variables are assigned the following values for a landing page test: V1a, V2c, V3b, V4b. The recipe in this case would be:

acbb

All recipes are unique, with the initial baseline test being assigned all a’s and all subsequent testing compared to this.

6. Search Space Size

Calculated by multiplying all the branching factors of the variables that are part of your testing, this represents how many unique recipes make up your testing.

Using the variables mentioned earlier, let’s say we have 2 headlines, 3 different versions of the sales copy, 5 calls to action and 3 button colours. Here’s how the search space would be calculated, with BF being the branching factor:

– BFV1 X BFV2 X BFV3 X BFV4

= 2 X 3 X 5 X 3

= 90

Obviously this is a very small size, and in reality this would normally be larger. But, too large a space may make testing very cumbersome and exceed what’s practical for tests such as standard A-B split testing.

With an understanding of testing terminology and methodology, you’ll be in a position to map out the testing strategy you need to take your internet business to the next level.

Contact a UK internet business coach for help defining and executing your landing page testing strategy.