Chapter 7 - Data Communications: Market Order 1973-1979
7.10 Micom: The Statistical Multiplexer 1976-1978
After the last minute financing in September 1976, Evans knew he needed to find a viable product opportunity if Micom was to escape its uninspiring past. To find a compelling product idea he began making calls with his manufacturing representatives. In one such call, he learned that Gibraltar Savings and Loan in Los Angeles had a Data General minicomputer and wanted to support offices in San Francisco and Fresno by generating mortgage and trust deeds on-line. The manufacturing rep had sold the account on using Micom’s TDMs, claiming they would support three terminals and a printer in each office when used with 9600 bps modems. Evans would inform the customer that using TDMs would not work. He tells what then happened:
So I ended up walking away, first of all having made a friend for life because the customer had actually for the first time talked to a salesman who had sold him off something, rather than onto something. Secondly, I walked away saying: ‘If we could solve this, we’d have a real business here.’ Third, I had at least learned enough from this guy to know that there were quite a number of other guys in the area planning to do similar things. So I was shocked in a way that I perhaps wouldn’t have been if I had grown up with it – the fact that these minicomputers were being used for business applications all over the place. They were growing like wildfire, and people were actually – in some cases had actually put multiplexers in to do some of this stuff, and there were several things that were obvious. One, they desperately needed retransmission on error, just to make the thing usable at all. Secondly, they were paying an arm and a leg for 9600 bit per second modems to be cut up into four 2400 bit channels where the terminals were hunt and peck in one direction and only sporadically used in the other direction. So the data traffic was minimal, and it was the classic case where a statistical multiplexer comes into its own.
In a flash of insight, Evans envisioned selling statistical multiplexers to minicomputer users. His instincts were so strong and certain, he assumed others had seen the opportunity as well, and if they hadn’t, it wouldn’t be long before they did. A sheer volume of what he didn’t know, what he had to know, and how much there was to do before he could be clear as to what Micom should be doing and when, preoccupied his thoughts and emotions. As for details, such as having to find a customer who would fund the development and not retain product rights (for Micom had no cash to develop and launch a statistical multiplexer) well, they were just details, and would be sorted out in due time. To be so certain yet faced with so many uncertainties might have stopped a less street-wise, self-confident entrepreneur, but Evans knew he had a winner and he was not going to let it get away.
Evans first sought to confirm that the Gibraltar situation was not unique, that the growth in use of minicomputers to support remote locations was substantial and unaddressed by other multiplexer vendors. To his amazement, the potential appeared larger than he would have ever imagined. He also learned system integrators sold and installed most minicomputers and they required a margin of 40% off list price – a discount structure he assumed Micom would have to offer as well. Furthermore, the telephone circuits between terminals and computer tended to be much shorter with minicomputers than mainframe computers. Shorter telephone circuits meant less expensive lines and reduced savings to justify purchase of a statistical multiplexer. Evans recalls:
As I talked to potential customers, I ended up concluding that the real market probably lay with a list price around a couple of thousand dollars for a four channel box. We also realized that, whereas Codex and Milgo expected to install their equipment with an engineer that understood communications, we were going to be reliant on systems integrators or the end customer himself, neither of whom knew anything about communications. So ease of use and built in test facilities were absolutely key, and do it yourself installation was much more important than anything else.
Next came the challenge of finding a customer to fund development. In a stroke of good fortune, an existing customer, Reynolds & Reynolds (RR) of Dayton, Ohio, sought a device that would connect terminals in their customer locations, automobile dealerships, to their minicomputers. They had a potential supplier of a “very dumb” product that might work, and queried Micom to see whether they could offer a better product at the same low price. If Norred and Evans had any questions as to the viability of the market they were contemplating, they were dispelled when they asked RR how many boxes they were interested in buying. Norred remembers:
They said: ‘Well, certainly we’re going to need at least 1,000.’ To us, a 1,000 of anything was just mind boggling. You just didn’t sell 1,000 of anything in our business.
Norred began evaluating design options not knowing how to solve a problem known as “flow control.” When too many terminals wanted to send data at the same time, a solution was needed to store the “overflow” data, presumably expensive memory chips. To Norred’s surprise, minicomputers had a “stop terminal” command that stopped terminals from sending more data. It was then that Norred began to believe a low cost statistical multiplexer, or “concentrator” as Micom would name it, was possible.68 Norred remembers:
It was that breakthrough, in my mind, that was the biggest, most significant element of why we could build a concentrator and make it a viable product.
Norred’s first design for RR proved way too expensive. In reviewing how much cost had to be eliminated to meet the objectives of both RR and Evan’s preliminary concentrator specification, both men doubted it could be done. Fortunately, Norred had designed in a microprocessor, the Zilog Z-80, which allowed them to substitute software for logic chips, and as Evans summarizes:
We threw out a chip here and a chip there, and then ended up concluding we still hadn’t thrown away too many chips to be able to implement statmux code and make it work.
By late summer of 1977, Norred and Evans knew that their concentrator would be ready to ship in early 1978. To their surprise, no competitor had announced a similar product, or one targeted to the rapidly growing minicomputer market. Evans remembers his wishful thinking:
If, when we’re ready to launch the statmux it looks like we really have the field clear, we’ll stop everything, focus on the statmux, and become a one product company for as long as it takes for us to dominate and have resources to spare, because the rest of the world’s being stupid enough not to see the opportunity now, but they sure won’t be once we bring the product to market.
In January 1978, Micom introduced the revolutionarily modest Micro800 Data Concentrator, Models 804/814 and 808/818, priced from $1,650 to $2,750. They were instant hits.69