redmond_aerialACCIDENTAL EMPIRES

CHAPTER THIRTEEN

ECONOMICS OF SCALE

We’re at the ballpark, now, and while you and I are taking a second bite from our chilidogs, this is what’s happening in the outfield, according to Rick Miller, a former Gold Glove center fielder for the Bosox and the Angels. When the pitcher’s winding up, and we figure the center fielder’s just stooped over out there, waiting for the photon torpedoes to load and thinking about T-bills or jock itch endorsements, he’s really watching the pitcher and getting ready to catch the ball that has yet to be thrown. Exceptional center fielders use three main factors in judging where the ball will land: what kind of pitch is thrown where in the hitter’s zone, the first six inches of the batter’s swing, and the sound of the ball coming off the bat.

So Miller watches, then listens, then runs. Except for the most routine of hits, he never looks up to see the ball until he gets to where it is going to land; he just moves to where it should land. This technique works well except at indoor ballparks like the Seattle Kingdome. The acoustics in the Kingdome are such that Miller has to watch the ball for the half-second after it leaves the bat, just like the rest of us would do, and it costs about 20 percent of his range.

I will never be a Rick Miller. Bob Cringely, the guy who says it shouldn’t take six years to learn to be a blacksmith, wasn’t talking about what it would take to be the world’s best blacksmith. I could start today taking Rick Miller lessons from Rick Miller, and in six years or even sixty years could never duplicate his skills. It’s a bummer, I know, but it’s just too late for me to make the major leagues. Or even the Little League.

Back in elementary school, when all the other boys were shagging flies and grounders until sundown, I must have been doing something else. For some reason—I don’t remember what I was doing instead—I never played baseball as a kid. And because I never played baseball, I’ll always be in the stands eating chilidogs and never be in center field being Rick Miller.

There’s only one way to be a Rick Miller, and that’s to start training for the job when you are 8 years old. Ten years and 200,000 pop flies later, you are ready for the minor leagues. Three years after that, it’s time for the majors—the show. There are no short-cuts. A robot, a first-string goalie from the New York Rangers, or a genetically engineered boy from Brazil could not come into the game as an adult and hope to be a factor. Remember Michael Jordan’s dismal performance in the baseball minor leagues.

Even if Rick Miller himself was doing the teaching, it wouldn’t work. He’d say, “Hear the way the bat sounds? Quick, run to the right! Hear that one? Run to the left! This one’s going long! Back! Back! Back!”

But they’d all sound the same to you and me. We’d have to hear the sounds and learn to make the associations ourselves over time. We’d need those 200,000 fly balls and the ten years it would take to catch them all.

There is no substitute for experience. And except for certain moves that I surprised myself with one evening years ago in the back seat of a DeSoto, there are no skills or knowledge that just spontaneously appear at a certain preprogrammed point in life.

My mother is unaware of this latter point. She bought me white 100 percent cotton J.C. Penney briefs for the first eighteen years of my life and then was surprised during a recent visit to learn that I hadn’t spontaneously switched to boxer shorts like my dad’s. She just assumed that there was some boxer short gene that lay dormant until making itself known to men after high school. There isn’t. I still wear white 100 percent cotton J.C. Penney briefs, Mom. I probably always will.

And now we’re back in the personal computer business, where there is also no substitute for experience, where good CEOs do not automatically generate from good programmers or engineers, and where everything, including growth, comes at a cost.

**********

For computer companies, the cost of growth is usually innocence. Many company founders, who have no trouble managing twenty-five highly motivated techies, fail miserably when their work force has grown to 500 and includes all types of workers. And why shouldn’t they fail? They aren’t trained as managers. They haven’t been working their way up the management ladder in a big company like IBM. More likely, they are 30 years old and suddenly responsible for $30 million in sales, 500 families, and a customer base that keeps asking for service and support. Sometimes the leader, who never really imagined getting stuck in this particular rut, is up to the job and learns how to cope. And sometimes he or she is not up to the job and either destroys the company or is replaced with another plague—professional management.

