An interview with John Walker - Part 3

This is the third part of the series documenting my recent interview with John Walker. Where the first part looked at the early history of the company and the second looked at the architectural evolution of AutoCAD and its APIs, this part focuses on Autodesk's business strategy and focus on its eco-system. I have not linked to terms previously linked to in the first part of the series, which should also be referred to for typographical conventions, if they're not clear.

Autodesk's eco-system and strategy

We've talked about AutoCAD's early architecture. How did this relate to Autodesk's business strategy?

From the outset the architecture of AutoCAD was intended to be as open as possible. We intended to leverage every partner that wanted to work with us, in any capacity whatsoever, to use their resources. This was something that we had a lot of trouble explaining to the financial community and to the traditional CAD analysts. Because they would say, "how can you possibly compete with a company like Intergraph? They have a training staff that's larger than your whole company. They have a field support staff that's larger than your whole company. etc. etc." And I said, "because I believe in the market. Because I believe that if you create the market, our dealers will become our field support staff, universities will become our training department." And, indeed, all of those things not only happened, they had already happened by the time I was trying to explain this to them, those things were already underway. You will have every community college teaching this product, and they're not going to teach the Intergraph product because they don't have $150,000 for an Intergraph workstation to teach it. But they've already got the PC: all they need to do is to put this disk in it. And in terms of documentation and training, we've got authors already writing books about this product. For themselves - we're not paying them. I believe in the market solving things. A lot of people laugh at that, but hey, it worked for us.

You mention the importance of the market and having a healthy ecosystem. How important has it been to Autodesk's success?

Absolutely central. I don't think the company or the product would have been viable, if there weren't that community of dealers and developers, of educators and writers of books. One huge contributor to AutoCAD's success was the book Inside AutoCAD. I think I have a first edition, somewhere. Because that was what you would now call the missing manual. The AutoCAD manual told how it worked, this told how to do things with it. And that book created a whole publishing company, basically.

So we really had... the number of people that were contributing to the success of AutoCAD and Autodesk was much greater than the number of Autodesk employees. And, even better, we didn't have to pay them. That was really the strategy. As I said, we were explaining that to venture capitalists in '83, when they were visiting us. In exactly the form I've presented it here, except it was more in the future then than it is now. They simply didn't get it. These are guys who, their job title includes capitalist, but they didn't understand capitalism! I mean, the whole idea is that you create a market, and people will pop up spontaneously to fill the needs of the market, because they want to get rich doing so. And that's exactly what happened.

When we got back from COMDEX, as I said, we were deluged by dealers calling... "how can I sell this thing?" Why were they interested in selling it? Because it cost $1,000! That's why they were interested in selling it. A heck of a lot better than selling a WordStar for $300. And what's more, they sold a big graphics monitor, and a card and a digitizer and a plotter - that's a $10,000 sale. As opposed to $2,500. That's why they were interested in it, not because of the technological capability.

And what did we tell them? "You want to be a dealer? You get 40% discount. You have to qualify to be a dealer. To qualify to be a dealer, buy more than one." That was our policy. We had a couple of hundred dealers, before too long, as what's more we had a lot of money coming in, because they had to buy two copies at $600 a piece. So that was $1,200 for every dealer we signed up.

You were also surprised by the quality of some of the dealers, who wouldn't have been eligible if you'd put in place rigorous certification from the beginning.

That's right. We had dealers who would have never been considered for any kind of certification. Some of our most successful dealers were practicing architects who bought one, and suddenly people would visit their shop and see it, and "well, I can set one of these up for you." And before long, their architecture business was gone and they had this thriving business selling and servicing AutoCAD. I think we had a community college professor who found himself in the same situation. He started teaching it, and suddenly his students were out needing to buy systems and they were starting to work for people. And he developed a business selling these systems.

One other contributor to Autodesk that's frequently forgotten, is that, among our founders, we did have four Europeans. So we were operating in Sweden, Switzerland and the UK from day one. And in fact some of our very first sales were there. And through much of Autodesk's history, through the '80s, about 40% of our business was in Europe. And 40% is your profit margin - if we didn't have that business, we would have been a marginally-profitable company.

A lot of US companies took a long time to get there. And we really knew from the start - because my little, piddling, ridiculous hardware company got about 40% of its business from Europe, and that was about my profit margin. So we knew that we needed the localized versions, that we needed translated manuals, that we needed to have a presence on the ground in order to address that market.

We were much slower out of the block with Asia. Just getting the first localized Japanese version was a real challenge. That was, of course, the "go, go, Japan #1" era, lifetime employment, you couldn't hire a programmer in Japan, you couldn't get them. An American company that's only three years old? Uh-uh. We ended up with a whole staff in California of Japanese Americans. So we'd get these machines from Japan. And of course these Japanese manufacturers were just tripping over themselves to get AutoCAD on their machines. But, of course, they came in and the machine speaks Japanese and all the manuals are in Japanese. And at that point the DOS for Japan - this was all pre-Windows - was a pretty weird creature. There was a lot of stuff that was different with it, with Kanji keyboards and data-entry and so forth. So that was a lot of work getting there. But once we finally got the work done it was a great success.

