Variety of approaches to optimize performance based on behavioral and situational signals are described. A productivity service initiates operations to optimize performance upon receiving a behavioral signal and/or a situational signal associated with a user. The behavioral signal and/or the situational signal are analyzed to infer a usage pattern associated with a component of an application. Next, a failure of the usage pattern is detected in a comparison to an utility threshold. In response, the application is instructed to interrupt an execution of the component.