In recent years, design of wireless sensor networks using methodologies and mechanisms from other disciplines has gained popularity for addressing many networking aspects and providing more flexible and robust algorithms. We address in this paper the problem of random walk to model routing for data gathering in wireless sensor networks. While at first glance, this approach may seem to be overly simplistic and highly inefficient, many encouraging results that prove its comparability with other approaches have been obtained over the years. In this approach, a packet generated from a given sensor node performs a random motion until reaching a sink node where it is collected. The objective of this paper is to give an analytical model to evaluate the performance of the envisioned routing scheme with special attention to two metrics: the mean system data gathering delay and the induced spatial distribution of energy consumption. The main result shows that this approach achieves acceptable performance for applications without too stringent QoS requirements provided that the ratio of sink nodes over the total number of sensor nodes is carefully tuned.