Damage tolerance‐based operation and maintenance of aircraft components require the estimation of probability of detection (POD) of nondestructive inspection (NDI) techniques. The standard approach to estimate POD of an NDI process is to perform demonstration tests on simulated or real components. To generate a realistic POD, MIL‐HDBK‐1823 suggests that flaw sizes should be uniformly distributed on a log scale covering the expected range of the POD curve in the transition region. This paper presents an alternative approach that is based on Rayleigh distribution of flaw sizes on a log scale covering the expected range of flaws. This approach is only applicable to NDI methods that provide a signal response or “a⁁” for a given flaw size “a”. The use of this approach increases the accuracy of the POD estimate by optimizing the distribution of flaw sizes in the range of transition between very low and very high POD. The paper also presents a new three dimensional representation of POD versus the false call rate, defined by the relative operating characteristic (ROC). The 3D presentation makes it easier to visualize the change in POD and false call rate with the variation of decision threshold set for flaw detection. © 2003 American Institute of Physics