Home > Articles > Programming

An Interview with Watts Humphrey, Part 16: Semiconductor and Software Quality and the Revolution of CI-105

  • Print
  • + Share This
In this transcript of an oral history, Grady Booch interviews SEI Fellow Watts Humphrey. In Part 16, they discuss Humphrey's vision of the future for IBM circa 1972 and IBM's push to improve quality on everything from semiconductors to installations.

This interview was provided courtesy of the Computer History Museum.

See the entire interview.

From the author of

Vision of the Future

Humphrey: Now another thing I didn’t mention, when I was on corporate staff, remember I had the job or I was looking at the IBM Business System?

Booch: Right.

Humphrey: One of the things I was asked to look at was where we were going as a business. We had various task forces looking at stuff. I remember we had a task force looking at terminals and communications and stuff. I ran a host of these things in terms of the future. I remember somebody saying, “How do we define an intelligent terminal?” Remember, this was in a meeting in about 1972 or something like that in corporate. I said, “The definition of intelligent terminal, I believe, is one that will run MVS.” People laughed. They thought that was crazy, but it’s true. But what was interesting to me was--

Booch: You foreshadowed the notions of workstations.

Humphrey: Yes, exactly. But I could see the networks coming. I could see the whole idea of vastly increased power and remote workstations. I could see the enormous needs for compatibility and dynamic networks and all that kind of stuff and the tremendous growth of programming. In fact, I was arguing strenuously when I was talking about the cost, customer’s cost. Where did the customers’ money go? Customers’ money basically was going increasingly to software. I showed these charts of where their money was going in terms of spending on support systems and application systems. They weren’t spending it on hardware; they were spending it on writing programs and installing operating systems. They weren’t spending it on buying programs, because they weren’t priced. We just started to price them. There wasn’t much money in it yet. But the cost -- more than 50 percent of their cost -- was now in stuff other than hardware and it was going up exponentially.

So my contention was we were looking at the wrong target. We had to move. No reaction, no interest whatever. But even though I could see that, I couldn’t put together enough of a picture to make a case. I must admit, the future surprised me as much as everybody else. I wasn’t the only one. I had a few allies, but not a whole lot, way back in those days.

Technology Assessment

Okay, now I was on corporate staff. This was in 1979. It was getting kind of late. I talked to my boss and I had been working with the lawyers. I was involved in all the lawsuits and I was advising the lawyers and all this stuff. But I wasn’t building anything. I wasn’t learning anything. I wasn’t doing anything. I had a modest staff and a big, prestigious job, but it wasn’t really terribly interesting anymore. I’d sort of been there long enough. I learned a hell of a lot. So I was talking about where to go next, and I was saying I would like to go back and run a lab or do something like that. I knew I wasn’t going to be king anymore. So I got a call to see if I would come down and work for a guy, Art Anderson, who had just taken over as the new group vice president running all the development divisions -- development and manufacturing. Anderson actually came out of research. He was kind of a funny guy to go into the job. I don’t think he really did it very well, but he was a very sharp guy.

So I went and talked to Art. He wanted me to be director of technology, basically technology assessment. I guess that’s what it was, Director of Technology Assessment. He told me about how they’d done assessments in the various research labs. Basically, it was a way to get groups to assess themselves of where they stood versus technology versus what was going on in the world. Were they really doing superior work? Where could they improve? It was a way to get groups to sort of wake them up and think about how they could move and what they could do better. So I agreed to do it.

I joined him, and I had to build a small staff of folks. I got people out of various labs to come and work with me temporarily on a one year assignment. I got people from Europe as well. I would get them out of the country managers. I’d talk to the country managers and see who they had, they recommended. It was an upcoming person. I got a guy out of Italy once. I got one out of the Netherlands. I got one out of Japan. I had people out of various labs in the U.S. So I had a bunch of folks.

Booch: If I can ask, where was this office physically set up? Where you working at this time now?

Humphrey: It was in White Plains, New York.

Booch: Okay.

Humphrey: It was the headquarters of all the development folks.

Booch: Got it.

