Tree creation should be based on not only our content, but our knowledge of our users, our research of their mental models, and our testing of the existing site structure.
We should consider whether the classic grouping schemes (by task, by audience, by topic, and so on) suit our needs for the top levels of our site.
We should consider whether our structure is best designed as wide/shallow or narrow/deep.
- from labelling and taxonomy
We should create several candidate trees (ideally team-sourced) so we can pick the strongest candidates to test against each other (and the existing structure).
We should make sure that our tasks (described in Chapter ~) target the differences between our trees, so our results can help us pick the winning design.
Next: Chapter 6 - Preparing a tree for testing