With the advances in mobile technologies is now possible to support learners and teachers activities on the move. We analyzed the functionalities that should be provided by a general mobile learning platform and identified a problem that is weakly studied, namely support for offline usage of learning material, called hoarding. Hoarding can use some techniques used by caching and prefetching schemas, but in most cases the goal of the last two is to reduce latency time, bandwidth consumption and/or servers workload, while in hoarding the aim is to reduce the size of the hoarding set, keeping the accuracy very high. We want to study the parameters that could help hoarding algorithm to face the peculiarity in m-learning scenario and our final goal is to provide an efficient strategy, taking into account additional parameters, extracted automatically by the system.