Retrieval and local storage of data at mobile devices is managed by employing a scheduling policy based on a prediction policy of a preloader, a pre-render policy, and with respect to a current viewport state to reduce waiting time and mobile device resource usage. Portions of data maintained on a server associated with a client device are retrieved and preloaded into the client device memory to enable a user to view documents seamlessly while navigating through a document without overburdening the client device memory and/or processing capacity based on a prediction. A current viewing position of the document and user navigation on the document via a touch action or gesture are detected and next portion of data to present another portion of the document is predicted. A request manager may retrieve the requested data from the server and pre-rendered so that it is available when requested by the user.