About Erwin Tomash
Erwin Tomash was born in St. Paul, Minnesota, and grew up in that state. He graduated from the University of Minnesota with his electrical engineering degree in 1943. After graduating, he went into the U.S. Army Signal Corps, where he worked with radar and was awarded the Bronze Star for his wartime activities. After being demobilized from the war, Tomash spent a brief time with the Naval Ordnance Laboratory and then joined Engineering Research Associates. There he worked on developing electronic computers, including the ERA 1103 or UNIVAC Scientific. He also studied for an M.S.E.E. at the University of Maryland, which he received in 1950.
In 1956 Tomash joined Telemeter Magnetics, a Los Angeles-based company. Soon he became the company's president and oversaw Telemeter Magnetics' design of core memories for computers. In 1962 he left Telemeter Magnetic, which had been bought by Ampex, and co-founded a new company, Dataproducts Corporation. Dataproducts specialized in computer technology, especially printers, and by 1970 had become the world's leading independent printer manufacturer. Since that time, Dataproducts has successfully competed in the core memory field. Tomash died 10 December 2012.
The interview spans Tomash's career, focusing on his years with Telemeter Magnetics and especially with his company Dataproducts. Tomash describes his education and early experiences with radar and radio, and then moves on to discuss his computer work. He explains his reasons for forming Dataproducts, and talks about the challenges and rewards of running one's own company. Tomash provides detailed descriptions of Dataproducts' successes with computer core memories and printers. He also discusses his opinions about the relationships between government and the computer industry, on research and development priorities, on the importance of innovation and long term investment, and on the challenges of financing new products development. The interview concludes with Tomash's views on international trade and future prospects for the computer industry and the world market.
About the Interview
ERWIN TOMASH: An Interview Conducted by William Aspray, Center for the History of Electrical Engineering, 19 June 1993
Interview #164 for the Engineers as Executives Oral History Project, sponsored by the Center for the History of Electrical Engineering, The Institute of Electrical and Electronics Engineers, Inc.
This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.
Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, IEEE History Center at Stevens Institute of Technology, Castle Point on Hudson, Hoboken, NJ 07030 USA. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.
It is recommended that this oral history be cited as follows:
Erwin Tomash, an oral history conducted in 1993 by William Aspray, IEEE History Center, Hoboken, NJ, USA.
Interviewee: Erwin Tomash
Interviewer: William Aspray
Place: Los Angeles, California
Date: 19 June 1993
Education and career overview
Perhaps we can begin by having you recount your career, starting with your education and going through your work experience.
I was born in St. Paul, Minnesota. I attended the University of Minnesota, which is located next door to my hometown. I graduated high school in 1939 and went right on to the University. I received my degree in electrical engineering in 1943. Because of the war, the courses were accelerated courses, and I graduated in the spring of 1943. In those days, the study options in electrical engineering were limited to power and radio. There were no other choices, and I selected the radio option.
Immediately after graduation I went into the Army Signal Corps. I was commissioned as a second lieutenant and sent to radar school at MIT and Harvard. I spent about six months there. Radar school was an eye-opener. It was my first introduction to more modern electronics. It was also a very good experience. I found I was able to compete quite successfully with the hundreds of young engineers from schools all over the country who were also going through these courses. It gave me confidence in my education and was the first indication of my ability to perform as an engineer in the field. I was in the Army until 1946 and spent most of my time in the European Theater of Operations. I did nothing particularly remarkable, though I was awarded a Bronze Star. But I'm not truly a combat war hero.
On my return to the U.S. after I was demobilized, I returned to the University of Minnesota planning to get a master's degree, and also to be an instructor. The GI Bill was then in force, and the University was swamped with returning GIs. They were eager to find graduates to help them. However, I learned very quickly that I didn't care for that and left the university after a couple of quarters. I then took a job with the Federal government in Washington, DC at the Naval Ordnance Laboratory (NOL). NOL was then in the process of moving to its new home at White Oak, Maryland from its wartime location at the Naval Gun Factory near downtown Washington. I quickly found that I didn't like working for the government. I didn't like all the routine paperwork, and I didn't like the measured pace of the work.
Within a year I left and found a job with a small company called Engineering Research Associates. The company headquarters was in Arlington, Virginia and to my surprise I found that they also had a factory in my hometown of St. Paul, Minnesota. Once I got security clearance, I discovered that the company was building electronic computers. I settled right in. I liked the work from the start and I spent long hours on some of the early vacuum tube electronic computers, which were then being designed to be used for cryptographic purposes. I stayed with ERA in Washington for two years and then transferred to St. Paul. During that time at ERA, I learned a lot about myself, my capabilities, and my inclinations. I found myself drifting towards the generalized rather than the specific. I found I didn't enjoy and didn't excel at circuit design and detailed component evaluation, and other parts of the nuts and bolts of design engineering, perhaps because I was not very good at it. I did find I was quite good at system design, problem analysis, logical design of machines, organization of projects and so on. I soon became a project engineer and then a troubleshooter on a variety of programs and projects.
I also found that unlike, a number of my peers, I was able to express myself pretty clearly and describe what we were doing to outsiders. I seem to have a natural tutorial inclination. I don't enjoy delivering a prepared lecture but I can give a talk or make a presentation from just a few notes. The ERA management called on me often to do just that, which ultimately led me into marketing. In the 1951-52 period I was a project engineer on a major computer development known in-house as Task 29. This later became known as the ERA 1103 or the UNIVAC Scientific computer. I was asked to make a presentation to Remington Rand management on this Task 29. It was actually a fait accompli, a hidden and unknown bonus they had acquired when they bought Engineering Research Associates. In the presentation, I suggested that Task 29 be commercialized and turned into a commercial product because we felt it was superior to anything IBM had at the time. Remington Rand accepted this proposal but on condition that I be assigned to help to sell it. I agreed and in 1953 I moved to Los Angeles to open the first West Coast office for Remington Rand Electronic Computers. My choice of location was excellent and the new office was quite successful until it was bought by Sperry in 1955. I left not too long after the purchase by Sperry because for my next career step Sperry Rand wanted me to move to headquarters in New York and become a full-time marketeer. I really didn't want to do that. I didn't want to move my family. We liked California and wanted to stay here.
So I left Sperry Rand in 1956 and joined a small leading-edge company in Los Angeles that was designing add-in core memory systems for computers. Core memories were then very, very new and they had not yet been incorporated into any of the major product lines. The company was called Telemeter Magnetics. I joined it to head up marketing. It was a very small (less than 25-man) company. Within about six months, the president quit and I was asked to replace him. So without any experience in general management I became president of a small, high-technology company.
Telemeter Magnetics (TMI) was owned by Paramount Pictures. How they had gotten into this is a long story and not worth going into here. Suffice it to say that I worked hard to build the company and in a few years TMI was a nicely profitable company doing seven or eight million dollars' worth of business. It had become one of two leading independent suppliers of core memories in the country. In 1959 TMI went public, that is we sold stock to the public--what is today called an IPO. A year or eighteen months later--in 1961--Paramount decided to sell its remaining interest, which was still a controlling interest. We ultimately merged with Ampex Corporation. So Telemeter Magnetics became the Ampex Computer Products Division and I became a vice president of Ampex.
Soon afterwards, Ampex itself went through some business difficult times followed by a management shake-up. Within a year, the people with whom we had made our deal were all gone. A new president was brought in to run Ampex and he had completely different ideas. I soon left Ampex. In 1962 I started Dataproducts Corporation. Six or seven key people all from the original Telemeter Magnetics joined me. We decided not to go into the core memories business because Ampex and others were doing a good job in that area. Instead, as Dataproducts Corporation, we planned to supply a variety of peripheral equipment to the then rapidly emerging independent computer industry. We visualized that we would produce a whole set of different products lines. One would be printers. Another would be disk drives. A third would be punch-card equipment. A fourth magnetic tape units, etc. We had a long list of possible products that we thought we could supply to the industry. We did indeed follow that strategy and in the next few years built the company in conformity with that vision.
After about five years we had four or five product lines going and it also became clear that we were making much more rapid progress in printers than in any of the other product areas we had selected. In particular, we were not keeping pace technologically in disk drives, where IBM had been making tremendous investments. Very rapid evolution in disk drives ensued and we were not able to keep pace. By 1970, through phase-outs and sell-offs, Dataproducts Corporation ended up as a specialist printer company. By then we had changed our name to one word and had become the largest independent manufacture of printers. We sold our products all over the world. At its peak Dataproducts was about a $500 million company, and had about 5,000 employees. I retired from active management about ten years ago. Just before I did so, I founded the Charles Babbage Institute. That's the story of my career.
Telemeter Magnetics (TMI)
Maybe we should go back to Telemeter Magnetics. Why don't you tell me about some of the management challenges you experienced? What were the really problematic things for you?
