The objectives of this study were to find ways to use online data for Computer Based Education (CBE) lesson development and evaluation, to find correlation with off-line data, and to find limitations in the use of online data. Lessons in mathematics, biology, accounting, and chemistry were given on PLATO to community college and some high school students. Measures like percent of questions answered correctly on the first try, or ratio of wrong answers to the number of questions, were computed for each student. These measurements were compared among the classes that studied the same lessons. Observation in classes helped in judging whether the data were reasonable. Lessons that showed low performance were changed and the new data were looked at to find a reflection of these changes. Graphic representations of various lesson and student data were generated to help in the comparison. Off-line test scores were correlated with online data and stepwise multiple regression performed. Except in cases of very special lesson programming, it was found that online data can be a useful tool to follow and evaluate lesson development; the few exceptions can be corrected. (Author/VT)