Semiconductor Quality

Humphrey: One of the things that I hit almost immediately turned out to be very interesting. I learned a lot about it, because I learned what assessments were and how powerful they could be -- self- assessments. We learned, by looking at some of this stuff, that the costs of our semiconductor chips, as manufactured in our Burlington, Vermont plant, cost us more to build than it did to buy imported chips from Japan. That was just not competitive at all. So Art and I went up to Burlington and met with the management team. He took a very straightforward position. That was either you can get competitive in manufacturing chips or we’re going to close Burlington. Burlington, Vermont must have had 30,000 employees.

It was an enormous place -- made all of our chips -- but they weren’t competitive. They did a self-assessment. I went back and forth helping them pull it together and learned a hell of a lot about chips and chip manufacturing and quality. They actually brought the yield of chip manufacturing up to a very high number. I don’t know what it was. I certainly don’t remember. I probably knew at the time, but their yield had been like 20 percent. I think it was up to like about 80 percent. It was purely quality management. That’s all they did. But they really went through it. They crawled through it all. They looked at what everybody else was doing. Of course, the actual yield achieved was a highly classified number. Nobody would tell it. It was competitively extraordinarily important. But they actually got to where they were the lowest cost manufacturer in the world. They did it fast. It was amazing to me how quickly they were able to absolutely change their stripes. They became the leader in memory chips for some time.

I learned a hell of a lot about quality and quality management and the W.E. Deming stuff and how you manage quality. That served me in good stead later with the software community. As a matter of fact, one question came up and that was: Could we take any of the lessons from that and use them in software? Art asked the question. As I say, Art had some very interesting perceptions. So we pulled together a group of experts from software and the semiconductor group and had a session on it to see was there anything we could gain from that, between the software and semiconductor guys. The software reaction was, “No, there’s nothing we can learn from this.” There was no real interest in looking any further.

Another thing: I mentioned Gartner, Gideon Gartner. While I was working for Art Anderson, increasingly we were having people leave to go work for Gideon Gartner, who just started this little group. They were basically a group spying on IBM, trying to get clues as to what IBM was doing. A good friend of mine, Tom Crotty, had joined Gideon Gartner. IBM was very upset with Gartner and they were trying to figure out, “How can we shut these guys down? They’re doing something illegal, etc., etc.” So Art asked me to go meet with Gideon, which I did. I had quite a discussion with Gideon on this whole thing. Interesting guy.

As you know, the Gartner Group is doing a lot of stuff these days. But I’m sure they’re still doing much the same. They were sort of a spy shop, checking on what everybody was doing. But nonetheless, I met with him, but I didn’t really make a whole lot of headway. We ended up suing him anyway.

Next, Art Anderson was moved on. I don’t know where he went. Jack Kuehler replaced him. Remember I mentioned that Bob Evans had said there were two guys who were on the list to be IBM president/CEO down the road?

Booch: Right.

Humphrey: It was me and Jack Kuehler. When I was running the Endicott lab, he was running the Raleigh lab. A great guy. We worked together a lot, because we did a lot of printer interaction work, and so we spent a lot of time working together with our groups. We had to develop a printer for one of their terminals. The people couldn’t agree. So basically, Jack and I said, “Okay, we’ll put them all in a room.” So we shut them all in a room in Endicott. It must have been a half dozen of his guys and a half dozen of mine. We basically said, “We’re locking the door and throwing away the key until you agree on a strategy.” So we went in and met with them periodically. But we gave them really firm marching orders. Every so often they’d come out with problems and questions and we’d keep going.

Booch: Did you burn some little things so that it either made a white smoke or a black smoke when you were done?

Humphrey: That’s right, pretty much -- the idea of just locking people up and giving them a clear charter and you’ve got to go back and help them. Then frequently they needed some motivation and some facilitation. But they did a great job. Did I tell you the PC story?

Software Quality and Process

Humphrey: Art Anderson moved out and Jack Kuehler now came in as group executive running all of development. So Jack called me in and said, “Watts, what are you doing here?” I was on this technical assessment job, Director of Technology Assessment. I said, “Frankly, with Art gone, I’m not sure.” He went through a career talk. “What do you want to do?” At this time, this was about 1982. So I was about 55 years old and Jack was the same age. I was off the promotion track. He basically was asking, “What do you want to do for the remainder of your career?” I said I really didn’t know. He said, “Well, let’s think about that.” He said, “Give it some thought,” so I did.

