Outage probability has been commonly used as the primal metric to investigate system capacity of cellular networks. Outage probability is a system-level metric that lacks key information with regards to outage performance within the lifetime of a particular session. Analysis based on the satisfieduser criteria has been recently recommended by universal mobile telecommunication systems (UMTS) where the outage percentage for each individual session is computed for a more genuine estimate of system capacity. In this paper, the session outage performance is modeled as an alternating renewal process with exponential holding times (ARP/E). Estimation of satisfied-user probability resolves to derivation of the total downtime (bad-time) percentage of an ARP/E. An exact formula for the satisfied-user probability is derived for speech services. The ARP/E analysis developed for speech is also employed as an approximation for the satisfied-user probability of WWW browsing users.