Wilkinson: Actually, I started in psychology. My area was psychometrics. I was in graduate school at Yale, and my advisor was Bob Abelson, who is actually a Fellow of the ASA [American Statistical Association] and is a statistical psychologist. A lot of his research is in social psychology. I went over to the statistics department and took a course with John Hartigan – Clustering, and he inspired me in every way because not only was the clustering course interesting but he was really interested in statistical computing. John has been one of the pioneers in that area, and also had a point of view that impressed me because it was very data-oriented – not rigid, ideological, and he was very tolerant of alternative points of view. And he was a very good listener. From there I ended up spending almost all my time over in the computer center. Yale computer center in the early 1970s was a very exciting place because unlike most computer centers, the person who ran it, that staff who ran it, let us near the machine. You remember a lot of places made users hand cards in at the window.

Aczel: Right, I was a student at Berkeley in those years. Were you an undergraduate or a graduate?

Wilkonson: I was at Yale as a graduate student.

Aczel: You were an undergrad at Harvard before that in the 60s.

Wilkonson: Yes, I went to Harvard Divinity School. My interest when I went to Yale was in going into clinical psychology and doing research and teaching in that area, but when I met Hartigan – things took a completely different path. My advisor also – we talked a lot of statistics. My advisor was Bob Abelson who, like Hartigan, had been at Princeton and had been influenced by John Tukey and the group, Tukey’s thinking about exploratory data analysis. There was in that generation a solid group of Tukey disciples. Abelson and Tukey actually did a paper together on multidimensional scaling.

Aczel: Then you took a faculty position? In psychology?

Wilkinson: Yes, I taught in the department of psychology at the University of Illinois. I was the methodologist there. I was hired in the methods and measurement division to teach courses in statistics. Then I took a sabbatical some years later and during the sabbatical I built a microcomputer.

Aczel: By yourself?

Wilkinson: There were no commercial ones. The Apple hadn’t been invented yet.

Aczel: The IBMs came much later.

Wilkinson: Yes, the IBMs were 82…83…something like that….84…

Aczel: How did you build it?

Wilkinson: It was a kit. There was something called a Cromemco and it was built here [San Francisco] – the manufacturers were here in California. It turned out to be the same computer that Jerry Pournelle in Byte magazine wrote all his columns on. The components were produced by Bill Godbout, one of the early hackers, manufactured right here in California. We put all the boards together, and soldered everything up.

Aczel: Was it like the Commodore that came with 16K, or did it have more?

Wilkinson: This actually was a very powerful machine. I still have it. When I finished with it, it actually had a megabyte of memory. Two large floppy drives, the big 8-inch floppy drives, with a megabyte each of storage space and lots and lots of serial and parallel ports and a printer and a plotter – I spent lots of money on this machine. I had been consulting at the University of Chicago and I spent all my consulting income on this machine. I had been consulting at the University of Chicago and I spent all my consulting income on this machine. My wife was kind of aghast. Later, when I started Systat, it all made sense. But what happened was that very early after that machine was going, friends and I were communicating across machines and also we had our machines looking like terminals to the mainframe at the University, so we were able to use the mainframe as the downloading printer and do all our computations in BASIC, and eventually I got FORTRAN and LISP and a variety of languages running on the system. I just got in deeper and deeper. I had written at Yale a fairly powerful general linear modeling program, so it did regression, analysis of variance, multivariate…

Aczel: How did you write it?

Wilkinson: I did that for my dissertation. I needed a repeated measures program and SAS didn’t have one at the time – nobody did – multivariate repeated measures. So I ended up just writing one and it got bigger and bigger.

Aczel: And what you wrote is pretty much what went into the Systat package we have today?

Wilkinson: Originally, yes. I took that program, it was called REGM – multivariate regression – and basically chopped it into little pieces because I had to fit it on a 64K microcomputer with overlays. The segments that had to be in memory at any one time were so complicated that I had to take snips of paper and lay them out on the kitchen table, and it covered a large area. What it meant was that I had in 1984, when I came here to the ASA, a system for regression, analysis of variance, manova, and so on that was basically as powerful as the mainframe versions. Since I had grown it from the ground up and they were coming from the top down they were not able to get theirs to run on 64K machines.

Aczel: That’s incredible! You wrote it at Yale on a mainframe but as you went to Chicago you were able to then transfer it to the machine that you built…

Wilkinson: Right. I had to tear the whole thing apart and that was the difference. It would not run on a microcomputer until I spent quite a bit of time devising the overlay structure, cutting the subroutines into very small subroutines.

I was able to develop a new code quite rapidly once I got the microcomputer. It’s interesting that it’s often quicker to write a new code than to try to revise the old code. So by understanding algorithms and then just sitting down and starting again, you have a tremendous advantage. That’s something I faced in the last couple of years because Systat has gotten larger and larger and I began to bog down and I realized we had to rewrite the whole system.

Aczel: But why did you have to rewrite it?

Wilkinson: Because the architecture of the machines and operating systems changes so fast that nowadays a program does not stay current if the basis for it is more than about three years old. You basically have to revise your core code every three or four years. Now, not many companies are doing that because it takes a tremendous amount of energy and renewal, and so there are some that are doing it and there are some that have great difficulty doing it. They have such a large code base. It’s just a law of nature that you have to rewrite every few years.

