This paper describes the methodology for the redesign of the sample for the commodity and services component of the U.S. Consumer Price Index to accommodate the introduction of outlet frames from the Telephone Point of Purchase Survey. This work represents a further expansion and revision of models developed for the 1998 CPI sample redesign. Models relating data collection costs and sampling variance of price change to item and outlet selection variables for the sample design were developed and estimated. With these models, data collection resources were allocated to minimize sampling variance of price change, subject to budgetary and operational constraints, using nonlinear programming techniques. Models for sampling variance and costs are given, and solutions to the design problem posed under varying assumptions are discussed.