Most of the on-demand routing protocols in WSNs are designed with the aim to find the best path based on a single routing metric such as minimum hop count, so that the End-to-End (ETE) delay can be minimized. In these protocols, the probability to select the same sensor node(s) all the time is high, resulting in overloaded nodes and increased energy consumption per node. Furthermore, in these protocols, each intermediate node acts only on the first request it receives and the subsequent requests are ignored during route discovery, leading to suboptimal routes. In this paper, the Congestion-aware energy efficient and traffic Load balancing Scheme (CLS) for routing in WSNs is proposed. This scheme utilizes the ignored information during the route discovery process and considers a composite metric that incorporates the consumed energy E, participation level P of the node and signal strength S of the link between the nodes. In addition, through extensive simulations using ns-2.35, we demonstrate the effect of weighted additive approach against the lexical approach for multi variable routing metric. © 2014 IEEE.