Overlap of coverage of the five geostationary satellites has allowed an inter-comparison of the FGGE cloud tracked winds. No attempt was made during FGGE to standardize the cloud tracking techniques. In spite of this potential for differences between data sets, the compatibility of the various cloud wind data sets was generally quite good. The vector magnitude differences between nearly co-located vectors showed similar cumulative frequency statistics for all data producers. A study of systematic biases which could affect a global wind analysis of any given synoptic period showed that image alignment errors caused less than 2 m s−1 bias for all data producers except the NESS high level winds which had an average bias of slightly greater than 3 m s−1. This appears to be caused by the manual alignment of images in the movie loops. Height bias studies showed the Japanese winds to be higher than other data producers by as much as 100 mb for both the high and low levels winds. Height biases appear to be caused by the differences in cloud wind height assignment procedures.