There comes a day when the founders start to disappear, and the suits appear, with their M.B.A.s and their ideas about price points, market penetration, and strategic positioning. And because these new people don’t usually understand the inner workings of the computer or the software that is the stuff actually made by the company they now work for, the nerds tend to ignore them, thinking that the suits are only a phase the company is going through on its way to regaining balance and remembering that engineers are the appropriate center of the organization.

The nerds look on their nontechnical co-workers—the marketing and financial types—as a necessary evil. They have to be kept around in order to make money, though the nerds are damned if they understand what these suits actually do. The techies are like teenagers who sat in the audience of the “Ed Sullivan Show,” watching the Beatles or the Rolling Stones; the kids couldn’t identify with Ed, but they knew he made the show possible, and so they gave him polite applause.

But the coming of the suits is more than a phase; it’s what makes these companies bigger, sometimes it’s what kills them on the way to being bigger, but either way it changes the character of each company and its leaders forever.

The great danger that comes with growth is losing the proper balance between technology and business. At the best companies, suits and nerds alike see themselves as part of a greater “us.” That’s the way it was at Lotus before the departure of Mitch Kapor. Kapor could use his TM training and his Woodstock manner to communicate with all types. As Lotus grew and some products were less successful than expected, Kapor found that the messages he was sending to his workers were increasingly dark and unpleasant. Why be worth $100 million and still have the job of giving people bad news? So Mitch Kapor gave up that job to Jim Manzi, who was 34 at the time, a feisty little guy from Yonkers who was perfectly willing to wear the black hat that came with power. But Manzi as CEO lacked understanding of the technology he was selling and the people he was selling it with.

Manzi was Lotus’s first marketing vice-president, and he was the one who came up with the idea of marketing 1-2-3 directly to corporations, advertising it in business and general interest publications that corporate leaders, rather than computer types, might read. The plan worked brilliantly, and 1-2-3’s success was a phenomenon, selling $1 million worth in its first week on the market. But for all his smarts, Manzi was also a suit in the strongest possible sense. He sold 1-2-3 but didn’t use it. He boasted about his lack of technical knowledge as though it was a virtue not to understand the workings of his company’s major product. His position was that he had people to understand that stuff for him. Being able to sell software so brilliantly while lacking a technical understanding of the product was supposed to make him look all the smarter, a look Manzi wanted very much to cultivate.

While he was totally reliant on people to explain the lay of the computer landscape, Manzi didn’t know any more about how to use people than he did 1-2-3. Five development heads came and left Lotus in four years, and each of these technical leads consistently went from making Manzi “ecstatic” with their progress to being “dickheads.” Programming went from being down the hall to “in the lab,” which could just as well have been in another country, since Manzi had no idea what was going on there, and his technical people felt no particular need to share their work with him either. At least three major products that would come to have bottom-line importance for Lotus were developed without Manzi’s even knowing they existed because of his isolation from the troops.

When all of Manzi’s emphasis was on 1-2-3 version 3.0, the advanced spreadsheet that was delayed again and again and would not be born, a couple of programmers working on their own came up with 1-2-3 version 2.2, a significant improvement over the version then shipping. By the time Manzi even knew about 2.2, its authors had quit the company in disgust, leaving behind their code, which eventually made millions for Lotus when it was finally discovered and promoted.

“May I join you?” Manzi once asked a group of Lotus employees in the company cafeteria, “or do you hate me like everyone else?”

Poor Jimmy.

“Manzi is a bad sociopath—one that is incapable of using friends,” claimed Marv Goldschmitt, who ran Lotus’s international operations until 1985. “A good sociopath manipulates and therefore needs to have people around. Manzi, as a bad sociopath, sees people inside Lotus as enemies. He could have kept a lot of good people who left the company—and he should have but saw them as dangerous.”

This attitude extended even to strategic partners. When Compaq Computer used some of his remarks in a promotional video without his permission, Manzi tore apart his own Compaq computer, stuffed it in a box, and shipped the parts directly to Rod Canion, Compaq’s CEO, with a note saying he didn’t want the thing on his desk anymore.

