Dozens of Ground Delay Programs (GDPs) are implemented each summer for San Francisco International Airport (SFO) in order to cope with reduced capacity caused by the presence of warm-season stratus in the approach zone. The stratus prevents the use of dual approaches to SFO's closely-spaced parallel runways, which essentially reduces the arrival capacity by half. In 2004, a prototype system for providing probabilistic stratus forecast guidance was transitioned from the research community to NWS Monterey. This system was intended to be used as a tool for improving the daily forecast of stratus clearing time from the approach zone, and correspondingly improve the efficiency of GDP implementation strategy. Since its transition to the NWS in 2004, the automated forecast guidance system has continued to produce reliable forecasts of daily stratus clearing time. However, this success has not adequately translated to a marked improvement in GDP efficiency. Analysis by the NWS indicates that the existing mechanisms for introducing the forecast guidance information into the GDP decision process, as well as the GDP implementation strategy itself, are not suited for taking full advantage of the forecast skill demonstrated by the system. A historical examination of SFO GDP implementation based on the probabilistic forecasts provided by the automated forecast guidance system is currently in process, with the objective being a recommendation for a more effective GDP strategy. An important consideration is understanding the risk/reward associated with the decision process. In this instance, the reward is increased efficiency seen as reduced aircraft delays, at the risk of creating increased delay, aircraft diversions, and controller workload in the event that an incorrect optimistic forecast results in the premature release of ground-held aircraft. This investigation is being performed in concert with the weather-integration objectives of the current FAA modernization program, particularly the integration of weather information that is delivered in a probabilistic format. Shortcomings within the current GDP strategy are described to provide context for potential improvements that exploit the probabilistic forecasts currently emerging from the research community.