Reflecting how users view your content

Companies often mirror the structure of their intranets and websites based on their organisational structure. This results in a very inward looking IA that only the company's employees will understand!

 

Card sorting is a generative method - i.e. it frames how the design takes into consideration the content users want. In an intranet, for example, this results in the greatest productivity gains as the structure reflects their employees' workflow.

The key steps in a card sorting exercise I carried out at Deloitte were:

1. All the classification terms gathered during the situation analysis were written on cards
2. A group of 30 people (split into three groups of 10) were selected from employees, content authors and knowledge managers
3. Each group was asked to go through these cards quickly, rather than thinking over them for a long time
4. They were then asked to choose the top 7-12 classifications that they would want to see at a top level and prioritise them based on the most important getting the highest mark and so on
5. The results were collated on a spreadsheet with all the classifications listed in alphabetical order
6. The respective marks of the participants, once entered and sorted with the highest mark first, shows how your users view your top-level classifications and how they should be ordered
7. This sample size I found was optimal as their feedback resulted in a correlation coefficient of between 0.90 and 0.95

The key questions to ask when carrying out this exercise are:

1. Is this classification in any way a duplicate of another classification already chosen?
2. While it's good to be brief, it's better to be clear. Does this classification have multiple meanings?
3. Is there a simpler word or phrase to use?
4. Is there a shorter word or phrase to use?
5. Does this classification follow web convention? Classifications such as Home, Contact, Help, Search, About, What's New, are what people are used to. Coming up with new names for the above will only confuse people
6. Is this classification roughly the same length as the other classifications at the same level? If 9 out of the 10 classifications at a particular level use one word, and one uses four words, it won't scan well
7. How does this benchmark with other intranets?

Where your audience is spread across cities and indeed countries, there are also a number of online card sorting applications that will replicate the above process.