With 1-2-3 the largest-selling MS-DOS application, it would have been logical for Manzi to have had a good relationship with Microsoft’s Bill Gates. Nope. Having barely escaped being acquired by Microsoft back in 1984, Manzi had no good feelings for Gates. He specifically tried to keep Lotus from developing a spreadsheet to work under Microsoft’s Windows graphical environment, for example, because he did not want to do anything to assist Gates. But trying to stop a product from happening and actually doing so were different things. Down in the lab, even as Manzi railed against Windows, was Amstel, a low-end Windows spreadsheet developed at Lotus without Manzi’s ever being aware of it. Amstel eventually turned into 1-2-3/Windows, an important Lotus product.

Manzi saw himself in competition with Gates. Each man wanted to be head of the biggest PC software company. Each wanted to be infinitely rich (though only Gates was). They even competed as car collectors. Gates and Paul Allen dropped $400,000 each into a pair of aluminum-bodied Porsche 959 sports cars, so Manzi also ordered one, even though the cars were never intended to be sold in the United States. Allen and Gates took delivery of serial numbers 197 and 198, and Manzi would have got number 201 except that Porsche decided to stop production at 200. Beaten again by Bill Gates.

Alienated by choice from the rest of his company, Manzi churned the organization with regular reorganizations, claiming he was fostering innovation but knowing that he was also making it harder for rivals to gain power. Taciturn, feeling so unlovable that he could not trust anyone, Manzi created development groups of up to 200 people, knowing they would be hard to organize against him. Such large groups also guaranteed that new versions of 1-2-3 would be delayed, sometimes for years, as communication problems overwhelmed the large numbers of programmers.

The bad news about Lotus was slow in coming because the installed base of several million users kept cash flowing long after innovation was stifled. In 1987, right in the middle of this bleak period, Manzi earned $26 million in salary, bonuses, and stock options. But the truth always comes out, and in the case of Lotus, even Manzi eventually had to take a chance and trust someone, in this case Frank King, an old-line manager from IBM who definitely did understand the technology.

Frank King had been the inventor of SQL, an innovative database language that somehow appeared from the catacombs of IBM. Like nearly every other clever product from IBM, SQL had been developed in secret. King and his group developed SQL in a closet, lied about it, then finally showed it to the big-shots who were too impressed to turn the product down. Frank King knows how to get things done.

It was King who set up five offices at Lotus, one in every development group, and spent a day per week in each. It was King who discovered the hidden products that had been there all along and who got the long-delayed, though still flawed, Lotus 1-2-3 3.0 unstuck. It was Mitch Kapor and Jim Manzi who made Lotus and Frank King who saved it.

**********

In a company with a strong founder, power goes to those who sway the founder. In most companies, this eventually means a rise of articulate marketers and a loss of status for developers. That’s what happened at Aldus, inventors of desktop publishing and PageMaker, which turned out to be the compelling application for Apple’s Macintosh computer.

Aldus was founded by a group of six men who had split away from Atex, a maker of minicomputer-based publishing systems for magazines and newspapers. Atex had an operation in Redmond, Washington, devoted to integrating personal computers as workstations on its systems. When Massachusetts-based Atex decided to close the Redmond operation, Paul Brainerd, who managed the Washington operation, recruited five engineers to start a new company. They set out to invent what came to be called desktop publishing. Brainerd contributed his time and $100,000 to the venture, while the five engineers agreed to work for half what they had been paid at Atex.

Aldus was originally pitched as a partnership, but, typically, the engineers didn’t pay attention to those organization things. That changed one day when they all met at the courthouse to sign incorporation papers and the others discovered that Brainerd was getting 1 million shares of stock while each of the engineers was getting only 27,000 shares. Brainerd was taking 95 percent of the stock in the company giving the others 1 percent each. The techies balked, refused to sign, and eventually got their holdings doubled. For his $100,000, Brainerd bought 90 percent of Aldus.

Paul Brainerd was into getting his own way.

“It’s common for founders of these companies to be abusive,” said Jeremy Jaech, one of the five original Aldus engineers. “Certainly Brainerd, Jobs, and Gates are that way. I looked up to Paul as a father figure, and so did most of the other founders and early staff. I was 29 when we started, and most of the others were even younger. We came to see Paul as the demanding father who could never be pleased. It was like a family situation where, years later, you wonder how you let yourself get so jerked around over what, in retrospect, seems to be so unimportant. ‘Why did I care so much [about what he thought]?’ I keep asking myself.”

