To evaluate a network's performance in processing communications related to a target transaction, a set of “reference” communications corresponding to the target transaction are compared to a larger set of communications in the network in a “production” environment, to identify the occurrence of the target transaction in the production environment. Preferably, the reference communications are recorded in a laboratory environment that models the production environment, or recorded from the production environment during a period of minimal other activities. A variety of filters are used to eliminate communications in the production environment that are apparently unrelated to the target transaction, including filters based on the time-order of communications among the nodes, the size of the packets being communicated, and the content of the communications. If necessary, after eliminating the apparently unrelated communications from consideration, the remaining production communications are compared to the reference communications to identify the most likely production communications corresponding to the reference communications.