Tony F. Hutchings,
Technical Director of Software Process and the Software
Engineering Technology Center

In the early 1980s, when the semiconductor and microprocessor industry was still relatively young, a few wise people recognized that the distinguishing factor for the winners in the race would be process, i.e., base technology, design methods, and CAD tools. They were right. Great processes are among the key reasons why Intel is today "top of the pile" and why our Alpha AXP chips achieve exceptionally high performance.

The formula works as follows: Brilliant, innovative people plus outstanding process produce consistently great results, repeatedly. This is in fact true of all product development efforts and is also therefore the case with software in the 1990s. We have thus devoted an entire issue of the Digital Technical Journal to software process and quality.

The most popular and effective models and methods for quality and process improvement hold several characteristics in common:

• All put the customer first, including knowing when customers and their requirements are being satisfied and when we and they are achieving desired results in the marketplace.
• All have a basis in applied measurement, using data from the application of the processes to help determine what changes to make.
• All are closed loop; that is, there is a clear path for feeding back observations to improve the current state of the process.

We are increasingly being asked: What is Digital's overall vision for software quality and process improvement? From a completely mature organization, the answer to that question would be something like the following: Every project sets its own clearly measurable, customer-driven quality goals; puts appropriate learning and improvement practices in place; continually monitors its progress toward its goals; and makes adjustments to process as needed to ensure it meets its goals. Fine words, but in reality we are not yet at that state in our corporate life. We have, however, developed a process-improvement strategy, or vision, which we hope will encourage all projects and groups to move toward the kind of state described above. That vision is best illustrated by the following diagram.

The strategy comprises three important concepts: Using Voice of the Customer techniques to implement the intention of being a customer-driven company; basing our process application on assessing our current levels of performance and therefore the opportunities for introducing new "best practices" to overcome our weaknesses; and continuously using quantitative and qualitative analysis to determine how we might achieve better and better results. Our Voice of the Customer concept embraces such powerful techniques as Contextual Inquiry (for understanding customers' work and what might delight them in the future) and Quality Function Deployment (for rigorously prioritizing customers' requirements and how to satisfy them with world-class product concepts). Our application of the Software Engineering Institute's (SEI) approach to improving processes relies on performing organization-wide assessments of process capability and on developing long-lasting improvement plans, drawing on the rich pool of best practices described in their Capability Maturity Model. Our notion of Continuous Improvements rests on empowering engineering teams to study the results of their work with measurable data, analyzing the root causes of any process problems, and systematically implementing improvements to their processes such that they achieve better results.

The relationship between these concepts is subtle yet vital: All our process work needs to be customer-driven, and yet these Voice of the Customer techniques themselves need to be open to improvement as we learn from real data coming from their application; the advice in the SEI's Capability Maturity Model is sound and we need to choose judiciously the most appropriate best practices according to the state of maturity of each organization; nevertheless, as these practices are applied, we need to learn what is working and what is not and adjust their definition and application accordingly; these practices should also be chosen, at all times, to maximize the benefit for our customers as well as for ourselves.

None of the three mutually reinforcing elements of the composite strategy is sufficient individually to drive the massive and sustainable changes we want to see in software engineering process at Digital. The SEI's Capability Maturity Model framework alone under-emphasizes the extraordinarily important and powerful Voice of the Customer and Market; the Voice of the Customer and Market alone provides insufficient structure on which to hang an entire process improvement strategy; Continuous Improvement alone, as likely to be practiced at Digital, is at a level of intervention too low to move entire organizations sufficiently quickly toward orders of magnitude improvement in productivity and quality.

How are others in the industry tackling the problem of improving their quality and productivity? Many of the techniques and processes which we are now mastering or planning to are also in use by other leaders in our industry. For instance, Voice of the Customer processes (as typified by Quality Function Deployment) are in regular use at Hewlett-Packard; Formal Inspection (called Peer Reviews by the SEI) is practiced at Hewlett-Packard, IBM, and a host of other industry leaders; the use of software metrics is commonplace at Hewlett-Packard and Motorola; Continuous Improvement teams abound at Motorola, IBM, etc.

We have made great strides in the past two years in the application of better and more modern quality processes in Digital's software engineering community. No longer is the notion of using Voice of the Customer techniques really contested; few doubt the cost-effectiveness of Formal Inspections as a defect-detection technique; there is a ground swell of support for the SEI's organizational assessment model and a belief that its associated Capability Maturity Model offers a rich source of really good advice on the steps to take to improve one's process capability; and so on. We are even beginning to compile case studies from within Digital that demonstrate the positive impact of these processes, techniques, and concepts on project quality and schedule. Of course, we need many more such experiences before we can say that we are truly "best in class" in these areas.

Readers may well ask how the various papers in this issue relate to the strategy described here. Different aspects of the application of our Voice of the Customer techniques are emphasized in two papers: Contextual Inquiry and Rapid Prototyping are discussed in the paper "Changing the Rules: A Pragmatic Approach to Product Development"; an approach to using Quality Function Deployment across different geographies is covered in "Defining Global Requirements with Distributed QFD." Examples of how we are applying SEI's assessment and Capability Maturity Model approaches are covered in "SEI-based Process Improvement Efforts at Digital." Another form of quality assessment is shown in the paper "Assessing the Quality of OpenVMS AXP: Software Measurement Using Subjective Data"; the business case for implementing SEI-like programs is covered in the paper "Modeling the Cost of Software Quality"; finally, in the paper "TP Workcenter: A Software Process Case Study," concepts are shown in practice: the use of a requirements analysis process, of defects metrics, and of overall continuous improvement.

Digital's software engineering processes are improving quite quickly and radically. To be completely successful will require a high degree of commitment and significant effort by management and engineers alike. The opportunity is, however, clearly there.