Although the Japanese market was - there really wasn't the Japanese equivalent of the IBM PC, except for the IBM PC - every big Japanese company had its own PC, which was nothing like anybody else's, and if you wanted to sell to Hitachi, you had to put it on Hitachi's machine, and same for Sony, same for Toshiba, etc. And, again, so you had this platform explosion, each to address essentially one customer.

But we had, really... we spent probably the better part of three or four years flapping around in Japan, listening to these analysts who told you, "the Japanese economy is entirely top-down, and you have to get into the big, integrated companies, and sell to the top, and they make a company-wide decision when they choose the product. And that's what you have to do." So we pursued them, we were doing all these ports, we were doing everything, and then we hired a new manager for Japan who was a Japanese national but he'd lived about half of his life in the US, and he basically said "you know that's all bull----, your market in Japan is going to be small business, just like it is in the US. There is a huge sector there. Because the huge companies, they already have a CAD lab with workstations, and things like that in it. They can afford the very highest-end CAD systems. You may get some AutoCADs out in their field offices, and so forth, and for things that you're not doing on the workstation, and we can get there eventually, but we need dealers, we need local support people, we need exactly the model you've got in the US." And so we said "go ahead." And up it went.

AutoCAD used to have a hardware lock outside the US...

In the early '90s, we had no locks in the US, and locks everywhere outside the US, and that created quite a bit of tension. I would always ask the regional managers, for Europe and Asia, "can you estimate for me the difference in revenue if we did not have the hardware lock on AutoCAD in your market. Universally the answer would be 'none'. Because there were so many cracks going around that anyone that wanted it could get it to run without the hardware lock. And there was simply no person that was willing to use it illegally who was unable to, that was the fundamental reality. And I said "well, why don't we get rid of it? Because it'll eliminate this tension that we have here." Particularly with copies flowing out of the US, with unlocked copies to Europe, which the European dealers were just enraged about. They wanted us to lock it in the US, and consider the Indian market, which speaks English. I mean, good grief. We had zero penetration there, everybody was running illegal copies from the US.

It would also, at that point, knock $30-40 off our manufacturing costs for every copy with that thing that we were buying, and so our profits would go up. And they said that you could never do it. And I said "why?" And they said "the dealers - the dealers would walk away from us in disgust." The dealers would perceive it as we're abandoning them to the pirates. Even though actually it would probably have had no impact on their sales.

And this wasn't just idle speculation: because at this point, most of the other major applications were not locked, and they were selling fine in Europe. So it wasn't like Microsoft wasn't selling any Microsoft Office in Europe.

The hardware lock was introduced briefly in the US at one point, I believe.

In 2.5 it was introduced around Easter, and pulled around Thanksgiving.

And it was then nearly put back in?

About three years later they were going to try to put it back in, and that's when I went to war. And the amazing thing, the astonishing thing about it is that, almost without exception, the people who were going to put it back in were the same people who had lived through having it in and taking it out the last time! It really is amazing how the human mind can flush unpleasant experiences, so you're set up to have another one.

Now it's certainly true: I designed the original hardware lock - I'm guilty as charged - and I can show you it, the prototype downstairs [John did, indeed, show me the first, prototype dongle for AutoCAD at the end of the interview]. And it wasn't very good, actually, it had a lot of problems. It created compatibility problems with a number of plotters, for example. And it stumbled into a bug in the serial port chip that a lot of PCs had. It was a fundamental bug in that chip, and the hardware lock found it. So it wasn't just that people didn't like the hardware lock, it's that they put the hardware lock in and - you've got a PC, you've got one serial port, you put your hardware lock on and you put your plotter on the other end - well, the plotter would plot for an hour and then go "BRRREEEEEEE" and start drawing to the moon, or something. Quite understandably people were irritated by that. And they perceived it was the hardware lock that caused the problem. Actually it wasn't the hardware lock that caused the problem, it was this serial chip but the adaptation of our serial port driver code to accommodate the hardware lock stumbled into this bug in the National Semiconductor chip. So true - if we'd not had the hardware lock,we wouldn't have changed the code, we would have continued to skate past the hole in the ice. So you can say the hardware lock caused it but in that case it wasn't actually the hardware lock. But it was just a marketing disaster to us.

And again, part of my reason in saying it wouldn't affect our sales is when we put it on in the US our sales didn't go up, and when we took it off in the US our sales didn't go down. So it really didn't have any impact.

What tenets most contributed to Autodesk's early success? From your perspective - with the understanding you haven't had much contact with the company in recent years - do you believe them to continue today?

When I attribute so much of the initial success of the company to luck, it's hard to say that's something that has continued in the long-term. In a way it has, because Autodesk - like every other established PC software company - has benefited from the fact that Moore's Law has continued, all this time. And so you've never really reached a ceiling in the platform, in terms of the size of the applications you can develop or the power that you need to run them. And that's a piece of luck that has come along, and it may not continue forever.

