End-use electricity demand forecasts play critical role in resource planning approaches that actively consider both supply- and demand-side options to meet customer energy service needs. Yet, in order to forecast peak demands by end use, utility and state planners have had to rely on both simulated and borrowed end-use and class load research data. This reliance has introduced additional uncertainty into an already complicated resource planning process, as questions arise regarding the veracity of these inputs. The data now available from recent end-use metering projects holds the promise of reducing these uncertainties and thereby improving the planning process and its outcomes. This report summarizes findings from a unique project to improve the end-use electricity load shape and peak demand forecasts made by the Pacific Gas and Electric Company (PG&E) and the California Energy Commission (CEC). First, the direct incorporation of end-use metered data into electricity demand forecasting models is a new approach that has only been made possible by recent end-use metering projects. Second, and perhaps more importantly, the joint-sponsorship of this analysis has led to the development of consistent sets of forecasting model inputs. That is, the ability to use a common data base and similar data treatment conventions for some of the forecasting inputs frees forecasters to concentrate on those differences (between their competing forecasts) that stem from real differences of opinion, rather than differences that can be readily resolved with better data. The focus of the analysis is residential space cooling, which represents a large and growing demand in the PG&E service territory. Using five years of end-use metered, central air conditioner data collected by PG&E from over 300 residences, we developed consistent sets of new inputs for both PG&E's and CEC's end-use load shape forecasting models. We compared the performance of the new inputs both to the inputs previously used by PG&E and CEC, and to a second set of new inputs developed to take advantage of a recently added modeling option to the forecasting model. The testing criteria included ability to forecast total daily energy use, daily peak demand, and demand at 4 P.M. (the most frequent hour of PG&E's system peak demand), We also tested the new inputs with tile weather data used by PG&E and CEC in preparing their forecasts.