Even though some progress has been made on the use of bulk longitudinal and/or shear waves for defect classification, the purpose of this paper is to exploit the potential of guided waves in flaw classification analysis. The multi-mode character of guided waves offers the required redundancy in improving the probability of detection and also in establishing a rich and complete data base for use in neural net flaw classification algorithm development. The use of Boundary Element Methods are also used to select promising features for reflector classification purposes. Neural nets are superb for algorithm development and implementation purposes in transforming a number of promising feature observations and expectations to a decision algorithm. Aspects of guided wave analysis are presented that are critical to use of neural net analysis, from data acquisition and feature source methodology, onto neural nets for the development of a decision algorithm. Along with general philosophic discussions, a sample problem of defect characterisation in tubing is presented, whereby specific feature sources characteristic of a tone burst swept frequency approach are used in the development of specific flaw classification decision algorithm.