The study tested the idea that full buffers mean a bottleneck downstream and empty buffers mean a bottleneck upstream. They found that this is true for balanced systems, but in unbalanced systems, the bottleneck direction is more likely to be biased towards the bottleneck.