There is the power law heuristic, that 90% of results are often due to 10% of phenomena.

If there are several filters, then they may differ in effectiveness by many multiples, and then it's semantics as to whether you say any beyond the largest are "great".