The City of Los Angeles has performed over twenty thousand residential water surveys since the Department of Water and Power initiated its Home Water Survey Program in the summer of 1990 as part of its voluntary conservation program. Marketing letters were sent out to single family accounts that ranked at the top 20 percent of water use and approximately five thousand surveys were completed. In the following year, the imposition of mandatory conservation targets resulted in a vastly increased response to a second offer of home water audits. A third marketing effort in 1993 was timed in conjunction with a new rate structure that imposed a higher price for higher levels of water use. Because of the design of the home survey--consisting of installation of toilet displacement devices, retrofit of old showerheads and faucet aerators, leak detection and repair, and performance of an outdoor irrigation efficiency audit--remained constant through the three phases of the program, these data represent a nearly ideal natural experiment. Statistical models of household water demand are used to link the inspection data on household water-using characteristics, climatic measures, and water rate data with changes in metered water use. These empirical models yield concrete answers to both how much water was saved and why the level of conservation may be different for the different phases of the program; different households within a given phase; different seasons of the year; and different lengths of time (i.e. the persistence of savings). The method and results of this study should interest conservation coordinators and water resource planners who need to know the reliable payoff from this type of demand-side investment.