INTRODUCTION OF THE AUTHOR
Uroš Čimžar is the director of Web agency Klaro, which developed from a company established by him and his classmates during their studies at the Faculty of Computer and Information science. Uroš is an expert in information architecture; he studies the different aspects of the web. He always follows technological and creative innovations offered by the internet.

Designing websites should not be left to coincidence. When faced with this task we often start with goals the page or website needs to achieve. When these are set we need to think about the means of achieving them. This is where things get interesting. There is rarely only one way of achieving a certain goal. When we start designing a website at Klaro agency, we establish a project team comprised of individuals with different skills and outlook. Everyone contributes their ideas. Certainly the number of possible solutions gets reduced through discussion and previous experience, yet three or four feasible solutions for one goal can remain. So, which one to choose? Before tempers start getting out of control we can suggest a different method of choosing: testing.

What to test?

Theoretically all of the elements of a website could be improved by testing: content organization or information architecture, design, texts, applications and functions and the combination of all of the above. Testing is most common by landing pages, sites selling products or services. Visitors are invited to them through different means – online advertisements, website optimisation et cetera. All of the above requires sizable investment so it matters how successful the landing page is in turning visitors into buyers. Sometimes even the slightest change in design, arrangement of elements on the site or in text can dramatically influence the sales results. We have seen cases where sales went up 40% because of small changes, which is a huge effect.
In practise testing all the aspects of a website would be too time-consuming and costly, so we have to make a few decisions at the start. Firstly, we have to decide which goals are important enough to require testing. We select an appropriate method of testing and evaluate its precision. When planning a test we mind the costs: if we are testing a site which sells individual products, we start the test with one product and then apply the most successful solution to all others. If these are similar products with the same or similar key audience, we will probably not be wrong.

How to test?

Choosing a method of testing depends on the goal we are following and partly also on tools at our disposal. If our task is to select the form and design of a landing page, which will lead the visitors to buying the product, we can do the following: we design two different landing pages and publish them online. The sites differ in one crucial property, such as: colour, images, parts of the text or location of elements. Visitors are brought to both pages with the help of ads on Google. The ad for both sites is identical. After the test, we measure the result in the form of submitted orders or buys. With analytical tools such as Google Website Optimizer, Google Analytics and CrazyEgg, we are able to monitor the behaviour of visitors on both pages – where do they click, how long do they stay on the page et cetera. We make sure the same person does not come across both variations of the page. This can be done by using cookies saved by the browser, which lead the user to the same page each time. This is how we decide on one variations of the site. The chosen site probably also has room for improvement. So we set up another version of the page, let’s call it the challenger, which is different from the chosen one in one detail. We repeat the process of testing and keep the winner. This cycle is repeated until our landing page reaches the goal we set at the beginning in content and if possible in numbers.
The described method is called the A – B or bi-variation testing. Multi-variation testing also exists, where a several properties are tested at the same time.

What can help us?

We have mentioned the analytic tools from Google Analytics and CrazyEgg. Both are accessible on the web for free and require basic knowledge. Google Analytics is an extensive system for tracking website visits, which tell us how visitors come to our site, how long they stay there and what they do. The tool can be modified to fit our needs and set other criteria of measurement. Google Analytics is activated by including a tracking code in the code of the website, which means we require the proper access.
CrazyEgg also offers different types of data and it’s advantage is the clear display of clicks on each individual site. It’s simple to use and in our experience the data is reliable.
If we do testing we quickly get interested in what the main attraction of our site is. To this end we could carry out complicated and expensive tests, where eye movements of an individual are tracked with a special camera. Yet if we test different sites at the same time and require quick results we can use a clever shortcut. AttentionWizard is a simulator of camera testing which uses a large amount of data to assess how the glance of our visitors would move through the site and what would attract their attention.
The decision to test can bring numerous advantages. We just need to be careful not to get trapped in a vicious cycle of trying to find the perfect version – while our web presence is neglected and should be filled with a working website or an online store. The beauty of internet is that even the already published sites can be improved and adopted every day.