ABSTRACT This paper will address the need for a statistical approach to validating and optimizing an Automated Optical Inspection (AOI) process. The AOI process is designed to measure and pass judgment on visual Printed Circuit Board Assembly (PCBA) workmanship conditions. The focus of AOI process improvement is to improve AOI's ability to issue an accurate verdict on these visual PCBA workmanship conditions. The majority of the AOI process improvement effort will consist of identifying the optimal imaging and analysis criteria for each inspection requirement. However, changes made to the PCBA manufacturing process in order to make the PCBA more compatible with the current AOI technology would also be considered part of this improvement effort. Optimizing an AOI process can be very challenging given the numerous inspection techniques and the variety of inspection requirements among AOI users. Much of the AOI usefulness as an inspection system is determined by its ability to tolerate levels of acceptable variation while at the same time identifying unacceptable variation for specific component and workmanship conditions. Many AOI systems lack the tools necessary to measure and analyze the levels of variation being experienced. Not being able to measure and analyze the measurement variation and margin to the measurement limits typically results in inefficient and inaccurate adjusting of inspection parameters. This paper will propose statistical methodologies and metrics that could be applied to an AOI process or to individual AOI algorithms. The objective of implementing such methods is to reduce the levels of incorrect component condition verdicts made by the AOI process. The metrics outlined in the document will be used to 1,2 : Provide a baseline of the current inspection process to track AOI improvements. Determine the nature of the AOI problem caused by the measured variation. Predict levels of AOI accuracy based on measurement distributions.