Problems with Poorly Scaled Models

A poorly scaled model is one in which the typical values of the objective and constraint functions differ by several orders of magnitude. A classic example is a financial model with some dollar amounts in millions, and other rate of return figures in percent. Poorly scaled models often cause difficulty for both linear and nonlinear Solver algorithms; the effect is often more severe for the nonlinear GRG Solver.

The Solver must perform many calculations where quantities derived from the values of the objective and constraints must be divided into and subtracted from one another. Because of the finite precision of computer arithmetic, when these calculations are performed with values of very different magnitudes, roundoff error builds up to the point where the Solver can no longer reliably find the optimal solution.

When the Use Automatic Scaling box in the Solver Options dialog is checked, the Solver will attempt to scale the values of the objective and constraint functions internally in order to minimize the effects of a poorly scaled model. We recommend that you check the Use Automatic Scaling box for most of your Solver problems.