One of the most popular and efficient methods for conserving energy in Wireless Sensor Networks (WSNs) is data aggregation. This technique usually introduces an additional delay in the transmission of data packets. The inherent trade-off between energy consumption and end-to-end delay imposes an important decision to be made by the nodes, mainly to determine the most appropriate time for aggregating local/transiting packets and forwarding the resulting packet(s) to the next hop towards the sink. Most of the solutions proposed so far are either unable to significantly reduce the overhead caused by the redundant transmissions, or require a long waiting time before aggregating the received packets. To overcome the above limitations, we propose a novel scheme that ensures efficient and fast packets aggregation in WSNs. This scheme defines optimal decision making policies at the cluster head level (i.e. in a cluster based topology) to determine the appropriate waiting time before aggregating the local data sampling, as well as data sampling received from other neighbor cluster heads. The obtained evaluation results confirm the efficiency of our scheme in terms of the achieved end-to-end transmission delay for periodic packets and the reduced overhead. These results reveal also that our scheme outperforms other three literature schemes (no aggregation, randomized waiting and full aggregation) as it ensures the best compromise for aggregation saving and delay reduction.