Distribution utilities serve millions of customers and have traditionally depended on measurements with low temporal resolution, historical data or pseudo-measurements. With increasing deployment of residential smart meters in recent years, they now have access to a very large amount of data and need efficient data mining strategies to effectively utilize consumer consumption information. This paper presents a computationally efficient strategy using k-means clustering for identifying suitable candidates in demand-response programs. The attribute selection approach used takes into account the average annual consumption of customers, shapes of their load profiles and also aims to minimize computation burden. The objective is to identify a small group of customers who can be incentivized to curtail their peak period consumption leading to significant aggregate peak-shaving for the utility. Performance of the proposed approach is demonstrated using real-world data from the Irish smart meter trial.