Aczel: It’s an incredible story. People read something like this in the popular literature like People magazine or Money magazine about Bill Gates and Steven Jobs. What you did, I think, as a statistician, is at least as amazing.

Wilkinson: In one respect I think I know very profile programmers, statistical programmers, around. I wouldn’t begin to say I’m the most proficient or certainly the most knowledgeable, but I do think in terms of the amount of code I’m capable of producing that I’m pretty skillful. I produce a lot of code in just a short time.

Aczel: So, going back to your development in that area, when did that idea first strike you that you should start Systat Inc.? Was it when you came to this conference 10 years ago?

Wilkinson: Back in 83. I expected to sell it. At the conference, I showed it at a poster session. Then I expected I would sell copies for source code plus copies for a low price.

Aczel: For personal computers?

Wilkinson: Well, not PCs but CP/M systems because there were no PCs in 83. They were just coming out. But basically like the Apple II and CP/M systems.

Aczel: You were the first with that? You came with disks then, right? And you were the first with disks? Nobody else had disks?

Wilkinson: No, that year there were two of us showing that year. There was Neil Polimus from Statagraphics. His system was written in APL. He had it partially running. I believe. I forget… it was on some high level microcomputer. The following year he introduced the PC version of Statgraphics and that was our main competition at that point.

Aczel: So there you were at that poster session. What happened?

Wilkinson: There was a lot of interest. And two people who were doing a review for Byte magazine – David Margenstein and Jim Carpenter from the Bureau of Labor Statistics – and they were just about finished. They said give them a copy, and I sent them a copy and a little xeroxed manual and lo and behold April rolled around.

Aczel: Then you knew it would make it?

Wilkinson: At that point I knew it would make it. My brother-in-law did a very cautious business plan, and we assumed the company would disappear by the end of 84.

Aczel: That somebody would buy you?

Wilkinson: No, actually I assumed I would go out of business. I would sell several hundred copies of it and then go out of business because everybody who wanted it would have it. I thought there’d be maybe 50 or 100 or 200 people who’d want to use this thing.

Aczel: So it was like a limited time company….

Wilkinson: Yes, I’m by nature conservative anyway – not politically – but I mean in that sense I don’t like to take big chances. April rolled around and I’d just put an ad in Byte magazine in March saying, Systat, but it didn’t have much effect, but then the review came out and the phone just lit up. It’s the April 84 issue – I’ll never forget it. It came out March 20^th or something and I got a panicked phone call from the bookkeeper. She said, "I don’t know what to do. The phone is ringing off the hook. You’ve got to come back home. This is terrible…" It was a business phone down in the basement but it had call interrupt and I’d say "please hold" and I’d go to the other one and it was just unbelievable. So we started to package these things. I a made a manual and got a looseleaf cover for it and I would trundle off to the post office and mail these packages.

Aczel: You were producing them in the basement?

Wilkinson: Yeah, at that point I was literally hand-assembling these little disks, copying them in a little computer. Every disk, I copied, stuck a label on, made the whole thing, and literally racing off to the post office. I wasn't using UPS yet. I’d walk up to the window with a stack about this high. The woman behind the window got to know me – she was very nice – and one day she asked me: "You’re not insuring these. How much are you charging for these things?" $400. She just was absolutely flabbergasted because she had seen me coming with stacks of these things and she knew it was a computer program. Well needless to say, we started expanding. I moved out of the house. I almost broke down. I was hysterical. I called an old friend and I said, "I don’t know what to do." He said, "I know what you do. You hire some people."

I literally had the White House calling. At one point this guy had this very deep voice and was saying "This is the White House," and I said, "Can I call you back?" "You can’t call me back, this is the White House. I want to place an order." It was amazing. People sort of yelling at the other end of the phone, "I want my package now. Do you know who I am? I’m Jet Propulsion Labs," and that sort of thing. It was a very hot time. People had just gotten these microcomputers, were told they could actually do things with them, and then read that there was software that would actually do this. They were desperate to get off the mainframes. So, anyway, we grew then. We moved into an office in a bank and then we moved again to a larger office and then in 1987 we moved to where we are now, large offices in Evanston. A very nice building in Evanston. I now have 50 people in the company. It’s been fun. One of the things is people. I’ve really learned about management and what not to do and how to get people who were skilled. I had to find and hire people who were good managers and every company you know goes through these stages that we’ve gone through. I think we’ve successfully managed to get past the entrepreneurial stage into a second stage where we’re now working in teams.

Aczel: Programming in teams…

Wilkinson: Programming in teams. The marketing department’s become very efficient. We’ve focused a lot on academic marketing. We have a very, very active documentation department. I have some terrific people in tech support. I’ve always had this free floating kind of work environment. All of the offices are glass so you can look right through anybody’s office out to the lake – Lake Michigan, and everybody walks around in socks. It’s very pleasant, very casual. We’ve managed to move into more formal production plans without losing that informality.

Aczel: The Silicon Valley brain tank approach

Wilkinson: Yes, very much.

feedback form | permissions | international | locate your campus rep | request a review copy

digital solutions | publish with us | customer service | mhhe home

---

Copyright ©2001 The McGraw-Hill Companies.
Any use is subject to the Terms of Use and Privacy Policy.
McGraw-Hill Higher Education is one of the many fine businesses of the The McGraw-Hill Companies.