Occupational mobility and its relation with economic fundamentals is obscured by mistakes when assigning occupational codes. We correct (‘de-garble’) occupation patterns, using the heterogeneous probabilities with which a worker in a certain occupation appears as working in another. This leads to stronger empirical patterns of occupational mobility with ‘task distance’ based on O*NET, across age, the business cycle and with wage changes. Miscoding between occupation pairs reflects task similarities and can be used as a distance measure itself, and appears helpful to distinguish among occupations close in task space. Overall, taking into account miscoding, occupations and tasks tend to matter (even) more for economic outcomes than standard approaches suggest.