Simple user picking
Something we all know by now is that a test that uses programmers when the product is intended for legal secretaries is not a usability test. But the number of legal secretaries in the world is quite large, so what subgroup of legal secretaries are we supposed to test with? Categorizing users in the novice and expert subgroups is quite common solution, as is looking for primary and secondary users. If we could combine these in a nice table, everything would be hunkydory, but adding dimensions (subcategories) makes the table complex. We need to visualize the solution in more dimensions. The following figure shows the user cube of the three main dimensions along which users' experience differs: age, experience with computers in general, and with the task domain.
Jakob Nielsen has also taught us that we only need to test with 5 users, but which 5 would that be? If we combine the user cube with this, and add something I call boundary user, we can add another dimension of the categorization that actually helps us in finding users that may help us best with testing a system.
The four boundary users are picked out since they are in the outskirts of the main group, hence they represent typical categories. Complete the user group with the one in the middle (another extreme point) and you get a five-user group that represents the two dimensions in the example above very well. For more dimensions, add more boundary users (another 4 for 3D, etc.).
 Faulkner, X. (2000) Usability Engineering, Palgrave
 Nielsen, J. (1993). Usability Engineering. Academic Press.
 Nielsen, J. (2000). AlertBox: Why You Only Need to Test With 5 Users