When I came there, Telemeter Magnetics consisted of three or four very bright engineers, several technicians and a small production crew. It had no product line. It had no marketing or finance departments. It was, in effect, a small project team working on two or three government contracts to add core memories to existing vacuum tube computers. As I said, the company was owned by Paramount Pictures. All the contracts had been underbid and were losers. My challenge was to turn the project team into a business. My first steps were to introduce some marketing concepts and to get the rudiments of a financial structure in place. I knew a little about technical selling, and a little about marketing. I didn't know anything about finance. I did recognize we needed a good controller. It was clear that we needed to know what projects cost, and how we were doing financially. I've never had a course in accounting or finance. However, I feel that anyone who has had a decent technical education, including the usual mathematics courses, can quickly grasp the rudiments of accounting and also how to read and understand a balance sheet.
One experience that I like to mention is the course in electrical measurements that Otto Schmidt taught to engineering students at Minnesota. His second lecture was entitled something like: "Beware of the difference of two large numbers." In it he pointed out that small percentage changes in either or both of two large numbers can make a gigantic percentage change in their difference. Of course, the standard accounting profit-and-loss statement represents the difference between two large numbers. Income is one large number and expense the second large number. The difference is the profit or loss. I soon came to understand that a very, very small change in either income or expense can generate a large swing in the profit--the measure of performance.
My basic challenge at Telemeter Magnetics was to set a marketing strategy and thus determine a course for the business. I soon determined that we needed a product line in addition to the contract work. What TMI had been doing was limited to helping people update their computers. For example, the Rand Corporation had designed the JOHNNIAC memory using Selectron storage tubes from RCA, but then RCA decided not to produce the Selectron in quantity. Instead, TMI built a core memory for JOHNNIAC. We also built core memory for an IAS-type machine at the Aberdeen Proving Ground which was, I believe, called ORDVAC. It was clear to me that add-on memories to update older machines was not a viable business. We did get another contract or two, as there were still six or eight different custom computers in use, but the need for computing power was rapidly being filled by "products" such as IBM 701s or UNIVACs. We clearly were not going to be able to build a business supplying IBM and Univac, who already built their own.
However, there were a number of companies entering the then-emerging computer business, and it did appear as though there was a business opportunity in selling core memories and supplying the cores themselves. Also in providing what we called "stacks," which were the cores assembled in planes called "arrays," and then stacked together to form the main memory component. As a first step, to become known as a leader in the market, I decided on a simple product, a small but complete box of memory that we could sell for $10,000. Even a small memory was sold for nearer $100,000 in those days, and I wanted to impress and indeed define the marketplace by offering a memory in the range of $10,000. At the time, 1956, transistors were just becoming available as production devices. Transistors were not yet being used en masse in computers, but there were several transistor computers in design. So we decided that our first product should be transistorized as well as low in price.
We settled on a unit with a thousand words of memory. It was contained in a metal box about 12 inches high that could be mounted in a standard, 19-inch rack. It included its own power supply and had registers for data input/output and the address. It was as we planned--a black box of memory that sold for under $10,000. To reduce costs, we did the address selection using a magnetic selection system and ended up with 1092 memory positions rather than the standard binary of 1024. The product was named the 1092BU8. (It had 1092 positions of binary digits each). The BU was meant to indicate that it could be used as a buffer store. That product launched TMI. We sold quantities of tens and hundreds to companies who designed it into their systems. The 1092BU8 established Telemeter Magnetics as a high-quality, low-cost core memory supplier.
Who were your typical customers?
Collins Radio bought units. They used it as a buffer store. GE was a customer, as were UNIVAC and Burroughs. We also sold units to laboratories. It was used to collect data from punched cards, magnetic tape and paper-tape readers. It was used to handle the data rate change between slow peripherals and the mainframe. It was seldom used as a central memory. But the product got us enough business to start a production operation. We were encouraged to learn that there was a memory market out there.
Then we got a big contract to do memories for GE for their ERMA system. Those two things, the new product line and the GE contract, gave us financial stability for the next couple of years. By then we had developed a whole range of core memory products, cores, core stacks, and complete memories up to very large ones. At the time Telemeter Magnetics was sold to Ampex in 1960, we were supplying very fast 1 microsecond access memories to Philco, RCA, and people like that. We sold memories to GE. We sold stacks to Burroughs, NCR, and UNIVAC. We sold to almost every computer manufacturer except IBM. We were really in the business of helping the nascent and budding competitors of IBM. Our challenge was to supply them with products which let them compete with IBM.
Did any of these computer system manufacturers other than IBM do their own work in this area? Were you a second source, for example?
Yes. We became both a sole source and a second source. Companies just starting into the business such as Philco or RCA or GE, tended to buy the complete memory answer they needed. Their challenge was to get into the marketplace quickly with a working computer system. If they could buy a reliable peripheral or memory, they did. In the second stage, a few years later, these same companies started to integrate more. That is, they wanted to build more of the system and buy peripherals and components. For TMI this meant selling cores and stacks. For example, Honeywell, which acquired the Raytheon computer activities and therefore had a start into the business, went that route and was a big stack customer of ours, but never bought memories. Very many companies--UNIVAC, Burroughs, and so on--had their own core-making and stack-making operations as well. Yet, they also bought from us. So to them we represented a second source.
Did you have any competitors that were independent core-makers or plane-makers?
Yes. There were several. One was the company that brought memory cores to the U.S. The square hysteresis loop was introduced into this country after World War II by General Ceramics (GC), a company in Trenton, New Jersey. Before and during the war GC was a standard products ceramic-maker which made a broad line of industrial and commercial ceramics. It was owned (at least in part) by German-Jewish investment bankers, the Arnholdt family. They came to America as refugees from Hitler and used General Ceramics as a base to build on here. After the war, a ceramist who had worked with the Armholdts in Germany joined General Ceramics. He had done research in Europe on a new class of ceramics-square-loop materials. Soon General Ceramics had developed and introduced these square-loop materials here. They also obtained strong U.S. patents on the material. They were a major factor in the business and certainly a leading maker of cores.
Core memory, the system using these cores, came from the work of Jay Forrester at MIT. There may be some small dispute over this. There was other early work by Jan Rajchman at RCA and An Wang at Harvard. The engineers at Telemeter Magnetics had worked with Rajchman, and the ceramists, the core-makers, had also come from RCA. RCA, of course, had a long tradition in technical ceramics. We had other competitors. Fabritek, in Minneapolis, got its start as a supplier to UNIVAC and Control Data. Even before the Ampex merger took place, a group of engineers and marketing people left Telemeter Magnetics and started a company called Electronic Memories. They also became a factor in the business. So we had Ampex as a major player as well as Fabritek, Electronics Memories, and General Ceramics. There were also a few smaller ones. Electronic Memories eventually bought General Ceramics. Our break-away group, a competitor, bought the originator and patent holder.
When computer system manufacturers were going out to buy these cores or the core products, what were the most important issues to them? Were they questions of cost, of reliability, of function, of speed of delivery, of reliability of delivery? What sorts of things were of most concern to them?
All of these were of concern. But of major concern were performance, uniformity, and reliability. Speed of delivery became important only later. Properly defining performance of the product was critical. Standards weren't yet established and specifications were uncertain. There were complex issues in how to test a core, and what the test indicated. It was unclear how to predict and specify core function within a stack and stack function within a memory. Stack design never did become standardized. So functional specification of core, stack, and memory performance was a critical matter as it related to the desired speed of operation. Reliability was primarily a function of the manufacturing process, but in order to get reliable and uniform performance from the memory system, you needed to start with very uniform cores. The cores had to be alike. Yet these were ceramic products made by an ill-understood process, an "art" that was lost regularly. I say "alike," but [Chuckling] of course nothing is truly "alike." Given that every core differed, within what tolerances must we operate? Those were the concerns.
Cost was important. One had to be able to earn a middleman's profit. That is, sell the core, stack, or memory at a price such that the OEM could mark it up to cover his costs and margins and still compete with (be priced beneath) IBM. The cost challenges we faced were in manufacturability of the stack design, and also in the making of the cores, but particularly in the testing. We had to invent and build the test equipment. It was nonexistent.
Much is made today of manufacturing techniques and technologies and of quality control issues. Are there things that you can say about these issues?
You have to characterize the type of market you're in before you can sensibly discuss these questions. Operating in an emerging market, one that's changing rapidly, is very different from operating in a more mature market. In the former, no standards exist and manufacturing, quality and process control have different meanings. For example, at Telemeter Magnetics we were shipping products while we were still developing the core manufacturing and testing process. Core technology was not yet diffuse and the processes were very proprietary. In the early 'fifties, the ceramists, the coremakers, were sort of cooks with tightly held secret recipes and processes. They themselves didn't understand what they were doing, only that this or that worked. Every core plant that I know of "lost the art" from time to time with the result that they couldn't make good cores for weeks or even months. Eventually they would get the art back and make good cores again.
