A performance test of a detection model for estrus and mastitis was perform
ed on four farms during several years. Sensor data of milk yield, milk temp
erature, electrical conductivity and animal activity were the inputs for th
e model. The sensor equipment differed between farms. On two farms leg tran
sponders were used for activity measurements, on the others neck transponde
rs were used. On three farms conductivity was measured of quarter milk, on
one farm conductivity of only mixed milk was measured. The sensitivity for
estrus varies from 63% to 80%, depending on the chosen settings, with speci
ficity varying from 94% to 98%. There are great differences between farms.
The sensitivity for clinical mastitis varies from 55% to 80%, depending on
the model setting, with specificity for mastitis from 94% to 99%. Also for
mastitis sensitivity there are great differences between farms. The equipme
nt can only partly explain these differences. Other explanations are indica
ted. The main conclusion is that a detection model can be useful but only i
f the equipment is monitored regularly.