We have come to accept that approximately onethird of projects are successful, one-third of projects suffer from overruns of time or cost or failed delivery of desired functionality and fully one-third are abandoned as complete failures. These dismal statistics have defied years of project management research and all manner of innovative ideas ranging from the early introduction of critical path methods, to the more recent critical chain methodologies, agile software development ideas and much behavioral research on project leadership and teams. Moreover, to manage expertise and share scarce resources the majority of projects today are executed in matrixed organizations with their attendant problems of confused lines of responsibility and complex relationships between the project managers responsible for the timely completion of projects and functional managers responsible for the management of resources. In this paper, we outline a three-pronged attack on these problems. Our main supposition is that projects with different size, complexity and economic or strategic importance should be managed differently. Important large projects should be assigned dedicated resources and managed as projects using accepted project management techniques. Medium-sized projects should be managed in matrixed organizations, while small repetitive projects should not be managed as projects at all. Rather they should be managed using techniques borrowed from production and process management. The paper develops the conceptual underpinnings for these ideas and describes how projects in the latter two categories can be best organized and managed. Our "divide and conquer" approach should help management deal with the need for a faster and more sure-footed response to the pressures for change in the face of global competition and ever changing technology.