Dataproducts went into the core memory business a few years after we established the company. By then, we were no longer concerned about being sued for using proprietary Ampex information. The reason we entered the core business was that some technical people came to us with a breakthrough idea for making cores more uniformly and at lower cost. We invested in their ideas, which worked out over time.
My point is that at the beginning of a market the issue is how to make the product and make it repetitively. Later on is when process and manufacturing engineering, comes in. And even later, when the fundamental technology is well understood, comes sophisticated test equipment, automation, and all the things that are normally done to improve productivity. Most of these steps can only be taken when the product and market have stabilized.
What about research and development at Telemeter Magnetics?
All of the R&D in all of the companies I've been associated with has been in the engineering area. It has been product development, not research. It has been a constant search for ways to make things, to do something. True research has been limited to the largest companies.
All growth companies do spend a considerable amount of their revenues--on the order of ten percent--on new product development. This is due to the rapid rate of product change as growth markets develop. Dataproducts always spent about 10% of revenue on engineering. We always felt we had to have a product advantage. When I introduced Dataproducts and Telemeter Magnetics to new employees, even in those early days, I said: "We can't get by with 'me too' products. We can't offer products like those of other companies and hope that somehow the customers will buy ours. We have to differentiate ourselves. There has to be a clear, fundamental difference that gives us a cost or performance advantage. We are not going to succeed based on volume. We are not going to succeed because we are nicer to deal with. We have to develop a sound technologically based sustainable competitive advantage."
Hence, for core memories we developed a superior process that was not just different. It yielded more uniform cores. For printers, we developed a print hammer mechanism that greatly simplified mechanical printers and made them more reliable. We sought to apply our efforts to achieving a fundamental lasting advantage. We did not concentrate on "manufacturing engineering" where one takes something someone else has developed and figures out a better way to make that part, but rather tried to innovate orders of magnitude cost or performance improvements.
In the development phases of a market sector real progress is made--real breakthroughs are made--by those who use new approaches to develop componentry and methods to make that componentry, and then design and build equipment based on these innovations. As the market sector matures, volume starts to matter. Dataproducts later on had great advantages due to economies of scale. But we never would have gotten to the point where volume mattered if we hadn't introduced printers that were much more advantageous cost-wise in the first place. Innovation got us into the business. In due course, our continuing advantage and staying power came from the economies of scale, which made it harder for other suppliers to compete.
When you were trying to decide at Telemeter Magnetics how much of your revenue to spend, how did you make those kinds of decisions? What were the trade-offs?
These decisions are really difficult. Management is under a variety of pulls and pressures. At least in the early days, I was able to arrive at and use a set of rules. We set targets. Our cost of goods was to run fifty percent, engineering would run ten percent, marketing would run ten percent, and G&A expense would run ten percent. That would leave twenty percent profit before tax to be reinvested in the business. Of course, we never could make those goals. G&A usually ran a little bit more. Engineering often ran more. Also, cost of goods rarely came in as low as fifty percent. As an OEM supplier we really weren't entitled to more than two times markup over cost. We had that lid on our price. One might ask: "why not just increase the price and have more to spend to do more?" We had to provide a printer to our customer which they could mark up (often double in price) because they were going to have to sell it, integrate it, provide the software, service it, and so on. They needed to at least double our price and still be able to compete with IBM. This meant that our costs had to be no more than one fourth of IBM's price. Roughly IBM's costs. We had to be able to manufacture a wide variety of products at lower volume and still match IBM's costs. That was our challenge.
Presumably some of your overhead was less expensive than IBM?
Oh, yes. Their overheads were higher. I was talking about manufacturing costs though. When IBM used a four times markup, their profits were large--but not four or five times ours. [Chuckling]
Their profits as a percentage of revenue were only slightly more because they were spending twenty-five percent for selling expense, fifteen percent for administration expense, and so on. Yes, they did have more overhead. As a percentage, bigger companies often have less overhead because of the volume, because of their scale of operations.
What was your relationship with your major customers, the DECs and so on? What role did they play in the design of your products, in determining your direction for business? What kind of interaction did you have?
In the early years, we had close business relationships, but not in the area of product design. The market was structured by IBM, and our challenge was simply to provide our customers with competitive products of high quality as answers to IBM. Especially in the early years, our product planning was done for us by IBM. A number of our customers had tried to differentiate themselves from IBM, but weren't able to succeed. The marketplace kept saying, "IBM is the standard." So the customers wanted products like those of IBM. They wanted simulation. The OEM's turned to suppliers like us particularly for the external peripherals. They limited their concern to the interface. We worked closely with them to make sure that the interface between us was convenient, economic, and non-duplicative. But by and large they left the insides of our printer to us. Occasionally there would be special needs. NCR, being strong in the banking business, wanted to be able to print OCR characters and the magnetic ink characters--the MCR characters--on checks. That need posed challenges to standard data processing printers, and NCR worked closely with us on that. They concentrated on specifying their need. I recall no instance where our customers' engineers worked with us on the design of the printing machine. They really just wanted a working answer for their system.
Core memory required more technical interchange as it was not a peripheral, but internal. Also there one found different levels of customer sophistication. Several of the customers were capable of designing their own memory systems and simply wanted a reliable second source. Honeywell was a case in point. We did business with them for years, particularly in Europe. We had a large plant in Dublin, and we were their major source for memories in Europe. They just wanted units identical to what they were making in the States. "And don't tell us about improvements." [Chuckling] On the other hand, RCA and also most of the smaller companies, would come to us for a total answer: "What can you provide? What we need is a faster memory." Or, "We need a cheaper memory." We would then work closely with them using the latest memory techniques. We would design in association with them and try to eliminate duplicated electronics, share test equipment, and so on. Anything to help their speed to market, to make it flow. So, sometimes the relationship was much more involved.
Essentially we remained an OEM supplier. Our marketing was the program needed to maintain liaison with our customers. We did no retail selling at all. Our marketing expense was therefore much smaller than had we been selling directly to end-users. We provided generic literature on our products to the OEM and we often printed it for him. We also offered training classes. We ran a quite large training program for service engineers. They would come to our plants, where we offered classes year round. But we didn't take field responsibility for the maintenance. That was up to the OEM.
What kind of technical training did your marketing people have to have?
They were almost all engineers. The reason for this was that our true customer was the systems engineer. Basically, we had to sell the OEM's engineering department on the superiority of our product. Their purchasing people would become involved only at the end. In later years, there would be more competition between printer companies. The engineering department would sometimes say: "Both these machines are acceptable and will work." But usually we were able to pre-sell the engineering department on our approach, and we were selected as the supplier. Then the purchasing department would come in, and they would negotiate terms and conditions.
As the company matured, the product line and the market broadened. We started to see printers used by communications and instrument companies. We then started using less technically oriented sales people who were not computer systems engineers by training. But we never used these people to deal with major customers like AT&T or DEC. At these large accounts, one really had to gain the confidence of the engineers. For larger systems deals, it might take two years to do one sell. But then, of course, that order might go on for three to five years.
Engineers as employees
What about the engineers you hired? What kinds of backgrounds did you want them to have?
From the technical training point of view, we were quite content with the output of any of the better engineering schools. We did a modest amount of recruiting each year, and we did a modest amount of connection-building with the faculties at UCLA, USC, and Cal State Poly. In general, we had great difficulty adequately staffing our engineering departments. We simply couldn't afford routine, non-creative engineering. We couldn't afford to do what is often done in major product programs, such as big government projects or major company programs--throw manpower at the problem. We expected a lot from our engineers.
I was an engineer by training and had a good feel for both market and product needs. But I did not have a good feel for detailed electro-mechanical design. Another of our founders, our chief engineer, the person who headed all our development engineering, was an outstanding and inventive designer. Fine engineers are very demanding people and are very seldom good managers. They tend to overwhelm, not develop others. They don't seem to be able to bring them along. It was a constant challenge for us to build quality technical teams that could get the job done. That was our experience in general. In particular, we found it almost an order of magnitude harder to get mechanical and electro-mechanical design done than to get electronic design done. The difficulty in electronic design lies in development and production of the components, not in the systems design, represented by the plug-in boards.
Of course, one can turn the designing of a board into a problem. It can be done poorly. But doing it well is not that difficult. For example, the proper design of the electronic system for a printer, as in the design of a computer, is based upon selection of the right components. The complexity and difficulty is in what Intel does. The next step is easier. It's how you put the components together in a system. Logical design, and software design do indeed have creative aspects. But engineering of the hardware of purely electronic systems is not a very difficult thing to do, as you can see from all the clones on the market today.
