Modeling customer demand in print service environments using bootstrapping

For simulation modeling, what-if analysis and optimization studies of many service and production operations, demand models that are reliable statistical representations of current and future operating conditions are required. Current simulation tools allow demand modeling using known closed-form statistical distributions or raw demand data collected from operations. In many instances, demand data cannot be described by known closed-form statistical distributions and the raw data collected from operations is not representative of future demand. This paper describes an approach to demand modeling where historical demand data collected over a finite time period is combined with user-input using two-tier bootstrapping to produce synthetic demand data that preserves the statistical distribution of the original data but has overall metrics such as volume, workflow mix and individual task and job sizes that represent projected future state scenarios. When the customer demand data follows highly non-normal distributions, a modified procedure is presented.