I have an existing online catalogue of more than 300 services to provide improvement to. One of the common complaints anecdotally is around the current navigation, that the labeling isn't clear and users can't find what they're looking for.
What's a good approach to arriving at a user-centered IA with so many elements? My initial thought was to do a closed card sorting exercise, but 300-some elements strikes me as too much to ask a user to go through, no? Most other references I've read talked about no more than around 100.
- Should I take a random sample of 100 services? Maybe use different samples with different users to cover all 300 services at least once?
- Do I take just the most popular services? I'm afraid this will just reinforce a bias toward those services that are already easily discoverable.
- Do I manually select a sample of services are representative of different aspects of the catalogue? In doing that, don't I risk introducing my own biases into the process?
- Am I not looking at the right approach? Perhaps there is something better suited to large collections of information that would let me get top-level structure, and then look at card sorting for each high-level category (assuming they would be smaller, say 50 to 100 elements)?