Model-based control methods like nonlinear model predictive control (NMPC) is a promising approach to control complex chemical processes. While there are many publications on theoretical aspects and simulation studies, the experimental evaluation of NMPC is limited. We present the implementation and experimental validation of model-based control for continuous emulsion co-polymerization in a tubular reactor. We compare two schemes to control the total monomer conversion and the polymer composition at the reactor outlet. The two control schemes are (i) an NMPC scheme and (ii) a heuristic control approach, which consists of PI control, ratio control and feed forward control. The process model consists of a set of differential algebraic equations which capture the polymerization kinetics as well as the dynamics of the auxiliary equipment in the setup. We evaluate the performance of the NMPC scheme in silico for different constraints on the maximum reactor temperature. The simulation studies show that the NMPC maintains the temperature constraints. However, decreasing the maximum allowable temperature in the reactor results in a deviation of the total monomer conversion and polymer composition from the desired specifications. We also experimentally investigate the performance of both control schemes. As expected, the experimental studies show that both control schemes are able to reduce the variation of the polymer composition compared to the case without supervisory control. Although a deviation of the controlled variables from their setpoint values is observed in both control schemes, this deviation from setpoint is less with the NMPC scheme than with the heuristic control scheme. (C) 2019 Elsevier Ltd. All rights reserved.