You need to compare it to the expected error rate.

If you expect for example 100 detections due to noise but you get 1000 detections in the measurement, that would be pretty significant.