Grid computing has been identified as an instrument to fulfil high computational demand, a promising approach for higher resource utilization, and an instrument for cost reduction. The full potential of cost savings can be tapped when incentives are set such that demand is shifted to periods or hardware with lower demand, thereby flattening the demand. To set such incentives, it is mandatory to know the preferences of the Grid resource consumers. We propose the application of the ADBUDG-function to estimate the willingness-to-pay function of prospective resource consumers and K-means clustering to determine optimal tariffs based on this information. We demonstrate the application of this approach in the financial service industry at a large European bank that is planning to move from dedicated servers for single business units to an enterprise Grid. Based on a sample of 21 project leaders and business unit heads with their own budget responsibility, we show how optimal price tariffs can be determined using our proposed approach. Based on the self-reported preferences, we determine tariffs that allow an additional cost savings of 7% or an increase in utility by at least 6% and 9%, on average.