A sampling network design model is presented that evaluates the trade-off between the varying costs of different types of data and the contribution of those data to improving model reliability. The methodology couples parameter-estimate and mode-prediction uncertainty analyses with optimization to identify the mix of hydrogeologic information (e.g., head, concentration, and/or hydraulic conductivity measurement locations) that will minimize model prediction uncertainty for a given data collection budget. Two alternative optimization algorithms are presented and compared: a branch-and-bound algorithm and a genetic algorithm. A series of synthetic examples are presented to demonstrate the adaptability of the methodology to different sampling scenarios. The examples reveal two important properties of this network design problem. First, model-parameter and model-prediction uncertainty analyses are important components of the network design methodology because they provide a natural framework for evaluating the cost/information trade-off for different types of data and different sampling network designs. Second, the genetic algorithm can identify near-optimal solutions for a small fraction of the computational effort needed to determine the globally optimal solutions of the branch-and-bound algorithm.