Traditionally, the sensor network design procedure was based on positioning sensors so that certain network monitoring capabilities (e. g., observability, redundancy, and error detectability) of key variables are assured at minimum sensors cost. We present a new approach that is based on maximizing economic value of information minus cost instead of the traditional approach that requires the satisfaction of performance targets. This article presents the conceptual aspect and computation issues of this new approach: the connection between the new approach and the traditional minimum-cost approaches is explored and the computational methods to solve the proposed problem are presented. (C) 2010 American Institute of Chemical Engineers AIChE J, 57: 2136-2148, 2011