Wide-area control design is one of the main challenges in moving toward a self-healing power grid. Historically, controls in power system were designed and operated using component-based models derived from linearizing nonlinear differential equations governing the system. With recent trends toward deployment of renewable distributed generation and introduction of new types of loads and various load controls, power system models are becoming increasingly inadequate for representing the system. Therefore, a measurement-based model, estimated online directly from the measurements, can serve as a good complement to conventional models. In this paper, we proposed a new wide-area control framework using online model-identification to improve the small signal stability. Based on model validation studies, three cases of common model inaccuracies were considered to compare component-based and measurement-based models for control design purpose. Our simulation results on Kundurs two-area test system illustrates that a control designed solely using component-based models fails to provide the expected performance due to model inaccuracies that are hidden to system operators in real-time. The proposed wide-area control confirms the benefits of control design using measurement-based models.