A free comprehensive guide for evaluating site structures
The tasks we create should cover the most common and critical activities that visitors will do at our website, and any “suspect” areas of our tree.
We should create as many tasks as we need to cover the key parts of our tree, but we shouldn't ask a given participant more than 8-10 tasks.
We should decide if we need different tasks (and therefore separate tests) for different audiences, or if they can all “reasonably pretend” to do each other’s tasks.
We need to be careful and consistent in marking our correct answers, because there may be more than we expected.
In most cases, we should randomize the order of tasks to reduce the learning effect.
In most cases, we should let participants skip (give up on) tasks to avoid user frustration and pollution of the results.