Brainerd’s money lasted six months, long enough to build a prototype of the application and to write a business plan. The first prototype was finished in three months; then Brainerd went on the road, making his pitch to forty-nine venture capitalists before finding his one and only taker. The plan had been to raise $1 million, but only $846,000 was available. It was just enough.

It wasn’t clear how venture capitalists could assign a value to software companies, so they tended to shy away from software, thinking that hardware was somehow more certain. The VCs were always worried that someone else writing software in another garage would do the same thing, either a little bit quicker or a little bit better. The money people were so uninterested that Brainerd found that most of the VCs, in fact, hadn’t even read the Aldus business plan.

You need a big partner to start a new product niche in the personal computer business. For Aldus, the partner was Apple, which needed applications to help it sell its expensive LaserWriter printer. Apple’s dealers had been burned by the failure of the Lisa, the HP Laserjet printer was out on the market already and much cheaper, and no software was available that used LaserWriter’s PostScript language. The situation didn’t look good. Apple was worried that the LaserWriter would bomb. Apple needed Aldus. Three LaserWriter prototypes were given to software developers in September 1984. One went to Lotus, one to Microsoft, and one to Aldus, so Apple had a clear sense of the potential importance of PageMaker, the first program specifically for positioning text and graphics on a PostScript printed page.

Aldus’s original strategy was to show dealers that PageMaker would sell hardware. They kept the number of dealers small to avoid price cutting. The early users were mainly small business-people. Compared to going outside to professional typesetters to prepare their company newsletters and forms, PageMaker saved them time and money and gave them control of the process. It was this last part that actually drove the sale. Traditional typesetting businesses didn’t pay much attention to customers, so small businesspeople were alienated. With Pagemaker and a LaserWriter, they no longer needed the typesetters.

Aldus surprised the computer world by taking what everyone thought was a vertical application—an application of interest only to a specialized group like professional typesetters—and showed that it was really a horizontal application—an application of interest to nearly every business. Companies didn’t produce as many newsletters as spreadsheets, but nearly all produced at least one or two newsletters, and that was enough to make the Macintosh a success. There was nothing like PageMaker and the LaserWriter in the world of MS-DOS computing.

The first release of Pagemaker was filled with bugs, but microcomputer users are patient, especially with groundbreaking applications. There was talk inside the company of holding PageMaker back for one more revision, but the company was out of money and that would have meant going out of business. Like most other products from software start-ups, PageMaker was shipped when it had to be, not when it was done. Three months later, a second release fixed most of the bigger problems.

By the late 1980s, Aldus was a success, and Paul Brainerd was a very wealthy man. But Brainerd was trapped too. When Aldus was started, the stated plan was to work like hell for five years and then sell out for a lot of money. That’s the dream of every start-up, but it’s a dream that doesn’t hold up well in the face of reality. Brainerd had discussions with Bill Gates about selling out to Microsoft, but those talks failed and Aldus had no choice but to go public and at least pretend to grow up.

Companies used to go public to raise capital. They needed money to build a new steel mill or to lay a string of railroad track from here to Chicago, and rather than borrow the money to pay for such expansion, they sold company shares to the investing public. That’s not why computer companies go public.

Computer companies generally don’t need any money when they go public. Apple Computer was sitting on more than $100 million in cash when it went public in 1979. Microsoft had even more cash than that stashed away when it went public in 1986. These numbers aren’t unusual in the hardware and software businesses, which have always been terrific cash generators.

It’s not unusual at all for a software company with $50 million in sales to be sitting on $30 million to $40 million in cash. Intel these days has about $8 billion in sales and $2 billion in cash. Microsoft has $2.8 billion in sales and more than $900 million in cash. Apple, with $8 billion in sales, is sitting on a bigger pile of cash than the company will even admit to. At the same time it is laying off workers in the United States and moaning about flat or falling earnings, Apple admits to having $1 billion in cash in the United States, and has at least another billion stashed overseas, with no way to bring it into the United States without paying a lot of taxes. None of these companies has a dime of long-term debt.

This habit of sitting on a big pile of money originated at Hewlett-Packard in the 1940s. David Packard figured that careful management of inventories and cash flow could generate lots of money over time. Hanging on to that money meant that the next emergency or major expansion could be financed entirely from internal funds. Now every company in Silicon Valley manages its finances the H-P way.

