Marc Andreessen, Act II

People don't change so fast. That's human nature. Everyone is in favor of change in the abstract, but when it comes to things that will specifically alter people's lives, they freak out. What's more, people don't change their reading or shopping behavior all that fast. I use instant messaging and Amazon -- which I didn't three years ago. I have a global-positioning system in my car and a Replay service on my TV. But that's about it. We're not going to take in three new Web sites a month. This is particularly striking when you look through Fast Company or Red Herring or Fortune, and you see advertisements from hundreds of companies that you've never heard of. Your response is, "There's not enough room in the market."

You just mentioned advertising. What have we learned about brand building on the Internet?

The ongoing myth is that brands get built by advertising. Actually, the evidence is exactly the opposite. Brands get reinforced by advertising, but they get built by grassroots adoption and word of mouth. That was true of Amazon, AOL, Yahoo!, and eBay. It was true of Netscape. In the past 10 years, all the new Internet brands got built through grassroots adoption before the companies actually had enough money to mount a nationwide advertising campaign.

So now we know what not to do. What advice would you give to a big company that still wants to "get" the Internet? Are there ways that the big guys can become nimbler as they compete with startups?

You need to set up a separate group -- and one that's sized for the market as it is today. If you put 3,500 people on the project and expect them to generate enough revenue to pay their way, you're going to be disappointed. If you put 50 on it, you're probably better off. A lot of the actual work is software engineering -- and that tends to work better with smaller numbers of people.

There's another paradox at work here. In the high-tech industry, at least, it turns out that dictatorial management works pretty well. It's been conventional wisdom for 20 years that you're supposed to involve all of your employees, build consensus, and have ideas coming from the bottom up. But the only way for a big company to change fast and capitalize on its strengths is to have an order from the top.

What you also discover is that the market for ideas is pretty efficient. Most of the good ideas are already out there. For example, if you said, "I think everyone's going to buy a lot of optical-network equipment in the next three years, so I'm going to start an optical-network company,'' that wouldn't get you very far. This is in fact probably the exact wrong time to do an optical startup, because there are so many of them already. Instead, you need some fundamental piece of data that other people just don't have, or some level of conviction that other people lack. It's very hard to fight conventional wisdom. It's one reason why I avoid industry conferences: They are a breeding ground for conventional wisdom. They will almost defile your thinking.

And yet no one wants to develop products in isolation. What's the right way to get to market quickly with a winner?

You can't shoot for perfection in your product. You have to get the product to market, and you have to target the right customer at the right time. If you do that, the product will eventually achieve perfection over the next 20 or 30 years in the way that Windows now has, for its purpose, or in the way that Oracle now has. But the luxury of having all the right features comes only if you're so entrenched in the market that you can afford the R&D to do that.

This isn't a matter of sacrificing product quality. It's a matter of sacrificing scope. I mean, there's no question that you have to get to market really fast. You have to get the product in front of a customer. You have to get real-world feedback, because every day that you're not hearing from an actual paying customer is a day that the market is moving further away from you. When you start these companies, you have only a vague clue about what the actual market opportunity might be. There's nothing scarier than a startup that is doing the same thing it originally set out to do 18 months later. Look at Microsoft. In no way was it founded to do what it does today. It got started to do software tools. At one point, Oracle was a consulting shop.

What you absolutely don't want is to get stuck with "second-system syndrome." Something works wonderfully the first time, and then for the next version, the engineers have a list of 130 things they want to fix. If they try to do everything, deadlines end up slipping two years or more. Either the next version never comes to market or what finally emerges is so big and cumbersome that no one wants it.