The design of granular media filters has evolved over many years so that modern filters have larger media sizes and higher filtration velocities than in earlier times. The fundamental understanding of filtration has also improved over time, with current models that account reasonably for all characteristics of the media, the suspension and the filter operation. The methodology for design, however, has not kept pace with these improvements; current designs are based on pilot plants, past experience, or a simple guideline (the ratio of the bed depth to media grain size). We propose that design should be based universally on a characteristic removal length, with the provision of a bed depth that is some multiple of that characteristic length. This characteristic removal length is calculated using the most recent (and most complete) fundamental model and is based on the particle size with the minimum removal efficiency in a filter. The multiple of the characteristic length that yields the required bed depth has been calibrated to existing, successful filters.