Having been approached by a large national charity to develop a new information architecture for their website, we were about to suggest a traditional user-centred approach - open card sorting followed by iterative testing of a prototype - when we learnt that the client had already experienced a card sorting exercise, along with usability testing, carried out by another agency.
The client felt the card sorting exercise was of limited value and was not inclined to pay to repeat the exercise. When we looked at the results of the card sorting work we were struck by the fact it seemed to throw little light on how a new information architecture might be structured. This led us to reflect on our own approach.
Anyone who has undertaken a card sorting exercise will know that what it often provides is a BGO - 'blinding glimpse of the obvious'. All the user goals you would expect to go together are put together by users. Goals relating to an organisation's structure, staff and policies are put together in a group and, surprise, surprise, when users are asked to suggest a label for this group they suggest 'About us'!
It is the large range of other goals that we, the information architect, have difficulty deciding how they fit together that users also have trouble sorting. Often each user comes up with their own very different grouping.
A further problem with card sorting is deciding what to write on the cards. Users seem highly influenced by the use of individual words. So the repetition of a single word in very different goals (e.g. policy) can result in them being grouped together when, if the goals are worded differently, they are not.
In this case we decided to undertake some attitudinal research with users to understand more about how users thought about their goals they had for this site. What this showed was that users initially looked for subject based links (e.g. in this case specific words covered by the categories: health, benefits, death etc) rather than the organisational focused categories being used by the site (e.g. information, products & services etc.).
The original card sorting exercise had not identified this because the goals had not been written in a way that would have surfaced this type of 'topic' categorisation. Our conclusion is not that card sorting is ineffective, but rather that you must be very careful about how you write the goals so not to produce spurious associations or exclude other more obvious categorisations.