In telecommunication network management systems, pull and push are the conventional methods for data collection. The manager (also called client) entity manages a large number of network elements (NE, also called server) that send monitoring data to the manager. The pull approach does not depend on data changes and is not efficient. The push approach serves the data strictly based on the data changes and is hence more efficient. However, the push approach does not consider the manager's capabilities; hence, the efforts spent on serving the data will be wasted when the manager is unable or unwilling to process the data. Ideally, data collection should be done based on the manager's capabilities, load and other service level criteria. In this paper, we present an adaptive push technique that combines the best of push's advantages while providing the manager with control over demand in data. This approach is compared with pull and push techniques in terms of metrics such as benefit, cost and optimum. © 2011 IEEE.