Today, enterprise IT environments are complex as never before with individual applications, tiers, or technologies segregated into individual management domains. Typically, the value of business applications and the dependencies between business and IT objects and IT objects among each other is unknown or at least not up to date. Thus, ultimately, the business value of individual IT tasks is unknown. Hence it is very hard to perform global management services such as performance optimization in resource-constrained environments. This deficiency is even more deeply felt by an internal or external services provider called in to perform an optimization or to improve an IT management framework. We propose a framework ITBVM for business-value driven IT optimization with particular emphasis on such enterprise environments. A key part is the use of discovery technologies to provide the link between business value and IT objects. As one instance of the framework, we show how discovery can improve a performance-optimization problem in an otherwise blackbox scenario. We validate these improvements through experiments in a controlled setup and through statistical interpretation of fine-grained dependency discovery in a large real enterprise environment.