A methodology for the identification of multivariable processes is developed that can achieve a higher level of controllability by increasing one or more time delays between the inputs and outputs. By examining the delays in the numerator and denominator of the inverse of the process transfer function matrix, a procedure is given for the identification of the delays which should be increased, and by how much these delays should be increased, in order to yield the best performance. A simple heat recovery network provides a practical example of the application of this technique. An enhanced multidelay compensator, designed on the basis of the process model with increased delays, yields a consistently better closed loop response than existing compensators.