Use the users

Wednesday, July 02, 2008

The greatest challenge with usability tests is to find time, money and not least users to test them on. Preferably without delaying the project. Furthermore experts must be employed and this takes time. Or are there other ways?

During the development of a new project or continued project most persons working with websites, either as consultants or responsible for it, have contemplated if the new ideas apply. Will the users think that the website is better and are the concepts we apply as clear for them as for us. Subsequently too little time and resources are allocated for testing before launching a website.

Usability tests expensive and time consuming

A contributing factor to websites not being testing in a larger scale is that testing is viewed as difficult and complicated to make. They are considered expensive and time consuming. Neither is it certain that there are resources allocated to amend what is considered a flaw during the tests.

Tests are made late

Tests are often made to late. The efforts to make changes are then frankly too large. Everyone who has done a usability test knows how hard it is to get a hold of representative users.

Must the users be representative?

What is a representative user? There is no underlying science behind it, and one test is actually better then no tests. A user, irrespective of profile, is better then no user!

Use the website's visitors

How hard can it actually be to find users that are typical for the website? A general corporate website for a larger Swedish company has 30-50 000 visitors in a month. Of these 20-30 000 UNIQUE visitors. That is substantially more than anyone has ever used in a usability study for their website! We can use them!

Make the usability tests on the basis of the website map

When you do a usability test – base it on the website map, and test the restructuring of the website before they get hits from the navigation.
The website map is traditionally a copy of the navigation structure portrayed as a hierarchal tree. It is a relatively small risk taken if we work with the different structures in the website map as it is merely a fraction of the users that enter here. The advantage of this is that we receive a limited group that we can study in order to see how the changes in the structure affect where the visitor clicks.

Example - Investor Relations and Media joined together

The content under the heading Investor Relations is sometimes also available under Media and vice versa. A solution can be to join the two into one section on the website in order to find out:

• Does the user understand this?
• Are there too many navigation choices?
• Does it confuse the users (journalists, investors amongst others)? 
• Do the users find the annual reports? Press releases?

Through measuring the usability and click frequency through the website map we can determine how well our thought through change will be perceived by the rest of the users.

Example - Different terms on content

We can also experiment with different terms on content on the website. For example income statements and balance sheets could be named ‘financial tables’, ‘financials’, ‘tables’ or ‘economical tables’. In the statistics we can quickly determine which terms have the highest number of hits. When we are satisfied we carry out the changes in the ordinary navigation. And measure it there as well.

Example - Content of recruitment

Information directed towards potential employees – students, jobseekers amongst others – is becoming more important as the competition for workforce increases. What makes the users continue to click? What are they looking for?

Should the section be called CV bank? Or ‘Career’, ‘HR’ och just ‘Jobs’? The questions are often many and the decision is more or less based on a gut feeling. Or even worse – the name of the section is maybe based on the name of the department which handles the questions at the company.

Once again – just change the website map. Measure. Change again. Measure. Change again. Measure.

Finally we will know, based on these usability studies, what is the clearest for the users. In the final stage this is the most important, as structure and navigation is all about leading the users to the right content.

What is needed in making a usability test?

In order to perform a usability test you primarily need a CM system which makes it reasonably easy to create and change manual versions of the website map.

Secondly a well functioning statistical tool is needed which makes it easy to follow up how the users click on the website map within different time frames. Here, it is not sufficient to use a free tool on the market but instead a bit more is needed. At H&H we mainly work with Omniture’s products, however WebTrends and ClickTracks give corresponding opportunities.

Why should I do usability tests?

One of the most important functions for the navigation structure is to lead the users to the information. In order to do this concepts and tags must be used which the user associates with the content rather than the wording you and I would choose. Our choices are coloured by our knowledge of the company, organisation, persons in the organisation and the fact that we work with this and therefore are not typical users. This means that we need to test and verify if we have made the correct choices. Frequently you receive an eye opener of how the user does and doesn’t.

So – don’t forget that you have an inexhaustible source to use for your usability tests!

Different ways of testing

There are many different ways of testing. Following is a list of some of the most commonly used test:

A/B tests

A/B tests applies two different solutions which alternates at random on the website. The statistics show which of the solutions that most effectively achieves the desired result; for example that the user clicks to an underlying page.

A/B testing is a very effective way of working with continued improvements of the website. Each new solution that is added lies parallel with the old solution for a shorter period of time (1 week is sufficient). If the new solution is not better you need to return to the drawing table. If the new one is better then it can take over. What mainly is needed is a good statistical program and a some time.

More advanced solutions go under the name ‘Multi variate testing’ where a number of variables are involved. It requires technical support from for example Omniture or Mazymyzer.

Grab your neighbour

'Grab your neighbour' testing as defined by Steve Krug is based on the thought that is it better to test a little than not test at all. Testing is rarely scientific and can subsequently allow for larger opportunities, so it is frankly simpler and cheaper to ask others i.e. ‘grab your neighbour’. This is also the source of inspiration for this article.

Steve Krug’s book is called Don’t make me think and you can read more about it at his website http://www.stevekrug.com/

User testing

User testing occurs in a lab or at a user where you can observe and record the user’s activities. User testing usually required more input and preparations. You most commonly use software such as Morae (www.techsmith.com) or similar products. According to Jakob Nielsen  (www.nngroup.com) you can use less than 10 users in order to receive an acceptable level of collected data.

Eye tracking

Eye tracking is an advanced analysis of the users where you follow how the eye watches the screen. Eye tracking is relatively expensive and gives the best result when other tests have already been made and the solution has been adjusted with simple methods.