Purpose – This paper intends to illustrate an application of data envelopment analysis (DEA) to assess library performance from an efficiency standpoint. Design/methodology/approach – DEA modeling was applied to a convenience sample of 37 libraries affiliated to a federal university in Rio de Janeiro. Data were collected from the university’s managerial database and refer to three inputs – number of employees, area and number of volumes – and four outputs – consultations, loans, enrolments and (user) traffic. Markovian analysis of transitions between efficient and inefficient states along time allowed a long-term distribution between those states to be computed. Findings – The retained DEA model provides a list of estimated scores that quantify efficiency status for each library unit and from which both rankings and operation plans can be determined for each unit to assist managers in their quest for library efficiency. In fact, (re)allocative measures, expressed as operation plans, indicate that, for each unit, some input(s) may be decreased and nonetheless some output(s) will increase. Those indicators may also be used to further or avoid either promised or planned changes. Originality/value – As long as the efficiency principle is accepted the paper provides a three-step procedure whereby any set of library units may be simultaneously assessed and ranked in relative terms and a set of quantitative operation plans may be used to (re)direct inefficient units toward efficiency. Whenever historical (e.g. annual) data are available, more adequate long-term efficiency profiles will be computed, as well as some (e.g. yearly) durations relating to time spent in or before visiting (in)efficiency states. This model, combining short- and long-term assessment, may be seen as a novelty contributed by the paper.