On the other hand, electro-mechanical design is extremely sophisticated. For the most part you cannot buy components; you design the components. So in this section you do have all the complexities that are inherent to what Intel does. Designing something that is low cost, that can be made, and is reliable is very challenging. There is an oft-repeated joke about mechanical engineers: a prototype is designed and built and during testing one of the bearings breaks down. The design engineer decides "We need a heavier duty bearing. During the next test, the heavier duty bearing holds up at which point the shaft breaks down. [Chuckling] They replace the shaft, and the next time the bearing goes again. The solution to mechanical problems, things that wear, that encounter friction is often subtle and complex.
Indeed, the reliability and cost breakthrough that Dataproducts made in its printers was due to its hammer mechanisms and the fact that we got rid of clutches and solenoids. A solenoid is actually a quite unreliable gadget. It has that clapper that moves and closes a circuit. It arcs in doing so. Over time the arc burns the contacts and that changes the distance to be traveled. So it doesn't close exactly at the same time or at the same point. A group of solenoids which you require to close at the same instant is a problem that haunted the telephone industry for years. This problem has gone away because of electronic switching. At Dataproducts we tried to get rid of all clutches and all solenoids. We tried to get rid of elements that wear and change with friction.
The best part, the most reliable part of our electro-mechanical design is the part that isn't there. That's a part that is not going to break and will never cause trouble. So we tried to reduce the number of moving parts to a minimum. This meant trying to understand the functions needed in a fundamental way. Doing this kind of design is very difficult. It's an art and is not taught very well. At least it wasn't taught very well in the years that we were hiring engineers. The path we found was through experience and on long apprenticeship. Engineers in this country aren't prepared to spend much time apprenticing. It is too easy to get managerial positions, to advance by working on large systems projects. We had a lot of difficulty finding electro-mechanical engineering skills. We did find good engineers from time to time and one at a time. We couldn't hire ten at a time, and when we found them we nurtured them.
Continuing education for employees
What about continuing education? Did you have any kinds of programs? Did you believe they were necessary?
We certainly encouraged all Dataproducts employees to continue their training and education. We had a company-wide program. We would pay the tuition for any employee that wanted to go to school. We encouraged people to get their Masters and MBAs. We encouraged engineers and technicians to take extension classes in engineering. We found in general that in purely technical matters, engineers learned on the job. We paid for any books they bought. We also brought in technical consultants. In one instance we thought optical fibers might have advanced to the point where we could use them to replace some electronics. We didn't know anything about fibers, so we brought in consultants to help us. It was that sort of continuing technical education. We didn't send everyone to school to learn about fiber optics. We remained essentially an electro-mechanical company. We certainly needed to keep up with the latest componentry, and to that end our people did go to conventions, seminars, and meetings where they heard papers. Also our customers were strong in electronics, and we would learn from what they were using.
When you were looking for those electro-mechanical engineers that were so difficult to find, were you likely to look to people just coming out of college, or did you look to other companies for experienced engineers?
Our experience was that we were better off hiring people after they had been out of school for a while. Not because they had printer design experience. There were only a few other printer companies and we respected only IBM. But because we found that there's a difficult transition period for engineers just coming out of school. We weren't prepared internally to nurture them properly, to deal with their uncertainties, and we weren't large enough to place than properly. We couldn't afford to cycle them through different departments.
Occasionally one or another of the professors we knew would tell us about a promising engineering senior, and we would hire him. If he was strongly recommended as a potential star, we would certainly make him an offer. Often the professors made their recommendations if they felt he would do better in a small company than in a big company. Otherwise, we found we did better with people we could recruit from the bigger companies who were at least two or three years out of school. They had settled down somewhat and had our idea of what engineering was all about. On the other hand, we rarely looked for twenty years of experience, although we did bring in a few sales people from IBM or from GE. They helped us because they brought with them knowledge of engineering discipline and organization.
How attractive was Dataproducts to a relatively young engineer? Were you at an advantage or disadvantage or was there no difference between going to a Hughes versus a Dataproducts?
I think we were able to compete. Salary and benefits were competitive. We were relatively attractive because we offered the individual engineer more control over his activities. They also were much closer to the decision center. They were aware that the work they were doing was important to the company. There were not several departments between them and the chief engineer, but perhaps only one person between them and the chief engineer. [Chuckling] Typically they were assigned a piece of a project and dealt with him directly on it. That was attractive. We were considered a good place to work. We were attractive to people who do not like to work on large-scale weapon systems and aerospace projects. We didn't have big rooms full of engineers' desks remote from the factory. Engineering was much more connected to the life of the company. So we were able to compete for people.
You alluded earlier in one of your answers to the fact that engineers often didn't make good managers. How did you solve your middle management/project management problems? How did you staff those positions?
Yes, we really did have a lot of difficulty with that. I can't say we ever really solved it. The best I can say is that we settled it from time to time. [Chuckling] We kept living with it. By the time the company was over $100 million in revenues, we had a more formal engineering department, a system of project reviews, and so on. By then engineering was certainly more structured and more organized, and we had technical people who viewed themselves as managers, not as engineers. I would say that after about ten years we had a group of very senior engineers, probably numbering half a dozen, who truly were creative engineers. Design engineering is what they wanted to do. We paid them well, and they had implicit authority, but they were not product or program managers. They actually continued to work on technical issues. There was the electronics guru, and the electro-mechanical guru, the optics guru, the design-for-production guru, and so on. We didn't try to turn them into managers. We had a set of program and product managers who ran projects.
We never really successfully solved the technical management problem of completing an advanced, innovative design and introducing it into volume manufacturing without great expense and delay. We tried many approaches but it remained a very difficult problem for us and we never satisfactorily learned to do it. Introducing products based on new technology into manufacturing is a major revolution more than a transition. A new version of an old product is easy. But a new approach utilizing new technology at the component level--that is a very difficult transition. I don't know of anything to do but work at it. [Chuckling]
Management and corporate growth
What about in your own role as a top manager? What kind of technical knowledge did you need? How did your background as an engineer affect the decisions you made? Did you feel you had the kind of technical knowledge to make decisions? Or did you have to rely on a certain set of people for that kind of advice?
Particularly for companies in newer developing markets, such as biotechnology today or the computer business in the 1950s and '60s, where there are as yet no standards, I think top management must have technical training. You must be able to understand what is going on in your field. You don't have to be a top-notch designer and you don't have to be a technology specialist. If the use of fiber optics is proposed, you don't have to take a course in fiber optics. You do have to be able to understand and judge what people are telling you. You need sufficient technical background to evaluate the business and program trade-offs, the risks. You are dealing with the future where it is easy to generalize and hard to be specific.
For example, there's no question that our increasing population will ultimately create a protein shortage in the world. Plankton in the sea is a great potential source of protein. Therefore there's no question that plankton is going to be eaten and used as a source of protein by mankind. Now the management question becomes, "should we open a plankton-canning factory?" To make that decision management must judge and understand all aspects of the situation regarding plankton. These are technical, marketing, financial, regulatory, competitive, etc. One cannot decide based only on the generalizations. One doesn't need to be an engineer to understand the generalizations. But one does need technical understanding in order to make judgements about the likelihood of success. In new, immature businesses, there isn't anywhere to turn for that.
My example may be poor in the sense that canning of seafood is not a new, immature industry. Today, the president of a canning company does not need to know a great deal about canning processes and canning machinery. Canning of food has been around a long time. The experience and knowledge base is diffuse and there are lots of sources of information--experts and consultants and so on. But when you're building a business in a new developing market that's where you need sound technical judgement. That is why start-up companies need key executives who are technically trained. Others, such as lawyers or marketeers, can take a company to the next stage or the third stage as the technology and market settle down. At that point the judgements can be based on established norms and general good business practice.
One of the most fundamental questions that torments management of every growth company is how to balance the future with the present? Do you operate to maximize profit in the short term? After all, in earning a profit you are creating a healthy, strong, more secure, financial base. How much of the profit should one spend on engineering of new products and thus create greater future strength? I tried to answer such questions.
The best answer I ever came up with on this was that we had a number of different constituencies to satisfy and that the answer was somewhat different for each of them. I started with the questions: What are we trying to do? That is, who do we work for here? I came up with five constituencies that I felt we were to serve. Their needs differ, their desires differ, and their priorities differ, but still we had to serve each of them every day. Customers were our major constituency. Next was our shareholders. Third was our employees. Then, we also needed to support our vendors, and serve the communities that we were in. Each of these viewed the company, Dataproducts, through a different prism. Your customers are not concerned directly about your shareholders. They want you to stay in business. Your shareholders should be concerned about the customers, of course. I could continue down the list. So what is management to do? Do you make as much money as you can now and return large dividends to the shareholders? Do you invest all profits into R&D and thus return little now but much more to the shareholders in the long run? That is, if you last. So management must try to balance the short and long term desires of the people they serve. Most of the constituencies would like you to do it all. They would like great short-term results with great long-term benefits. [Chuckling] Why not? And great companies sometimes provide these.
