Ultrasonic thickness C-scans provide information about wall thickness of a component over the entire inspected area. They are performed to determine the condition of a component. However, this is time consuming, expensive and can be unfeasible where access to a component is restricted. The pressure to maximize inspection resources and minimize inspection costs has led to both the development of new sensing technologies and inspection strategies. Partial coverage inspection aims to tackle this challenge by using data from an ultrasonic thickness C-scan of a small fraction of a component’s area to extrapolate to the condition of the entire component. Extreme value analysis is a particular tool used in partial coverage inspection. Typical implementations of extreme value analysis partition a thickness map into a number of equally sized blocks and extract the minimum thickness from each block. Extreme value theory provides a limiting form for the probability distribution of this set of minimum thicknesses, from which the parameters of the limiting distribution can be extracted. This distribution provides a statistical model for the minimum thickness in a given area, which can be used for extrapolation. In this paper the basics of extreme value analysis and its assumptions are introduced. We discuss a new method for partitioning a thickness map, based on ensuring that there is evidence that the assumptions of extreme value theory are met by the inspection data. Examples of the implementation of this method are presented on both simulated and experimental data. Further it is shown that realistic predictions can be made from the statistical models developed using this methodology.