What’s ironic about all these bags of money lying around the corporate treasuries of Silicon Valley is that although the loot provides insurance for hard times ahead, it actually drags down company earnings. “Sure, I’ve got $600-700 million available, but who needs it?” asked Frank Gaudette, Microsoft’s chief financial officer. “I’ve got to find places to put the money, and then what do I make—12-15 percent, maybe? Better I should churn the money right back into the company, where we average 40 or 50 percent return on invested capital. We’re losing money on all that cash.”

But not even Microsoft can grow fast enough to absorb all that money, so the excess is often used to buy back company stock. “It increases the value of the outstanding shares, which is like an untaxed dividend for our shareholders,” Gaudette said.

While computer companies are aggressive about managing their cash flow, they are usually very conservative about their tax accounting. Most personal computer software companies, for example, don’t depreciate the value of their software; they pretend it has no value at all. IBM carries more than $2 billion on its books as the depreciable value of its software. Microsoft carries no value on its books for MS-DOS or any of its other products. If Microsoft managed its accounting the way IBM does, its earnings would be twice what they are today with no other changes required. That’s why Wall Street loves Microsoft stock.

So computer companies don’t go public to raise money; they go public to make real the wealth of their founders. Stock options are worthless unless the stock is publicly traded. And only when the stock is traded can founders convert some of their holdings in Acme Software or Acme Computer Hardware into the more dull but durable form of T-bills and real estate—wealth that has meaning, that makes it worthwhile for cousins and grandnephews to fight over after the entrepreneur is dead.

Bill Gates never wanted to take Microsoft public, but all those kids who’d worked their asses off for their 10,000 shares of founders’ stock wanted to cash out. These early Microsoft employees—the ones walking around wearing FYIFV lapel buttons, which stand for Fuck You. I’m Fully Vested—were millionaires on paper but still unable to qualify for mortgages. They started selling their Microsoft shares privately, gaining the attention of the SEC, which began pushing the company toward an initial public offering. Gates eventually had no choice but to take Microsoft public, making himself a billionaire in the process.

Companies that don’t grant stock options to employees have no trouble staying private, of course. That’s what happened at WordPerfect Corp., the leading maker of PC word processing software. Started in Utah by a Brigham Young University computer science professor in partnership with the director of the BYU marching band, WordPerfect now has more than $300 million in annual sales yet only three stockholders. The company also has more than $100 million in cash.

Paul Brainerd was one of those founders who wanted to stabilize his fortune, giving his kids something to fight over. Overnight, Brainerd became very rich by making a public company of Aldus Corp. But Brainerd’s secure fortune, like that of every other entrepreneur turned CEO of a public company, came at a personal cost. Start-ups are built on the idea of working hard for five years and then selling out, but public companies are supposed to last forever. CEOs of public companies stand before analysts and  shareholders, promising ever higher earnings from now until the end of time. Like other entrepreneurs-turned-corporate honcho, Brainerd is rich, but he’s also trapped at Aldus, by both money and ego. His enormous holdings mean that it would take too long to sell all that stock unless he sells the whole company to a larger firm. And there is an emotional cost, too, since he believes that he can’t do it again. This is his chance to be a big shot. Brainerd has a large ego. He needs power, and if he left Aldus, what would he do?

There are two kinds of software companies; one develops new concepts and pioneers new product areas, and the other works at continuing the evolution of an existing product. These two types of companies, and the people they need to do their jobs well, are very different. Aldus used to be the first type, but today it is very much the second type of company, and the people of Aldus have had to change to fit. Their primary job is to keep improving PageMaker. Public companies with successful products put their money into guaranteed winners, which means upgrades to the core product and add-on programs for it. At Aldus today, all the other products are viewed as supplements to PageMaker, which must be protected. PageMaker is the cash cow.