For instance, stock analysts are notorious for this. They come to the company to visit, and if you're quite profitable, they quickly ask, "Well, that's nice. But what are you doing about future investment?" But if you're investing heavily they say, "Well, what about your current earnings?" [Laughter]
The incentive and reward system that we have in this country is oriented to the short term. This influences management greatly. I feel this is a great weakness vis a vis the rest of the world. Even so, I've met many people who started companies and very few that did so just to make a lot of money. Everyone likes to make money, but stronger is the desire of the engineers and marketeers to build something. They want to build a company. That's often a stronger drive than making the money itself. Those dollar rewards may come with success but they're not the driving force they are often touted to be.
Engineering and production equipment
You mentioned earlier on that there wasn't the right kind of test equipment when you first were involved in Telemeter Magnetics. What were some of the other issues about tools for engineering and production that came across your way?
Components are probably the best illustration of what I had in mind. If you look behind a component manufacturing business, you'll find that the continuing challenge is in production testing. In order to test something adequately, the tester has to be an order of magnitude better than the testee. So if you are building a high-precision, high-performance whatever, the test equipment has to be high precision and high performance squared. It's a continuing problem as the product line advances.
The ferrite core is a good example. In the early days a core was made with a tableting machine, a machine borrowed from the pharmaceutical industry. The magnetic materials were ground up and mixed in the form of a powder. Then this powder was handled just the way they handle aspirin powder. The tableting machine had a small hopper in which the powder was placed. The hopper would then swing over a die and deposit an amount of powder equal to one tablet's worth of ferrite core. The top of the die would be brought down to form the toroid by pressure. The machine would then eject this formed core. These soft cores would then have to be taken to a kiln to be fired and turned into a ceramic. There were several processing problems that arose. Core memories require cores of much greater uniformity than this process easily yielded, and stringent production discipline was required. Not to mention the firing process in the kilns, which also added non-uniformity. So when the process is complete and we measure the magnetic properties of 10,000 of these cores that we've just stamped out and baked, we find a wide range of values and characteristics. A number of different things were happening. First of all there's a little bit more powder in one than another. You can lose a grain or two of powder because the machine was just dumping a charge in and around the die. Secondly, ferrite powder is abrasive. It's iron oxide. So the die is getting worn just a bit each time you use it. If you had made 10,000 cores with the same die, you can be sure that the dimensions of the toroid are different in toroid number 10,000 and toroid number one.
The other important dimension is the thickness of the toroid, and that depends on the pressure used in the die. At first we used a purely mechanical encapsulating machine taken over from the pharmaceutical industry. Later we used a pneumatic system, and we still found that core density could vary over a wide range. That is, some of the cores were packed more tightly than others. Then as they went through the furnace, did all the cores experience the same temperature? This depends on the cross-section of the kiln and whether the cores in the middle of the kiln got more heat than the ones on the outer edge, or on the trailing edge. Next is the question of cooling or quenching. Did some cores start to cool faster than others? You can see that in any complex manufacturing process like this, there are an endless number of places where the equipment and the process used influence the end result.
Returning to my 10,000 cores, before using them I must do 10,000 measurements. We had to develop mechanical equipment that automatically handled cores: that would select a core and probe it without chipping. The electronic equipment would run some kind of a test on it while it was held and then based on the test fed into a yes or no vial. In the early days, we found that the variations were so great that we had ten bottles to feed into. That is, we tried to get groups of cores nearly alike out of a wide mix.
There are two approaches to this type of problem: one I call the manufacturing engineering approach, and the other the development engineering approach. In the manufacturing approach, the engineer considers the defects of each step in the process and tries to cure them. For example, he puts an improved valve in the pneumatic system so it accurately supplies the same amount of pressure every stroke. Indeed, he senses the pressure, and uses a feedback loop to control the valve so that the pressure is constant. He automatically weighs each core after each time he handles it, and so on. He accepts that the core die is going to wear, so he counts usage and replaces dies on a regular schedule. In a manufacturing engineering approach you improve the process step by step through tighter controls and by engineering little artifacts and aids that help. If there is a problem with the heat profile of the kiln and its cross-section, perhaps he'll only use the center, or he'll redesign the carrier. If necessary, you redesign a new kiln, one that will redistribute the heat so that it yields a more uniform result. That's a manufacturing engineering approach.
The development engineering approach is to try to develop a new process. It says: I'm not going to use powder at all. I'm not going to use dies that wear. What I'm going to do is make a slurry of all this powder, put in a binder and roll it flat into a very thin sheet. I'll use two adjustable rollers. I can adjust those very accurately and know that well within the tolerances I need. I will have the same thickness and density all the time. I'll also put a lubricant into the slurry, one that will survive the heat of the kilns, and which will oil the dies when they cut. That way the dies have long life because I'm lubricating them. And I've got a kind of a cookie cutter process.
This latter process is indeed what we did. With it, Dataproducts went back in the core business. We set our competitors on their ear because our costs for a good core were one tenth of their costs. We didn't have to shake any powder. We didn't have to change dies. We didn't use a tableting machine. We cut thousands of cores at once. We didn't have one die; we had whole rows of multiple dies. By the way, we introduced a whole new set of problems that the manufacturing engineers had to come and help us with. But we did solve the basic problems in the powder manufacturing process by developing a way to get around them rather than fixing them. It's a challenge to balance these two ways of doing things.
Similarly, in the area of test equipment we went to two-stage testing. We could quickly measure with multiple-headed testers the rough magnetic characteristics. We could quickly and cheaply get rid of truly bad cores. We didn't know if the rest were good yet, but we knew at least they weren't bad. We got rid of bad cores on which further effort was a waste of time and money--perhaps five percent of them. The next step involved a whole new set of automatic core testing equipment. Here the optimum solution was to go the other way, to go to small inexpensive handlers and testers that we could adjust individually.
The correct answer to these issues is, of course, to apply the latest technology in both the test equipment and the process. This kind of thinking was quite traditional in the chemical industry. When we were doing this in the late 1950's and early 1960's it was a new set of issues for the electro-mechanical and electronics industry. So the net result was that we were able to create better, lower-cost components.
Management and competition
You mentioned a group of people leaving Telemeter Magnetics to start their own operation. I suppose one always has to count on a certain number of people leaving, but not necessarily leaving to form their own competitive company. Can you talk about the kinds of issues that you face as a manager when that kind of situation occurs?
It was particularly difficult for me because I really have a strong, deep entrepreneurial streak. On the one hand, I felt that we had built loyalty to the group and to the company, and that we treated people well and built an environment for team work and strong association. So it really hurt when it happened. I couldn't criticize, much less dislike, others for trying to satisfy their own entrepreneurial impulses. In the 1960's, start-ups seemed to happen mostly when people got uncomfortable with their situation. It happened less because they wanted to start their own company than because they were unhappy doing what they were doing. At the time, there was not too much venture capital around. This may not be a valid observation today. But I left Ampex unhappy and started Dataproducts because Ampex wasn't going in the direction I wanted. The fellows that left Telemeter Magnetics to form Electronic Memories left because Ampex was coming in, and they didn't like what was happening with Ampex, didn't like Paramount selling us out, and so on. In addition, I may have triggered one start-up by a reorganization in which I reduced the responsibilities of one individual and promoted two others. Whereas I had had one number two person, I now had a trio. The one who was previously the number two person left to form another company.
Dataproducts spawned several companies, some of them quite successful, some that failed. You asked me how I handled it. Intellectually, I felt as long as they played fair--and fair, to me, meant that while they certainly took with them the content of their heads [Chuckling] (you can't take that out), they didn't do anything unethical; they didn't take their project, their designs, customer lists, etc. They didn't bad-mouth us, and if they competed, it was indirectly. Under those conditions, I had a lot of trouble getting up a big head of steam over their leaving. Emotionally, as I said, I was always hurt and disappointed.
In the case of Pertec, which I think was one of our first Dataproducts breakaways, they went into the magnetic tape drive business. We didn't have products in that area. Later Pertec went into the disk business, but we had already gone out of the disk business. They didn't go into the printer business. In another case, the group went into the printer business and eventually failed. They took one of the design approvals that we had set aside as being not too successful, and they pursued it. They also took a few of our people.
