I need to consolidate two web products that have some similarities and many differences in labels and groupings. For round 1 testing, I plan to do open card sorting with content from both sites. I would test product 1 users, product 2 users and internal group with industry standards knowledge. I'm hoping that by triangulating with the 3 groups I will have enough insight to build an IA to test with TreeJack. Is there anything else I can do to accommodate both product users' mental models?

3 Answers 3


In the technology and business context:

In the technological context, product consolidation is important when both systems have a common framework (which also means that maintaining code involves simple replication) it is important to define the skeleton with common elements at the beginning and then, thanks to the final results of segmentation, to diversify the functionality (define differences in IA, based on the existing common core)

in the context of segmentation and research:

it is worth defining the target groups of individual products and based on them design the difference (mental models), depending on the priorities, such elements as category trees may vary. In the initial phase, however, I would minimize the difference just to preserve the common core. Then, however, it gradually introduced differences depending on the preferences of end users and business goals.


The exercise that is intended seems to be in path of creating a Unified Experience. Apart from card sorting there are few other qualitative analysis that can be done based on availability of time -

a. Mental model diagrams mapping each web platform provides detailed insight of how users perceive and think about various parts. This ideally will help to map/trace back the IA for its robustness and effectiveness. Any gaps that may emerge will be also backed up by Card sorting exercise

b. Pain matrix : Devise a 2x2 matrix with x axis on scale of pain against y axis with frequency of pain. This pain matrix helps to understand the discoverability/usability of the platform.

You could also run a quantitative analysis if the user population is large through survey to get more observations.

While the above methods are relatively not done through making users 'do/perform' a task - since there is larger difference between 'doing' vs. 'sharing/reciting'. If this would be critical, few mock-ups could be tested/Guerrilla testing with users as participatory method to approach the problem from bottoms-up (ie. through design)


Depending on how many categories your open sort produces you might normalize and consolidate those into a very de-duplicated handful of categories which you can then use in a closed sort. This is a good intermediate step that lets you start to evaluate assumptions (your normalized category names) before you try to validate an entire taxonomy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.