In the choosing of a launch vehicle for a given mission or in the determination of insurance coverage and premiums for a given launch, accurate estimates of the probability of success of the different launch vehicles provide important information. There are three general approaches to estimating the probability of launch success. The first is to use a probabilistic risk analysis, decomposing the system into its subsystems and components and estimating the probability of each of the failure modes. The second is to rely on expert judgment about the vehicle's success rate as a whole, without a functional decomposition of the system. The third is to use statistical data about the past performance of the system to estimate the vehicle's success rate. The focus is put on this last approach, using Bayesian probability theory to make better use of vehicle-level performance data. The procedure is demonstrated by an analysis of the success rates of most of the major families of launch vehicles currently in use in the world. A family of launch vehicles includes all variants of a particular type of vehicle from a specific manufacturer, for example, the Delta 2. For vehicles with a small number of launch attempts, the Bayesian approach provides the advantage over classic statistical approaches of yielding estimates of both the mean future frequency of success and the uncertainty about that mean.
