Setting unrealistically short schedules and low staff levels in the early stages of software projects frequently leads to lower product quality and reduced customer satisfaction. Effective cost-and schedule-estimating models of the software development process, however, can suggest appropriate staffing and reasonable schedules. Commercial models for this purpose have been available since the early 1980s, but differ widely in their algorithms and in the factors they use to adjust estimates to local environments. This paper reviews efforts in AT&T Research and Development to develop and use such models on AT&T projects. R&D models have been formulated and incorporated into software tools. The tools are being integrated into the planning process, and early results have been promising. When the tools were tested, the estimates were generally within 20 percent of actual experience.