One of the least understood and most ignored aspects of implementing on-line software is the problem of effective monitoring. This problem takes on many facets: gathering data on user interactions and errors, storing and processing these data, and generating summary reports which can be used to pinpoint deficiencies and to suggest modifications and improvements. Northwestern University has developed techniques to monitor an on-line information retrieval system, by using that system both to monitor itself and to process the data collected. RIQS (Remote Information Query System) is both a batch and on-line generalized, information retrieval system which operates on the CDC 6400. It is used by faculty and students for developing and searching data bases in medicine, political science, computer science, anthropology, astronomy, university administration and planning, etc. An on-line monitor program was developed to produce a RIQS-compatible data base of user interactions with RIQS-ONLINE. Using RIQS, this data base can then be interrogated and searched to generate reports, to create plots, or to produce input for a statistical package. Typical parameters which are studied include total real-time per user access to RIQS, real time and central processor time per query, number of records searched, number of records retrieved, number of search terms used, type and size of reports generated, number and type of language or syntax errors made by user, etc. This paper illustrates the use of these techniques by providing summary reports and plots from use of RIQS-ONLINE during the current academic year. System improvements are suggested from analysis of these reports.