In this work we address the challenge for an IT service customer to select the cost-optimal service among different offers by external providers. We describe the customer’s optimization problem by considering the negative monetary impact of potential service incidents on its business. First, we demonstrate that the information currently used in service level agreements may lead to suboptimal customer decisions. Second, we discuss how providers’ private information about the behavior of service delivery environments could be leveraged by the customer when selecting service offers. Third, we propose a procurement auction as a mechanism to optimize total cost for the customer – choosing from different service offers by risk-neutral providers. In introducing this approach, we suggest that customers and providers collaborate to define service performance measures, which allow providers to better tailor service offers to customers’ business requirements.