Intelligent Character Recognition has been frequently suggested at ETS as a means to read handwritten numeric digits for processing Financial Aid Forms and other documents as a substitute for key entry. Since it is not likely that all digits will be intelligible, a character recognizer usually converts the numbers it can recognize and passes unrecognizable numbers to human operators for interpretation and keying. Presented in this paper is an evaluation of a commercially available neural net recognizer and its effectiveness in processing data collected in the late stages of the 1991-92 Financial Aid cycle.