As high volume semiconductor manufacturing approaches sub-65nm transistor technology nodes, process excursions during production and their avoidance are consuming sizeable engineering resources. Daily fabrication production can be significantly impacted by process induced defects and result in increase wafer scrap caused by poor wafer handling, tool aborts during process and resultant suppressed wafer yields [1, 2]. Examples of process excursions in manufacturing range from the obvious plasma ignition abort to not so obvious reduction in backside wafer cooling gas flow. Process excursions are generally detected, in best cases, at downstream OLPM (on-line production monitor) data collection. Unfortunately, worst case scenarios are not detected until electrical testing at the first metal layer or end-of-line wafer sort yields. The detection and containment of process excursions is facilitated by intensive statistical analysis of manufacturing process data and application of aggressive statistical process control (SPC). The process engineers involved are presented with interesting and challenging data mining and multivariate analysis across multiple database types and architectures. This paper will examine data mining methods such as slot tracking, first-wafer-effects, chamber mining and spatial data de-convolution. Examples of electrical test and yield degradation via process excursions will be used to highlight the importance of data mining and statistical analysis to semiconductor manufacturing and product yield.