Software process improvement is gaining momentum throughout the software industry. Many cities now have nonprofit SPIN (Software Process Improvement Network) groups. In a visible sign that process improvement is now a mainstream technology, SPIN's national conference in Boston last spring drew several thousand attendees. Another sign is the frequency of journal articles devoted to process improvement. However, as often happens with software, process improvement articles tend to be theoretical. Comparatively little solid, empirical data is being published on three important topics: what does it cost to improve software processes? How long will it take to make tangible improvements? What kind of value can be expected in terms of better quality, productivity, or user satisfaction? The paper condenses some of my findings, which are based on studies of leading software producers in the United States, Europe, South America, and the Pacific Rim.