In The design phase: creating new trees in Chapter 3, we strongly recommended doing more than one round of tree testing to get a site tree performing well:

• Two rounds is very common – round 1 for the original tree and the proposed replacement, and round 2 to test revisions to make sure our changes actually solve the problems we found.
  
  
• Three rounds is ideal – round 1 for the original tree (along with an open card sort to generate ideas), round 2 for a few alternative trees to compete against each other, then round 3 to test revisions to the tree that won out in round 2.
  
  
• If parts of our chosen tree are still not performing well, we can easily do further rounds to keep improving our results.
  
  

At some point, however, we need to be done:

• In the best case, we’re done when our tree is performing well (typically a success rate of 70% or more) and there are no more useful revisions we can make to the tree – or at least, no revisions that are going to make a real difference to our tree-test results. We may still tweak the tree in later stages of the project (e.g. after usability testing), but for now, we have a solid foundation to build on.
  
  
• If time or budget (or motivation) run out before we achieve a high-performing tree, that’s not optimal, but it’s still much better than doing no testing at all. We have made substantial revisions, and we’ve identified where the trouble spots remain, so we can keep an eye on them (through usability testing and analytics) and fix them later when circumstances allow.

Next: Chapter 14 - key points