In this case, we competed hard to protect our customers from their product. Really, I do think start-ups are a kind of tax that a successful company has to pay, [Chuckling] and it puts something back in the well. It's not too bad. It's a process similar to your children moving away and setting up their own place. As long as it isn't stealing, which I define as taking designs, taking customers, and so on, it's a beneficial process. I was upset for a while with the Electronic Memories founders because they did go into direct competition with us. We were in the core memory business. They did try to bad-mouth us. All of this was unnecessary, I thought. As I look back on it, maybe it was necessary for them to bad-mouth us in order to justify their own actions and to make themselves feel a little bit better about it. But all my annoyance faded away after a year or two.
Can you tell me about the start-up of Dataproducts? How did you make the decision to go into the business? What kinds of people did you look for to pull it all together?
First I had decided to leave Ampex. When I did so, I really didn't have starting a company in mind. I didn't know what I was going to do. But I was unhappy with the new Ampex management, and I had plenty of confidence in myself. It was in 1962, and I was 41 years old. Immediately after I resigned three of the key people in the company, the chief engineer, the head of marketing, and the chief financial officer, came and said: "Hey, let's start a company." [Chuckling] My first reaction was, what are we going to do for money? Most of us had held options in Telemeter Magnetics, and when it was sold to Ampex, all options were cashed in. So among us we had a little money, but we certainly didn't have a lot of money. I started to inquire around. I went to see people. I went to see an investment banker that I knew. I went to see the lawyer for Telemeter Magnetics, who incidentally has remained a strong personal friend all my life. I also contacted my old friends in Minolta. Almost everyone encouraged me. They felt we could raise money if we put together a business plan.
The next step was to decide what we were going to do. We decided we'd go in the computer peripheral business, and that we'd serve the systems industry rather than compete with it. We then developed a little theoretical model. We decided to see if we could get a head start by buying a company in existence and so gain a little momentum. I started to look at a few possibilities. I did this myself. I was the only person unemployed. Everyone else was still working hard at Telemeter Magnetics, which by then was a part of Ampex. It wasn't at all clear that we were going to be able to start a company.
Then Arnold Ryden from Minneapolis phoned me. He was a business associate of my old friend from ERA, Bill Drake. Ryden was chairman of Telex and was trying to expand it. Telex was originally a hearing aid company. Ryden had started a division to develop disk drives in St. Paul with some people from Univac. He asked me if I would come up and run Telex. They were looking for a president. I said, no, I wasn't interested in most of what Telex did and the idea of running a mini-conglomerate had no appeal. But in the conversation I said, "Why don't you sell the disk drive business to us?" They said, "We can't sell it. We've put a lot of money in it. It's important to our shareholders. We're a public company. It doesn't have much net worth. It will be worth much more later. Besides, you don't have much money anyway." It was true that we didn't have any money at the time.
A few days later. Ryden phoned again with a new idea; he was quite a brilliant financial man--"we'll spin the disk drive division off to our shareholders. We'll create a new company so that there'll be Telex the hearing aid company--the old Telex--and this new company which is the disk drive business. We'll sell you stock in the new company, which will get new money into it. Your group will have major ownership in a new public company." So that's what we did, in effect. We formed Dataproducts, put our money in it, merged with the disk drive part of Telex and then spun it off to the Telex shareholders. That's how we got Dataproducts started. We then raised about a million dollars from three SBICs to get the working capital for it.
Telex had a contract with GE to deliver disk drives, but it couldn't build them. That was our first challenge. Looking around Telex, we also found a printer project. This was located in Detroit, an operation with four or five people. They had offered a printer for $5,000. Printers were then selling for $30,000. We hadn't particularly thought of going into the printer businesses up to that time.
I went to St. Paul the day we made the deal and signed all the papers. Things were happening everywhere. We had inherited a union in St. Paul. They went on strike the day after we made our deal. April 1st we made the deal; April 2nd they went on strike. I was there to see what we really had bought into. GE was trying to cancel their contract because Telex hadn't delivered. I asked Graham Tyson, one of the key people from Ampex/TMI who joined us, to go to Detroit to see what was there. Two days later he reported. "There's nothing here. There are five or six people in total. They've got one printer they're getting ready to ship to DEC. It has never run. They haven't tested it yet. They haven't completely put it together yet." Tyson said, "I'm going to shut it down." So I said, "Before you do, are there any good engineering people?" He replied, "I think there are a couple of them that aren't too bad." So I asked Cliff Helms, who was our founding chief engineer, to go to Detroit and interview them before we shut it down. The next day he reported, "Graham's absolutely right. There's nothing here. If this thing ever does go together, which I doubt, it won't last fifteen minutes." But he said, "There's a germ of an excellent idea here for a print hammer mechanism. They're using a coil in a magnetic field, like a loudspeaker instead of solenoids. The way they're doing it, I don't believe will ever work. But I like the idea of getting rid of the solenoids."
The upshot of all this is that we did shut it down; nobody would move. It turned out there was only one engineer we made an offer to, and he didn't come. After we shut it, we started a printer program in Los Angeles under Helms' direction. The basic idea was that instead of a solenoid, which all printers then used, we would use a free floating coil in a magnetic field. A pair of magnets on either side of a moving coil. It was similar to a loudspeaker, but instead of pushing a diaphragm to make sound, we would push a hammer to strike the paper. This idea was implemented very, very poorly in Detroit, but the basic notion was a good one. On it we built the whole printer business, eventually. When I started Dataproducts, we were flooded by people from Ampex who wanted to join. But we chose carefully. We picked a key financial man, who really was surplus because Ampex had its own chief financial officer. So we had Bill Mozena as our financial man. We had Cliff Helms as our key technical person. Graham Tyson was the head of operations. He was also a good engineer. Russ Dubois was the head of marketing. He had done the marketing at Ampex. That was the team that signed on April 1st.
We had to quickly assess the problem with the disk drives in St. Paul. It became evident that they'd made a mess of the electronics and that the logical system was just a jumble. They hadn't really had a decent logic designer at all. So Irv Weiselman, a proven logic designer from Ampex, was hired because we needed someone to sort that out. We also felt the need to get somebody to handle GE in Phoenix. We needed to hold their hand, and keep them in the family until we could get the product fixed. What was doubly worrying was that they could cancel the contract and retain full manufacturing rights. They didn't really want to cancel. They didn't really want to manufacture. But these disk units were essential to their program. If they lost faith in us and felt we couldn't fill their needs, we would drive them into the disk business, disk drive them into the business. What we needed to do was establish and maintain their confidence. We picked another Ampex/TMI person by the name of Jack Ogg and asked him to join us. I think he was our seventh or eighth employee. He went to Phoenix to keep things calm. At the time, we didn't know if we could fix the problems because we didn't know what was wrong yet.
That was the start of Dataproducts. We ultimately succeeded in keeping GE-Phoenix as a customer. We built a few hundred big disk files for them. They used big, huge platters for disks. They were not as large as this 4-foot table, but I think they were 32 inches across. It was like the early IBM RAMAC. We ultimately built some even larger units. I saw one of these a couple of months ago in the Science Museum in London. We shipped a huge unit to Cambridge for their Atlas machine, and it is in working order at the Science Museum today. It's on display there.
Anyway, the GE contract gave us a start. In the meantime, Cliff Helms came back here to California and started the printer product line using the venture capital we had raised. It took us about three months. Once we got things fixed we had the GE contract for cash. Just as we were running out of money, we started shipping to them. In the meantime Cliff and Tyson started to build a little operations nucleus to work on the printer. It was a very small effort at first because Cliff was just figuring out what to do. He was literally the project engineer himself. He was ideally suited for that challenge.
In what way?
First of all, he's a very, very good engineer. I didn't say that lightly and I don't say it about many people. He's truly an engineer. By my definition, a person who's really an engineer can deal with almost any technologies, can teach himself what he needs to know, and has a fundamental understanding. Cliff starts with F = ma when he has to and works it out from there. He states the problem, analyzes it and solves it. He understands technology in the fundamental sense. Cliff also has insight into mechanisms and is very innovative in development. He came up with the idea that we would build a printer with no solenoids, no clutches, no brakes, and feedback control in the paper drive. As I said, he used the moving coil principle, but he mounted the coil on flex pivots, that is, spring steel legs. The current was conveyed in the coil right through the legs so that there were no coil contacts to make and brake. There was no friction wear; there were just pivots and spring legs. So the hammer has virtually endless life. There are Dataproducts printers working in the field that are thirty years old. The hammer mechanisms are still going. Spring steel that's flexed within its tolerances just keeps flexing. He is that kind of an engineer. He tends to solve things in a fundamental way rather than tinkering with them. He sat back, thought deeply, and came up with his solution. Of course the printers five years later had continuously evolved--smaller, faster, simpler, and all that. But the first idea was from his notebook of the first six months. He got rid of the clutches and brakes used in the paper feed by using printed circuit motors, which were just coming on the market. You could use electronic feedback and an optical disk to lock the paper, so you didn't have a clutch and its associated wear.