Aldus programmers concentrate on new versions of PageMaker, while most other applications sold under the Aldus name are actually bought from outside developers. Freehand, a drawing package, came from a company in Texas called Altsys, which gets a 15 percent royalty on sales. Persuasion, a package for automating business presentations, is another Aldus product gotten from outside, this time with a 12 percent royalty but a bigger down payment. Although it pays 15 percent royalties for products developed outside, Aldus, like most other established software companies, budgets only 6 or 7 percent of sales for internal development projects. This is frustrating for the programmers inside because they are responsible for the vast majority of sales yet are budgeted at a rate only half that of acquired products. Aldus expects more of them yet gives them fewer resources.

Successful software companies like Aldus quickly become risk averse. They buy outside products for lots of money with the idea that they are buying only good, already completed products that are more likely to succeed. Internal development of new products suffers because of the continual need to revise the cash cow and because the company is afraid of spending too much money developing duds.

For an example of such risk aversion, consider Aldus’s abortive entry into the word processing software market. Although PageMaker was a desktop publishing program, it originally offered no facility for inputting text. Instead, it read text files from other word processing packages. When Aldus was working on PC PageMaker, which would run under Microsoft Windows on MS-DOS PCs, it seemed logical to add text input, and even to develop Aldus’s own word processing package for Windows. Code-named Flintstone, the Aldus word processor would have had a chance to dominate the young market for Windows word processors.

By early 1988, a prototype of Flintstone was running, though it was still a year from being ready to ship. That’s when Bill Gates gave Paul Brainerd a demonstration of Word for Windows—Microsoft’s word processor that would compete with Flintstone. Gates told Brainerd that Word for Windows would ship in six to nine months, beating Flintstone to market. Afraid of going head to head against Microsoft, Brainerd canceled Flintstone. Word for Windows finally hit the market two years later.

**********

While Lotus was a technology company with good marketing that became a marketing company with okay technology, some computer and software companies have always been marketing organizations, dependent on technology from outside. Even these firms can run aground from problems of growth and the transition of power.

Look at Ashton-Tate. George Tate’s three-person firm contracted in 1980 to market Wayne Ratliff’s database program called Vulcan. Vulcan was a subset of a public domain database called JPLDIS that Ratliff, an engineer at Martin-Marrietta Corp., had used on mainframe computers running at the Jet Propulsion Laboratory in Pasadena. Some have claimed that Ratliff wrote JPLDIS, but the truth is that he only wrote Vulcan, which had a subset of JPLDIS features combined with a full-screen interface, allowing users to seek and sort data by filling out an on-screen form rather than typing a list of cryptic commands.

Ratliff tried selling Vulcan himself, but the load of running a one-man operation while still working at Martin-Marrietta during the day was wearing. Rather than quit his day job, Ratliff pulled Vulcan from the market, later selling marketing rights to George Tate. The product was renamed dBase II and became the most successful microcomputer database program of its time. Ratliff, who had hoped to earn a total of $100,000 from his relationship with Tate, made millions.

Ratliff worked for Martin-Marrietta until 1982 while continuing to develop dBase II in his spare time, as required by his contract with Ashton-Tate. There was no program development at all done at Ashton-Tate’s headquarters in Torrance, which was strictly a marketing and finance operation. By 1983, when introduction of the IBM PC-XT with its hard disk drive made clear how big a success dBase II was going to be in the PC-DOS market, Tate bought rights to the program outright and installed Ratliff in Torrance as head of development for dBase III.

It was at this time, when dBase III was as successful in the database market as Lotus 1-2-3 was among spreadsheets, that George Tate snorted one line of cocaine too many and died of a heart attack at his desk. Suddenly Ashton-Tate had a new CEO, Ed Esber, who had been hired away from Dan Fylstra’s VisiCorp to be marketing vice-president only a few weeks before. Esber, who was 32, was a marketer, not a technologist, and except for the vacuum created by Tate’s sudden death probably would not have been considered for the jobs of president, chairman, and CEO that fell to him.

In his new position, Esber made the mistake of tipping the balance of power too much in the direction of marketing, then toward finance, and all at a major cost in lost time and bad technology. Marketing figured out what the next program was supposed to do; detailed specifications were written and then distributed to a large number of programmers, who were expected to write modules of code that would work together. Only they didn’t work together, at least not well, in part because the marketers didn’t have a clear concept of what was possible and what wasn’t when the specs were written. These were marketers acting as metaprogrammers and not knowing what the hell they were doing.

