In this work, the problem of identifying third-order Volterra models from process input-output data is considered. As a metric of model fidelity, the prediction error variance expression is employed. Tailored input sequences are designed to estimate the linear, nonlinear diagonal, and subdiagonal kernels for the third-order Volterra model. Random binary signals (RBS) are used to estimate the second- and third-order off-diagonal kernels. These input sequences are plant-friendly, and they are designed to take advantage of the Volterra model structure. For comparison, Volterra kernels are identified using a cross-correlation approach. A mildly nonlinear system, which has been previously employed, serves as the simulation case study (Congalidis et al. AIChE J. 1989, 35, 891; and Parker et al. J. Process Control 2001, 11, 237). On the basis of both static and dynamic analyses, the tailored identification algorithm estimates kernels that capture the nonlinear system behavior. Furthermore, the estimates from this algorithm are superior to, and require less data than, estimates using the cross-correlation technique.