In many of the cases, industrial processes are uncertain and the dynamics cannot be expressed accurately. System identification is a powerful tool which allows a user to estimate a process model, only with the information of input/output. Mostly, with the system identification, an integer or a fractional order model is estimated. An experimental study on process model type versus control performance is presented in the paper. To execute the work, first a fractional and integer order model of a two-tank level-loop is estimated with system identification and experimental data sets. Next, controller settings are obtained for both the models with two methods. In the first method, the controller is designed with internal model principle with maximum sensitivity. Whereas, the second method is based on target-loop with phase margin. Finally, the simulated and experimental results for both type of models are compared for the performance and robustness comparison.