Dataproducts and the growth of the printer industry
As the printer industry matured, and as Dataproducts got larger, what were the kinds of new challenges the company faced?
At the beginning, disk drive shipments carried the company as a business. Our printer product challenge was to get the design done, get into production, and to gain credibility with customers by getting them to test them. There was an established supplier of printers, Analex, in Boston. Their printers were priced at about $30,000. They used solenoids and their printers were notorious for needing maintenance and realignment. People regarded Analex printers as successful if they printed pretty well for six or eight hours. After that you had to spend an hour or two to realign and maintain them But they were the only game in town if you weren't IBM.
We had the challenge of getting into production and getting first customers for an unproven product. We readily had great difficulty to take this new design into production. It meant learning a whole new set of technologies and for the first couple of years we were struggling to get those right. Of course there was the challenge of product costs. You can't control costs when you're first learning. When you don't know what you're doing or how to do it, you can't cut costs. You have to have stability in what you're doing and then you say, "Now I'm going to do it a better way."
Our printer depended on printed circuit motors and depended on our new hammer banks. The hammer banks were by far the more difficult problem. We had three new challenges. New very strong magnets; they were a new type of Alnico which was still not uniform, so we had to take into account individual variations in the magnets. A typical engineer might simply write a tighter spec--in effect, transfer the problem to the vendor. That's all very well, but usually the costs go way up because the vendor ends up selecting magnets. The proper approach is to work within the characteristics of the production material.
The second challenge was inherent in the hammer design. When any print hammers strike the paper, there's a force of nearly 1000 G experienced. In our design this impinges directly on the coil. How do you insulate the coil and keep it from shorting internally. The solution was to use flat wire. We made a sturdy package of flat wire so that the turns lay one on the other instead of crossing and jumbling. The shock, when it hits, tends to travel through the coil. But that meant using a flat wire. We found that we could buy flattened wire. It's made in the instrument industry for ammeters and things like that. It's about a dollar a foot. We were talking about literally hundreds of miles of wire, and we'd like to pay less than a penny. So we developed wire-rolling mills to flatten the wire and managed to do it without changing the tensile strength. Indeed, once we got going and started buying aluminum in quantity, people from ALCOA showed up to see what we were doing. They'd never seen this before.
Third was the problem of coil and contact insulation. For this solution we borrowed epoxy then starting to be used in the aerospace industry. Epoxies were then brand new. We used epoxy to coat and thus insulate the flat wire. We wound the coil and then baked it. It became a solid mass that wouldn't move internally regardless of how many G's were applied. We used a different epoxy to insulate the connections. I mentioned that the coil carries the hammer and is attached to flex pivots. The current is sent through the pivots. So, somehow, you have to insulate the coil from the rest of the structure.
The reason I describe this in detail is because it only took us another year and a half to develop these manufacturing techniques and machinery to make the printer hammers, to develop the disciplines, and so on. That was a serious difficult challenge.
Our next challenge was getting the product introduced and accepted. We used price. Our price was about $14,000 versus $30,000 for ANALEX. We had to get the attention of the market. Engineers quickly saw the merit of our design. But decisions to commit are based on more than just the design. If you're already using ANALEX, you have trained service men, you have a large inventory of spare parts, you have machines you haven't yet written off, which you're renting out. There are many factors to be considered.
After about a year and a half, we were out of money. That was another big challenge. We had not yet delivered printers in any quantity. We were shipping disk drives to GE, but that was not enough to keep us afloat. We had used all the initial money to fix the disk drive product and to finance the printer development. As is often the case, we stood on the brink of success--out of money. We decided to make a rights offering to our shareholders, as is more commonly done in Europe. We offered stock at a low price and we got our investors to exercise their right to purchase the shares. That got us enough money and stabilized the company. Then our challenge became developing a flow of competitive products.
A few years later we ran into other serious problems as I said in my remarks about the disk drive business. We finally found we couldn't afford the engineering it took to keep abreast in that business We simply didn't have the financial resources.
Venture capital and government in technological industry
Do you want to make any more general comments about capital, its availability, and the way it affects the running of a technological business?
Yes. I have some observations and experience that I should talk about it. I think the concept of venture capital funds is a very good idea and is very important for start-ups. However, I think that it has matured into something that really doesn't fulfill this function any longer. Venture capital today has become a big business of large funds. It's very difficult for these funds to manage a large number of very small investments. Therefore, they tend for this and other reasons to select only for large opportunities. They want to invest in companies that will be able to go public at the end of five years. In today's markets that means a $50 million revenue company with a chance to be a $500 million company. Therefore, they tend to invest in teams of people, in clearly identified products, and in markets that are already established, not new. Fifty million and 500 million are not start-up markets. The result is over-funding of start-ups in large established market sectors. These don't do the economy or the funds any good.
For example, we managed to start so many disk drive companies that no one has made any money. Seagate did develop, and there are a half a dozen others that may survive, but the venture capital firms had twenty or thirty of them going. There is indeed a big market for disk drives. But by over-funding they made certain that no one would be a big success. Venture funds nowadays put large pools of capital together. They invest a few million each and the deal ends up at 25 million, so it must be a big deal. We need more smaller venture capital firms that have more entrepreneurial direction and undertake higher risk deals such as true start-ups.
It's very hard for an individual with only one or two other people involved to get funding today from venture capital. The venture capital specialists demand that you have a complete management team and a clear product idea. They demand too much sophistication at the outset, and demand too much from the lone entrepreneur. Yes, he needs to have a business plan, but it ought to be a modest, not very sophisticated, plan--one that can really be put together by asking some simple questions. He needs half a million dollars, not $5 million. But then he also needs support and faith. In almost every business you go through at least one valley in the early days where it's questionable whether you're going to make it or not. Maybe two. I know that was true of Federal Express, a venture deal that almost shut down. That's when you need the courageous venture capitalist to give it the extra money to push it over the top.
Other countries use government-supported bank capital to finance new ventures. Our banks are just not in this start-up, small company thing at all. I think that's a weakness. We rely on the stock market both for money to grow companies and for the payoff. The public offering is the window where the venture capitalist presents his winning tickets. In Germany, Nixdorf, who was a customer of ours for years, was financed by a regional or state bank who wanted job creation and therefore wanted Nixdorf in their area. They backed him until he could go public when it was a substantial company. That's an example of patient money. The Japanese use their banks for equity money and have more patient money. The difficulty with our method of using the stock market and venture capital oriented to the stock market is that it really does focus management on short-term results. All the rewards and incentives are short term. Some people use the phrase "investing," while other people use the phrase "speculating" to describe the same actions. In any case, you want the shares to go up. It puts an undue pressure on the management of young companies to perform quickly and to strive for quarter to quarter improvement. The pressure is to avoid the best kind of investment, which is the investment in people and know-how and infrastructure.
For example, at Dataproducts, when we were smaller, I could buy a million-dollar machine (if we had a million dollars) and it's cost did not greatly impact profits. If it does not become obsolete, you are able to write it off over eight or ten years. The impact on earnings is $100,00 or $125,000. If I decide to spend a million dollars to develop a new product line, engineering and development salaries amount to a million dollars spent and are "lost" immediately. But which would you rather have? A piece of machinery or a new product?
If, for instance, you decide that the thing to do at Dataproducts is to double our marketing capability and open new offices because that's going to give us the channels and the penetration we want and that costs a million dollars, that's a loss. If I bought a piece of machinery, it's an asset. Which would you rather have? A doubled marketing capability or a piece of machinery? So the public market measurement scheme--there's nothing wrong with the accounting per se; you have spent the million dollars, you don't have it, it's gone, you shouldn't put it on your books--but according to the measurement scheme that comes with the public market, earnings just went down by a million dollars. Therefore, instead of adding $10-$20 million to the shareholder wealth, we have hurt valuation of the company. We've hurt all the shareholders by doing what's right for them. That gets me back where I started--with the long-term/short-term. So what all the shareholders would like is increasing profits and investment for growth. That's a very, very difficult thing to do. This tells me we have more companies going public than are able to meet these expectations. There are fewer companies being financed than deserve it. Many companies don't need to be public, and they aren't in Japan or Germany. But they are worthy job-creators, worthy of fulfilling the entrepreneurial urge of the founders and good for society in general. They pay taxes. They do a lot of things we as a society want them to do. They become suppliers to big companies. They make money for the owners. They fulfill an important role. Some of them will merge and become public later. All in all, I think the short-term syndrome is too tough on a small new enterprise.
I want to ask a similar, general kind of question about government and its role in these kinds of businesses.
