The idea of Matched-Pulse (MP) reflectometry is here introduced, as an improvement for fault-detection techniques in wire networks. As opposed to existing reflectometry methods, the testing signal is network-dependent and self-adaptive, and designed in a way as to maximize the probability of detection of an echo associated to an eventual fault. The detection performance of this novel technique is investigated by means of a mathematical analysis comparing it to standard Time-Domain Reflectometry (TDR) methods. It is proven that the more complex the network the better the performance the MP approach will provide. This is opposed to common understanding, as standard TDR techniques yield degrading detection probabilities in the case of the complex networks. Furthermore, the MP approach is shown to be particularly interesting for soft-fault detection, as it provides an improved performance as soon as the fault is weakly reflective. Our findings show that the MP approach is expected to bring a substantial improvement in the likelihood of properly detecting faults in configurations that are deemed critical for standard reflectometry techniques, particularly when dealing with noisy NUTs. Experimental results measured on coaxial complex networks support these claims.