This paper presents an analysis of some of the definitions of power factor as applied to nonlinear and unbalanced systems and gives alternatives to the use of a single power factor. The three most important practical functions of power factor for linear balanced circuits are the computation of the increment in the transmission losses, the computation of the required reactive compensation and the information to specify (or design) machines. So far, the three have not been considered simultaneously for nonlinear and unbalanced circuits. In the paper we show that, for the general case, the three purposes cannot be achieved consistently with only one power factor. We propose alternatives that are completely general and applicable to linear and/or nonlinear, balanced and/or unbalanced circuits. In addition, we present a strategy for the optimal reactive compensation that yields the minimum transmission losses.