Virtually all variables used to forecast water demand are uncertain: p
opulation and household formation, employment and industrial growth, i
ncome and water use characteristics by customer sectors, conservation
targets and performances, and weather. The key to accurate forecasting
is to isolate or specify the effects of each variable so that changes
in water demand that are attributable to changes in the variables ove
r time can be accurately included in the forecasts. After a forecastin
g model has been properly specified, it is far more efficient to recog
nize the inherent error level and work with it to identify upper and l
ower bounds of potential water demand than to persist in fine-tuning t
he model to extract the last ounce of efficiency from a limited databa
se. The objective of this article is to briefly discuss the major meth
ods of demand forecasting and then concentrate on measuring conservati
on performance and integrating conservation targets into long-run dema
nd projections.