State space process models and feedback control laws are simplified by eliminating selected state variables using a modified least squares technique. The revised technique uses random perturbations to excite the high order system and the resulting reduced‐order models tend to be more robust than those derived using conventional least squares and deterministic inputs. Low‐order controllers were designed for a pilot scale, double effect evaporator using the modified least squares approach. These controllers performed well in both experimental and simulated response tests.