Business Analytics and Optimization in software development – Experience at IBM Rational

Tsutomu Kamimura, Rational CTO for Asia Pacific, IBM Software Group

Abstract: Increasing complexity in today’s economy, vast proliferation of data, and pressing need to stay competitive have motivated many companies to explore analytical techniques to leverage for business improvement in various areas from making investment decisions, improve operations, to better customer support. Business Analytics and Optimization is sometimes used to refer such approach. As software becomes increasingly critical in our society, it is natural to consider software delivery activities as business process rather than engineering process and to use analytical techniques to improve software activities from business perspectives.
This talk will present examples of such efforts at IBM Rational. IBM Rational produces about 100 software products with over 2000 software engineers at over a dozen of major world wide locations. We have a need of producing high quality software with right level of features, quality, cost and schedule. We will discuss three examples of applying metrics and analytics to three different types of activities, i.e., sustaining and maintenance work, active development effort and planning for product portfolio. In each of these efforts, there is use of metrics and analytics for achieving business goals that are interesting in their own right.
There are also some common lessons learned, such as the need of tight linkage with business needs, the need of sound measurement framework, and use of tools to support data collection and analysis with common infrastructure. The talk will discus these to share our experience with broad audience.

Profile: Tsutomu Kamimura graduated from University of Tokyo with BS and MS degree and from University of Delaware with Ph.D. He was an associate professor at University of Kansas from 1979 to 1986. He joined IBM Tokyo Research Lab in 1986. After conducting research and research management in software technology, he moved to development organization in IBM. He led Rational development team in Yamato, Japan for 2003 to 2007. From 2000 to 2002, he did software strategy work at IBM Somers, New York, and from 2007 to 2009, he worked on future version of change management tool for Rational at IBM Lexington Lab in Massachusetts. Since 2009, he works as Rational CTO for Asia Pacific. He has published over 20 papers in programming semantics, programming languages and software tools. He received Takahashi Award in 1987 from Japan Society of Software Science and Technology.

Measurement Impossible: How a Measure for Value Saved NASA JPL’s Software Assurance Program

Daniel Port, Associate Professor of Information Technology Management, University of Hawaii at Manoa

Abstract: Critical software systems destined for space are risky. Developers at NASA’s Jet Propulsion Laboratory (JPL) know first hand that a single software defect can result in mission failure or catastrophe. Because the stakes are so high, JPL has always followed the famous Russian advice, “doveryai, no proveryai”, or “trust, but verify.” That is, the software must be assured. Assurance provides independently performed efforts that go beyond testing such as activities as process compliance checks, artifact audits, and traceability validation. But while trust is free, these activities can be costly and their benefit, particularly in managing risk, is poorly understood. This inevitably leads to tough questions of value such as “Is this worth doing?” and “How much assurance do we actually need?”
For many years JPL, and indeed the software assurance community in general, has been grappling with how to rationalize investment in software assurance and answer such questions. Numerous works have suggested a myriad of value models, yet none justifiably link assurance activities to measurable benefits and outcomes. The result is that when JPL assurance managers were challenged to justify their budgets, they frequently found themselves first on the chopping block when cuts were made. The JPL assurance group was in as terrible downward spiral.
Yet there is a pervasive feeling among stakeholders that assurance is important. Recently we have discovered a simple measurement model of assurance value that identifies the connections between assurance activities and measurable benefits and outcomes. There are some surprising implications of this model. For example assurance value does not depend on the number of defects found, rather on the degree of coverage an assurance activity provides. This work carefully and systematically explores this value model and illustrates how it addresses tough questions of value through real-world, non-trivial value measurement examples, and ultimately saved the JPL software assurance group from extinction.

Profile: Dan Port, Ph.D. is an Associate Professor in the Department of Information Technology Management. He is a veteran in the computing industry with over 30 years professional and academic experience. Previously he worked in Professional Services at NeXT Computer, Inc. and served as the director of technology at EC2, the Multimedia Business Incubator Project at the Annenberg Center for Communications at USC. Dan has extensive entrepreneurial experience founding and participating in technology intensive startup companies. His primary research area is Value Based Software Engineering where he currently specializes in software assurance. He is an active visiting affiliate in the Software Product and Process Assurance group at NASA’s Jet Propulsion Laboratory and a frequent software assurance consult to JAXA. Dan teaches ITM and Financial Engineering courses at the Shidler College, where he has received best teacher awards.