Since the introduction of the relay feedback test by Astrom and Hagglund (Automatica 1984, 20, 645-651), autotuning of PID controller has received much attention, and many commercial autotuners have also been designed accordingly. Without knowledge of the model structure, most of these relay feedback autotuners use Ziegler-Nichols-type tuning rules to set controller parameters. This can lead to poor performance in some cases, because no single tuning rule can work well for all model structures over the entire range of parameter values. Luyben points out that additional information can be obtained from relay feedback tests, namely, the shape of the response (Ind. Eng. Chem. Res. 2001, 40 (20), 4391-4402). In this work, relay feedback tests are conducted on processes with different orders and a wide range of dead-time-to-time-constant ratios. On the basis of the shape of the response from the relay feedback tests, these processes can be broadly classified into three major categories (model structures). Procedures are given to find parameters for the corresponding model structures, and then different tuning rules are employed to find appropriate PI controller settings. The procedures are tested against linear systems with and without noise. Simulation results clearly indicate that, by incorporating the shape information, improved autotuning can be achieved in a straightforward manner. Moreover, possible dead-time compensation and higher-order compensation can also be devised when necessary. It should be emphasized that the improvement is obtained from the conventional relay feedback test and no additional testing is required.