Booch: Did IBM have a mandatory retirement age at that time for its executives?

Humphrey: No. Senior execs had to retire at 60, like the CEO. I don’t know if the senior group executives did or not. But the CEO, Tom Watson, had basically decreed that. A lot of people thought that he put it in place so that Learson would be forced out in two years. I think Tom was scared that Learson would be a little roughshod and he was. Tom had real insight. So I think it was just the CEO had to retire at age 60. Everybody else was 65. In any event, I had five to ten years to go. I didn’t know what I wanted to do. So Jack said, “Well, look, there are a couple of things. You could either run one of these programming shops we’ve got that are running the lab computer shops.” He said, “That’s one. Let me think about some others.”

A few days later he came back and said, “One job I’d really like you to consider is to take over as Director of Software Quality and Process.” It’s the group in Poughkeepsie. We were living in Darien, Connecticut at the time. It would have been impossible to commute. But I looked at it and thought about it and I said, “You know, maybe that isn’t a bad idea.” So I agreed to do it and I did. We didn’t move to Poughkeepsie. Barbara never wanted to move to Poughkeepsie. So we moved back to Chappaqua and I had about an hour commute each way to get to work in the morning. I had about 70-some people, included a bunch of people working on process work, quality work, etc., some very good folks.


It was a fascinating staff job, but it got me back into programming and it got me back thinking about programming quality. There were a bunch of things involved with that. One was fascinating. It was called CI-105, Corporate Instruction 105. IBM had put out a corporate instruction on quality. It was for the company’s products, all products. It required that each new product have better quality than its predecessor IBM products and better quality than the best competitors, the leading competitive products.

Booch: Did you have a sense for how to measure quality then, so you knew what you were going against?

Humphrey: Good question. The hardware guys had a measure that everybody had been using. It was called RAs or repair actions. The maintenance community that did the hardware maintenance had lots of measures of RA. The thing that was absolutely astounding about actually measuring and tracking quality and managing it step by step, they had to project what the quality was going to be before they could announce the product. They had to show what they were doing. People had to review and concur with it. They tracked it. When people missed their quality goal, they had to come in with action plans. And so it was really managed and people were really paying attention from the top. On one file system they had a quality improvement in RAs of over 1,000 times.

The numbers were absolutely extraordinary. They had another thing with the shipment of IBM’s largest machines out of the factory. For instance, out at Kingston they’d ship a large computer system -- it would fill a 42-foot moving van. It was a lot of stuff, and that’s how they moved it nationwide -- in big moving vans. Worldwide, they typically went by air. It cost too much to let it float around. But when they’d arrive, they’d go to put them in, and they’d be missing stuff, and things wouldn’t fit and all kinds of stuff. So they started putting in tracking on the installation quality on these systems. Typically at that time it would take a week to 10 days to install a new machine.

Amdahl at that time was installing machines in about three to five days. It was an enormous problem. When you have a great big system and a machine room, people can’t shut their whole business down for 10 days. Most of them don’t have backup systems. At least they didn’t then. They don’t have all the machine rooms, so they really had to be able to pull them out and put them in quickly. So they set a goal, and the people in Kingston thought it was crazy. The goal was we’re going to have defect-free installations. They were to track every defect on every installation and come back and figure out how to find it or prevent it. What was amazing to me, before I retired, which was in 1986, about 80 percent of their large system installations were going in defect-free. They would arrive at the customer’s office on late Friday and they would be up and running Monday morning. It was an extraordinary achievement. And they did it with a very routine way. They just basically measured every defect. They went back to find out what caused it or how they could find it earlier and they laid in procedures and practices to put that in place. They basically went through with that and that’s what they did. And boy, did it work. That was a marvelous lesson for the software community. So CI-105, basically that was what they were after.

  • + Share This
  • 🔖 Save To Your Account