When I say we were lucky, I mean that if we hadn't been, at that time - at that unique time in the market - with the unique product that we stumbled upon as much as anything... it started out as an application that ran on my Marinchip machine. If I hadn't had the Marinchip machine, the application would have never been written or it would have been written for something else. I'd have never known of it. Never thought of it, basically. Do you call that luck? Yeah, I call that luck.

And from that point, I think it was really just a question of doing my favorite three words, "whatever it takes", and being driven by the market. Listening to what the dealers were saying, listening to what the users were saying, listening to the feedback we got from the big CAD analysts, when they finally started to notice us. I think there were a lot of times when it was tempting to go off on some kind of technological crusade, to try to do something really ambitious and neglect a whole bunch of really boring stuff that just needed to be slogged out, that people were asking for.

I remember, this was probably '87-'88, in that epoch, we started a thing where we would invite the CAD managers of big companies to come and have a day at Autodesk and see all of our developing technologies. And they would fly down at their own expense, and they thought they were really getting something. What we were doing was picking their brains, obviously, for this whole time, because - again - free market research with some very eminent and distinguished people. And I remember sitting at a round table, and having the CAD manager of Boeing - now, to me it was an honor to be in the room with the CAD manager of Boeing, right, not to mention three or four other guys like that - and he said, "well, you've asked me what do I think you should do. And I'll answer, I'll tell you what not to do. Don't try to build CATIA: because we've already got CATIA, and it doesn't do enough for us, and it takes us three years to train an operator to do what it does now, and we've got people developing on it and you can spend your life there and you'll never catch up, because they're moving faster than you move. What you have to do is to do what suits your users, in your sector. And your users are not CATIA users. They're completely different communities." And I really took that to heart, because we were thinking, "big CAD, Numerical Control, Finite Element Analysis integration".

And if I hadn't heard that from his mouth, we might have done some really stupid things. That, again, wouldn't have wrecked AutoCAD, but they were just missed opportunities: with opportunity cost you're spending your resources in places that you're never really going to be successful.

Are you at all surprised about AutoCAD's continued success?

Yes and no. I think by the late '80s, I had really just one worry, two worries, maybe: Autodesk is going to do something really stupid and screw it up - either a disastrous update release or something that destroys the developer community or wrecks the dealer channel or something like that - and secondly, Microsoft. I thought Microsoft was going to come after us. Just look at their history: every major application sector, they've launched a product against it. And they did eventually: Visio. And they blew it. I actually - in March of '93 - I met Gates and I asked him, "why didn't you come after us?" And he said, "at the point that we really could have, I didn't think it would ever get to be a half-billion dollar business. I never believed that it could be as big as Office." And, actually, at that point AutoCAD was not as big as Office, but it was as big as Excel or Word, taken independently. And that was big enough to interest him. But he just never thought that it would be there. He thought - I guess I think he probably thought that...

Well, AutoCAD, to my knowledge, did something in the software industry that has never, ever happened with any other product. It sold more than a million copies at more than $1,000 a copy. And that never happened. You've got expensive software but a tiny market, cheap-large market. That's what it really took to hit that revenue figure, and I think that Gates probably having seen the prices of his own products erode, over time, as the market became larger, assumed that by the time we got to the millionth unit, we'd be down in the $200-500 price range. And that not only didn't happen, but the price went up, over time. And that created that unique profile. And that profile, I think, cost Autodesk a great deal in the '80s and '90s, because it created an impossible standard. Whenever Autodesk was looking at doing something new, it had to be another AutoCAD. Another Microsoft Word wasn't good enough, another DBase II wasn't good enough. And there aren't any - nobody's ever found another AutoCAD.

It was an absolutely unique circumstance, at the time. And I think - as much as anything - created by the fact that it took such an expensive hardware configuration to run it on, that people didn't look so much to the cost of the software. They weren't just looking for a box to run on their existing PC. And that did create problems, but really, those two things not having happened, it doesn't surprise me because the fact is, when you get to an 80-90% market share, other than messing up or having a competitor with unlimited funds come after you, you tend to stay there. A Wall Street analyst whose name I forget, who was one of the first... he basically made his reputation on having predicted that Lotus was going to be a huge success before anybody else. And he said - and this was like in the '80s, when people were talking about the big CAD companies coming after us, and they were all readying their PC packages - and he said, "they're all going to fail, you're going to succeed, and you're going to succeed because they all under-estimate the power of the franchise and of an installed base. That you have. They don't have community colleges teaching their products, they don't have shops with ten stations that are buying three or four more. They'll do fine cannibalizing their high-end systems, selling low-end systems to their existing customers, but they're not going to hit you." And I've always kept that in mind, and I think that when you hear about... the risks that you run are the ones that are a completely different thing. That the world has changed and you haven't. It's Windows is here and you're five years from being on Windows. It's Google launches Google CAD, and you have nothing to respond to it. Those are the things that are the really big risks, now.

[Tune in next week for Part 4 - Past and future opportunities.]