My first million: 'We thought if we delivered on time we could charge what we liked'

Peter Stephens used to be an academic, but in the past 16 years he has parlayed his experience into a computer company, Edinburgh Portable Compilers, which is today worth more than pounds 4m.

Our business began with a colossal piece of foolishness. We would never have started if we hadn't been pushed. My partner, Geoff Millard, and I were writing commercial software that brought in more than pounds 2m to the University of Edinburgh over the years. But in 1982 an edict came down saying we should stop. It was a bolt out of the blue.

To get around it we set up EPC. Two years later the universities were subject to severe grant cuts and Edinburgh set up a commercial wing to do exactly what it had told us not to do. In some ways the move out of academia was unfortunate. We would have been further ahead if we hadn't had to spend so much time on management.

At first EPC worked from the university, paying a token rent. We also paid royalties on software we took with us. After a while, when they saw the bill for our modem connections to places like Texas and Florida - which ran to pounds 1,000 a month - they insisted we get our own phone lines. In 1986 we moved to our present premises.

Geoff and I both came up from England in 1966 when the Edinburgh Regional Computing Centre was started in response to former prime minister Harold Wilson's "White Heat of Technology" election campaign. The system we helped set up grew into Emas, which in the 1960s and 1970s had only one rival, Mulics, at the Massachusetts Institute of Technology. It's still better than systems such as Unix and MS DOS available today. It was one of the first mainframes which allowed lots of people at different terminals to share time on its processor.

One of the things we had to do for Emas was write a compiler - the software which translates a program into a form that particular computers can run properly. They cost about pounds 1m to write, which isn't much in this industry, and act like piston rings in an engine. They're not expensive but without them, or with a faulty one, you've got a serious problem.

When EPC got started it had pounds 1,000 of equity and pounds 15,000 in loans from the founders. Initially, we had a third partner, Linda Griffiths, who was a programmer and project manager. Shortly afterwards she became ill with cancer. It was overwhelmingly sad from a personal point of view, and terrible for the company. The bad news came out slowly. At first there was every hope there would be a cure, but two months' sick leave stretched into six, and nine months after we launched she died. We managed to get through it by covering as much as we could between us. One consequence is that our bank, with which we have a pounds 150,000 overdraft facility, has insisted ever since that we take out large insurance policies on the executive directors.

The tragedy of computing has been customers' lack of discernment. Second- and third-rate solutions have won out because of marketing. Unix as a technical choice would be nowhere, and MS DOS, which has made Microsoft rich, would be even further behind. The same goes for chips - Motorola's 88110 chip is far better than Intel's, but in fourth place in terms of market share. We've succeeded in carving a niche for ourselves - although most of our end users don't have the required skills to evaluate our product technically - through word of mouth and because we've saved several big deals.

For example, Sequent, a mid-sized US manufacturer, sold equipment worth millions of dollars to a large US wholesaler that specialises in just- in-time deliveries to drug stores. However, the compiler it had wouldn't run the customers' old programmes, and the drug wholesaler was on the verge of cancelling the deal when we stepped in to solve the problem.

Even in those early days computer programmes had a reputation for being late and over-budget. We thought that if we could deliver on time we would be able to charge what we liked. The lights here often burn on Saturday and Sunday as people work late to hit a deadline. For a long time we succeeded, but when a new programming language, Fortran 90, came along, things began to fall behind.

It was supposed to be called Fortran 88, but even after the name change it didn't become a fixed target until 1993. Standards for new languages are drawn up by international committees. It's like trying to write the rules for the English language - whether you should put "a" or "an" before a word beginning with a vowel. We tried to get ahead of the game but some of the things we thought they would include were voted down and vice-versa. Part of the problem was that the big users wanted the new Fortran to run all the programs written for Fortran 77, the old standard. It's a schizophrenic way to design a computer language.

Our early business plan was to make enough money selling Fortran 77 and other compilers to pay for the development of the first Fortran 90 version. We eventually succeeded in being first, but not by as big a margin as we would have liked. We ran out of the self-generated pile of cash in early 1994 and we knew it would take two years to boost sales. We approached five venture capital companies to help us bridge the gap, and got offers from three. That enabled us to negotiate with them, and eventually to get 3i to reduce the portion of its minority investment that was in the form of repayable preference shares. They put in pounds 450,000, of which we have used pounds 300,000.

The venture capital firms didn't know what to make of us. We haven't got the enormous drive of some entrepreneurs. They kept asking about our management experience, or lack of it, and were keen that we employ someone who had been with a large company.

One firm we had worked with from the earliest days was Gould, a US conglomerate that started 50 years ago making batteries for Detroit's motor industry. From there we recruited Derek Leadbetter, a Scot who had lots of computer experience, to be our managing director. He has enabled Geoff to spend more time doing sales work in the US and Europe, and allowed me to devise technical solutions to problems.