I am of a mixed mind there. My experience with government in one aspect is very positive. In my judgment, the seeds for the computer industry were planted by the government's risk-taking. The government bought computers when no one else did. Even though some of the early customers were aerospace companies, they were all on government contracts. The government bought all eighteen of the IBM 701s, most of the 1103s, and all but one of the early UNIVACs. Because of their needs and demands they also encouraged a lot of development work which later showed up in commercial products. The commercial pull came later. So the role of the government as a major risk-taking customer, one who has large-scale problems and can afford expensive new technology, is a very valuable function. In general, I'd rather see the government be a leading-edge customer than see it try to set the direction for industrial or technological development. I think the economy gets helped more fundamentally when the government asks for a network to use and they're willing to help finance it, than when they sponsor network software development and logic. It's the solving of real live problems and creating demand that has pushed our industry along. There I think government has a positive role, and I wish that somehow that role could become more socially acceptable and palatable.
The government in its regulatory role has a very, very heavy hand. This is particularly true for small new enterprises, and it's on all levels of government. You can't build a building in the County of Los Angeles in less than two to three years. California has a Workman's Compensation system and state laws regarding employment that are outrageous. Federal OHSA, EPA regulations are no better.
We pride ourselves on all the jobs we've created in the services sector. In my opinion, a significant fraction of these, those that are not minimum wage, are non-productive and have been built through government regulation. The tax law is an accountants' and attorneys' retirement act. The public pays for people who are going to protect them and help them, and facilitate for them against another set of employees called civil servants that they are also paying for. With racial, sexual, age and gender discrimination laws, it is a wonder small industry is willing to risk hiring any one. No wonder the "temporary" employee business is booming. The challenge is to maintain a balance. We want a fair, equitable society, but no one feels we will get there by government regulation.
I don't know if I would even consider starting a company today. I was an acting CEO for a little while recently. It came to me as a shock to learn that as a practical matter you can't just fire somebody for incompetence. You must first build a case, and that takes at least six months to get the file right. What does this do? It prevents you economically from replacing that person for at least six months to a year. You continue to suffer the costs of that incompetence. You lose the time. Then finally you do replace him. I suppose we have prevented abuse, but at what cost? For whatever percentage of abuse we had, is it as large as three percent or five percent? We've spread this cost on everybody to prevent this abuse. The government's hand, that indirect hand, in regulation and in an over-legislated society is very costly. It introduces distortions and aberrations make it very, very difficult for anybody to start a company or to carry on a business.
What about an international trade role of the government?
I think the government is right to try for a level playing field. However, the mechanisms that are available for protection are a joke for small business because they are so expensive in both time and money. If you really feel that you're being discriminated against and you'd like to bring a modest, straight forward case to the Federal Trade Commission, it can cost a hundred thousand dollars a month in attorneys' fees. The process we have requires you to prove damages. But to prove damage takes a few years. By then you're out of business.
If I may generalize for a moment, in European countries where civil servants are not looked upon as potential enemies, they in turn don't have an adversarial attitude towards industry. In those countries, civil servants are expected and encouraged to exercise judgment. By contrast, our system does not reward risk-taking and encourages the avoidance or use of judgment. It tries to substitute rules for judgment. If I could show that this glass here is being sold in Japan for $50 and is being sold here for 5 cents, and I could bring you the sales slips to show you that and a picture of both items, it could still take years and millions of dollars to get a court order dealing with that. No one in government will make a judgment until they get a court ruling. No one will decide anything major short of a court decision. So you see by emphasis on process and regulations we simply handicap ourselves and slow our whole response mechanism. I think that American companies can compete well abroad on a level playing field. Even without a level playing field they do and should be encouraged to do it. Not that they need much encouragement. The products that sell are the items that are attractive because they represent leadership. We sell jet airplanes because other countries don't make them as well. We sell ultra large, ultra fast computers, and so on.
In the printer business Dataproducts no doubt had Japanese competition. As I understand it, Dataproducts now has a Japanese parent company. Do you want to talk about doing business in competition with the Japanese?
Yes. Dataproducts was able to compete in the OEM market very successfully against the Japanese. They tried to compete and later told us they couldn't make any money selling printers against us. We sold worldwide on our own. We offered quality products at good value. The market was relatively small, and it was only a niche market. Dataproducts had perhaps forty percent of the business, and it was a $500 million company. Relatively speaking, it was a small business in terms of worldwide markets. Where the Japanese excel is in mass production of standard products, and there we couldn't do as well. We didn't think mass distribution or mass production. So Dataproducts missed the personal computer wave, and Hewlett Packard caught that wave. Hewlett Packard showed how to join with and compete with the Japanese. Their two best selling products are based on technology. The case is made in America, and the electronic board is made somewhere off-shore--not Japan. But the Canon engine inside is what makes that printer, and that comes from Japan.
The Japanese have excelled at mass production, in volume, particularly on smaller units sold to the end-user. We didn't do well there. Whether we could have done well, I don't know. I do think Dataproducts as a company could have done better in this last decade than it did, and I fault the management. Dataproducts didn't try to lead in non-impact printing. It stayed too long with the hammers and its established technique. It let the next wave catch them. Today, we may think that laser printers are the cat's meow, but I assure you--although I don't know what it is--something will come along to obsolete lasers. The new technology will be cheaper, better, perhaps provide color, whatever, but inevitably something else will replace lasers.
There are several disadvantages intellectually, from a technical point of view, to the laser printer. It's not a direct printing approach; it's indirect. Then there is the need for replacement of the cartridge. The process has a great deal of friction inside the cartridge. One reason they went to a replaceable cartridge is that toner is abrasive. You need new parts to account for the wear. Intellectually, there's room for someone to come along and say, "Hey, let's get rid of the abrasion, let's get rid of the indirect process. Let's make the marks with ink right on paper instead."
But certainly Dataproducts was too slow to respond. I think we should have done it in partnership with the Japanese. If we'd wanted to go the laser route, we didn't have the xerographic technology. In America there was no mass producer of small laser engines. There were great producers of big engines, such as Xerox and Kodak. But not small engines. The small engine was needed for that printer. But even before lasers, in the time of dot matrix printers, we really never did compete with the Japanese in manufacturing. Dataproducts had some high quality, heavy-duty printers, and you see them even today in ticket-printing applications. American Airlines still has several thousand of those, at travel agents desks around the country. Where you need a heavy-duty printer, those little light printers from Japan wouldn't stand up. But ours were priced in the $2,000 range as compared to their $200 printers. An order of magnitude different. I'm not sure that there's something fundamental at work here. I think America can mass produce items in competition with the Japanese. I don't know that we can mass produce them in competition with the Philippines, though. It seems to me that one lives on an ever-shifting ground. The advantage the Japanese had they no longer have. Korea has it now, but won't have it very much longer. Singapore won't have it too much longer either. Perhaps Singapore is a little different. They've got an endless supply of cheap labor across the border. [Laughter] As long as you have large sectors of the world that are relatively underdeveloped, there's always going to be a place with cheap labor.
I think we can compete with the Japanese today, and indeed isn't that happening? The Japanese are opening assembly plants here. They're now competing with plants in Japan. I don't have much trouble with the nationality of whoever owns the factory building if the money is being circulated here and the value is being added here. We get the jobs and the development of subcontractors and the improvement of the whole structure that occurs.
I think that there is a larger issue. The United States really faces some very difficult problems, as do the rest of the developed nations. We have enjoyed in this last century a much higher standard of living than the rest of the world. We feel we merited that because of the industrial revolution and its products and byproducts. We have furnished goods and services that were desired by the rest of the world, for which they were willing to pay a premium. They have furnished raw material, for which we did not have to pay a premium. I think two things are going to happen: as the population increases and the competition for materials increases, their value will go up; and second, the costs of industrialization are equalizing.
The true cost of industrialization is just emerging as we begin to understand the environmental impact. We really ought to put that cost in the price of the product, but we don't know how to do that. It is not a lack of will so much as ignorance. When we sell a gallon of gasoline, we know that it's going to create pollution and that we're going to pay later to try to clean it up. We really ought to put those costs in the price.
What I think is happening in these last ten years, and at an accelerating rate, is that we are seeing a lowering of the standard of living in the developed countries due to these forces. It's not a temporary phenomenon due to an incorrect political policy or cyclical recession, but a real basic societal change. There is no reason why we should be permitted to maintain a superior standard of living if we can't produce desired goods and services for people. If they can do these things for themselves and are willing to live at a lower standard of living, our standard will inevitably have to come down. We won't deserve or get that premium. Don't despair, yet. I feel any individual will be able to stave this off for a while. It won't be really apparent for decades. Perhaps a century is a better time span to see what's happening. Some historian will be sitting here in the year 2200 and talking about what happened in the major shift of the year 2000.
An Interview with Erwin Tomash conducted by Arthur L. Norberg, 15 May 1983, Charles Babbage Institute.
Erwin Tomash and Michael R. Williams. The Erwin Tomash Library on the History of Computing: An Annotated and Illlustrated Catalog (2008).