Usage-based pricing is commonly used by service providers of, among many others, wireless and fixed-line telephony, Internet access, and digital music. The optimal pricing of these services requires the simultaneous prediction of each consumer's decisions on service subscription, tariff choice, and consumption. This prediction is challenging because all the decisions are interrelated and the services differ not only in prices, but also in their service attributes. This article expands previous research and shows that some of these service attributes affect tariff choice via the subscription decision, and others affect it via the consumption decision. This difference is important because the two mechanisms differently affect willingness to pay and the resulting optimal prices. Our empirical study indicates that failure to model accurately the influence of service attributes leads to non-optimal prices and profits that differ by up to 17% from the optimal solution. In addition, this research shows how discrete choice models can capture the innovative subscription plans of bucket pricing that are nowadays more widely used for innovative products such as wireless data plans for the iPad (e.g., 250 MB for $14.99, 2 GB for $ 25.00).