We often look for a best-fit function to a set of data. This article describes how a "pretty good" fit might be better than a "best" fit when it comes to promoting conceptual understanding of functions. In a pretty good fit, students design the function themselves rather than choosing it from a menu; they use appropriate variable names; and they vary parameter values by hand--sometimes with the aid of residuals and other tools--to help their function fit the data. (Contains 8 figures and 2 online resources.)