System and method of intelligent application diagnostics of a user device (UD). Upon every first time period, the key performance indicators (KPI's) of the processes running on the UD are determined. The top predetermined number (N) of processes that are using the most computing resources are filtered and their KPI's are stored in a local memory. The KPI's of the remaining processes are discarded. This process is repeated at every first time period. At every second time period, the filtered KPI's of every iteration of the first time period are retrieved from the local memory and sent to a remote monitoring server for analysis and storage. The filtered metrics are removed from the local memory.