Ashton-Tate began to have the same problems bringing out its next version of dBase—dBase IV—that Lotus was having with 1-2-3 version 3.0. The company bought outside products like Framework, an integrated package that competed with 1-2-3, and MultiMate, a word processor, but even these were allowed to bog down in the bureaucracy that resulted from an organization whose leaders didn’t know what they were doing.

“Esber thought management of a development group meant going over the phone bills and accusing us of making too many long-distance calls,” said Robert Carr, who wrote Framework and was Ashton-Tate’s chief scientist in those days.

When dBase IV finally shipped, it was nearly two years late. Worse, it didn’t work well at all. The product was seriously flawed and the programmers knew it. Still, the product was shipped because the finance-oriented company was worried about declining cash flow. They shipped dBase IV only to help sales and earnings. But bad software is its own reward; the resulting firestorm of customer complaints nearly drove the company out of business.

Ratliff left, and competitors like Nantucket Software and Fox Software created dBase-like programs and dBase add-ons that outperformed the original. Despite having 2.3 million dBase users and over $100 million in the bank, Esber was forced out during the spring of 1990 when Ashton-Tate posted a $41 million loss.

The week after he was pushed from power, Ed Esber had his first-ever dBase programming lesson.

**********

The suits first appeared at Microsoft in 1980, right around the time of the IBM deal. Prior to that time, Microsoft was strictly a maker of OEM software sold to computer companies and maybe to the occasional large corporation. Those corporate deals were simple and often clumsily done. In 1979, for example, Microsoft gave Boeing Commercial Airplane Co. the right to buy any Microsoft product for $50 per copy, until the end of time. Today most Microsoft applications sell in the $300 to $500 range, ten years from now they may cost thousands each, but Boeing still would be paying just $50.

When Microsoft realized its mistake, a blonde suit in her twenties named Jennifer Seman was sent alone to do battle with Boeing’s lawyers. First she dropped the Boeing contract off with Microsoft’s chief counsel for a legal analysis; when she came back a few days later to talk about the contract, it was on the floor, underneath one leg of the lawyer’s chair, still unread.

That was the way they did things when Microsoft was still small, when what people meant when they said “Microsoft” was a group of kids wearing jeans and T-shirts and working in a cheap office near the freeway in Bellevue. The programmers weren’t just the center of the company in those days, they were the company. There was no infrastructure at all, no management systems, no procedures.

Microsoft wasn’t very professional back then. A typical Microsoft scene was Gordon Letwin, a top programmer, invading the office of Vern Raburn, head of sales, to measure it and find that Raburn’s office was, as suspected, three inches larger than Letwin’s. Microsoft was a company being run like a fraternity, and, as such, it made perfect sense when one hacker’s expense account included the purchase of a pool table. Boys need toys.

But Bill Gates knew that to achieve his goals, Microsoft would have to become a much larger company, with attendant big company systems. He didn’t know how to go about creating those systems, so he hired a president, Robert Towne, from an electronics company in Oregon called Tektronix, and a marketing communications whiz, Roland Hansen, who had been instrumental in the success of Neutrogena soap.

Towne lasted just over a year. The programmers quickly identified him as a dweeb, and ignored him. Gates continually countermanded his orders.

Hansen’s was a different story. He dealt in the black magic of image and quickly realized that the franchise at Microsoft was Bill Gates. Hansen’s main job would be to make Gates into an industry figure and then a national figure if Microsoft was to become the company its founder imagined it would be. The alternative to Gates was Paul Allen, but the co-founder was too painfully shy to handle the pressure of being in the public spotlight, while Gates looked forward to such encounters. Paul Allen’s idea of a public persona is sitting with his mother in front-row seats for home games of his favorite possession, the Portland Trailblaz-ers of the NBA.

Even with Gates, Hansen’s work was cut out for him. It would be a challenge to promote a nerd with few social skills, who was only marginally controllable in public situations and sometimes went weeks without bathing. Maybe Neutrogena soap was a fitting precedent.

To his credit, by 1983 Hansen managed to get Gates’s face on the cover of Time magazine, though Gates was irked that Steve Jobs of Apple had made the cover before he did.

