The engineering of cable networks for IPbased voice and data services presents new planning challenges to cable operators. Unlike the broadcast video services for which cable networks have traditionally been designed, the traffic of these new services is directed to individual subscribers. We describe algorithms for estimating capacity requirements to support IP-based services at acceptable levels of QoS as well as the traffic models on which they are based. The algorithms account for the efficiencies realized from the statistical multiplexing of independent traffic streams of different subscribers. We provide examples of their use in investigating various "what-if" scenarios. By combining the capacity estimation algorithms with methods for deriving the parameters of the traffic models from network measurements, one could create a monitoring and planning system for provisioning IP services on cable networks.