Loss of performance due to stiction in control valves is an important problem that needs to be solved. Industrial practitioners have articulated that stiction diagnosis from routine operating data is a component that is urgently needed for automated controller performance monitoring and improvement. While the need is well recognized, a solution that is industrially acceptable is extremely challenging. Industrial solutions preclude either invasive testing of valves or maintenance of a large database on valve characteristics for efficient stiction diagnosis. The two key industrial requirements for any stiction diagnosis technique are: the technique has to be non-invasive, and should work with routine operational data. This paper deals with the development of stiction diagnosis and compensation algorithms.