Massaging Bill’s image did nothing for organizing the company, so Gates went looking for another president after Towne’s departure. By this time, Paul Allen had left the company, suffering from Hodgkin’s disease, and Gates was in total control, which meant, in short, that the company was in real trouble. Fortunately, Gates seemed to know the peril he was in and hired Tandy Corporation’s Jon Shirley to be the new president of Microsoft. Shirley was not a dweeb.

Gates had been Microsoft’s Tandy account manager when Shirley was head of the Radio Shack computer merchandising operation. Although Shirley had made mistakes at Tandy, notably deciding against 100 percent IBM compatibility for its PC line, that didn’t matter to Gates, who wasn’t hiring Shirley for his technical judgment. Technology was Gates’s job. He was hiring Shirley because he had successfully led the expansion of Tandy’s Radio Shack stores across Europe. Shirley, who joined Radio Shack when he was a teenager, had literally watched Charles Tandy build the chain from the ground up to 7,000 stores worldwide. Shirley was to management what Rick Miller was to center field. Growing up at Radio Shack meant that Shirley knew about organization, leadership, and planning—things that Bill Gates knew nothing about.

Shirley’s job was to build a business structure for Microsoft that both paralleled and supported the product development organization being built by Gates based on Simonyi’s model. The trick was to Create the systems that would allow the company to grow without diverting it from its focus on software development; Microsoft would ideally become a software development company that also did marketing, sales, support, and service rather than a marketing, sales, support, and service company that also developed software. This idea of nurturing the original purpose of the company while expanding the business organization is something that most software and hardware companies lose sight of as they grow. They managed it at Microsoft by having the programmers continue to report to Bill Gates while everyone on the business side reported to Shirley.

This was 1983. Microsoft was the second largest software company in the PC industry, was incredibly profitable, was growing at a rate of 100 percent per year, and had no debt. Microsoft was also a mess. There was no chief financial officer. The only company-wide computer system was electronic mail. Accounting systems were erratic. The manufacturing building was the only warehouse. The company was focused almost entirely on doing whatever the programmers wanted to do rather than what their customers were willing to pay for them to do.

One example of Microsoft’s getting ahead of its customers’ needs was the Microsoft mouse, which Gates had introduced not knowing who, if anyone, would buy it. At first nobody bought mice, and when Shirley started at Microsoft, he found a seven-year supply of electronic rodents on hand.

Then there was Flight Simulator, the only computer game published by Microsoft. There was no business plan that included a role for computer games in Microsoft’s future. Bill Gates just liked to play Flight Simulator, so Microsoft published it.

In one day, Shirley hired a chief financial officer, a vice-president of manufacturing, a vice-president of human resources. a head of management information systems, and a head of investor relations. They were all the same person, Frank Gaudette, a wisecracking New Yorker hired away from Frito-Lay, who at 48 became Microsoft’s oldest employee. Six years later, Gaudette was still at Microsoft and still held all his original jobs.

To meet Gates’s goal of dominating world computing, Microsoft had to expand overseas. The company was already represented in Japan by ASCII, led by Kay Nishi. In Europe, operations were set up in the United Kingdom, France, and Germany, all under Scott Oki. Though Apple Computer didn’t know it, Microsoft’s international expansion was financed entirely with payments made by Apple to finance a special version of Microsoft’s Multiplan spreadsheet program for the Apple IIe. Apple needed Multiplan because Lotus had refused to do a version of 1-2-3 for the IIe. Because Charles Simonyi had designed Multiplan to be very portable, moving it to the Apple lie was easy, and the bulk of Apple’s money was used to buy the world for Microsoft.

Even with real marketing and sales professionals finally on the job, accounting and computer systems in place, and looking every bit like a big company, Microsoft is still built around Bill Gates, and Bill Gates is still a nerd. During Microsoft’s 1983 national sales meeting, which was held that year in Arizona, a group of company leaders, including Gates and Shirley, went for a walk in the desert to watch the sun set. Gates had been drinking and insisted on climbing up into the crook of a giant saguaro cactus. Shirley looked up at his new boss, who was squatting in the arms of the cactus, greasy hair plastered across his forehead, squinting at the setting sun.

“Someone get him down from there while he can still father children,” Shirley ordered.