This hasn't been documented officially, unfortunately. But, it is based on the speed each mesh unit communicates back to the primary during the mesh test benchmark. Anecdotally, I have seen units that get 100Mbps on that benchmark being reported as "good", and I know at a number higher than that they will be reported as "great". Sorry I don't have the detailed threshold between those numbers. But, in general, you really want to have your mesh units placed such that all of them show a "great" mesh test result. If any of them aren't great, the channel will end up being used for longer periods of time to service traffic to and from that unit, taking capacity away from everything else.