Fault detection in planet bearings is difficult. This is particularly true in wind turbines, where the main rotor shaft is under 20 rpm, such that the planet fault frequency can be sub 1 Hz. This papers analyzes a missed fault on a wind turbine planet bearing, and discuses how changes in the analysis configuration then allowed this type of fault to be detected. From raw data from ten machines, a strategy for fault feature identification was developed, to include, the evaluation of window selection, biasing of the data set with faulted components, and the use of improve analysis techniques. This allowed meaningfully appropriate thresholds to be set.