A spin-off from a hardware firm named Britton-Lee, Sybase had come up with the first database specifically designed for the new paradigm of client/server computing. With Sybase’s architecture the database lived in one computer, the server, while the application would live in the client machine. Although Oracle’s Version 5 supported client/server and 5.1 contained a feature that enabled so-called distributed queries, allowing a single query to access data stored in multiple locations, Oracle’s database had not been designed from the beginning for the separation of client/server. Ellison says, “The key thing that Sybase had that we didn’t was a programmable server, the ability to write programs called stored procedures that ran inside the database. This allowed Sybase to build and advertise two features—referential integrity and two-phase commit—that they had and we lacked. These key differentiators gave Sybase immediate traction in the market.”
Two-phase commit made it possible to execute a transaction spanning a number of databases on different computers. If you have two databases and you make a savings withdrawal out of one computer and a checking deposit into the other with a separate database, that whole thing would have transactional integrity—either both updates would occur or neither would. Referential integrity provides for an automatic existence check to make sure that if, say, you enter “department 57,” there is in fact such a thing in the department tables as 57. One of the things that customers liked about Sybase was that with its programmability and stored procedures it was a database that could be used to enforce business rules across an organization.
Although Sybase was a lot smaller than Oracle, its advertising was getting plenty of attention from both Oracle’s customers and, to an even greater extent, Oracle’s sales force, who were demanding something to hit back at Sybase with. A few months before Oracle’s IPO in early 1986, Ellison had declared that the company would double its revenues every year. Indefinitely. A year later, with revenues of $131 million, Oracle claimed to be the world’s biggest database software company, having overtaken a rapidly failing Cullinet on the way down. However, Ellison, sensing that the original code base would not allow the product to evolve as it would need to, had ordered a complete rewrite of the kernel, effectively the entire “bottom half” of the database’s software, for the next major release. “Version 6 included a complete rewrite of the most complex part of the database. It took so long and had so many bugs, it almost killed the company. But its advanced architecture laid the foundation for much of what differentiates our database today—including clustering. Version 6 was something we had to do to move ahead of the competition.6 But while we were trying to finish Version 6, Version 5.1 fell further and further behind Sybase.”
With Oracle’s ambitious revenue targets, its sales force (now under the relentless Gary Kennedy) screaming for a new product, and Ellison’s fear that Sybase was gaining real traction, the development team found itself subjected to appalling pressure to declare Version 6 ready to ship. Ellison says, “There was an overwhelming sense of frustration. We had delayed Version 6, and then delayed it again and again and again. Delaying it one more time seemed unbearable. We had done a lot of testing, and I thought that it was good enough to go. I was wrong. When we did release Version 6 in 1988, it had more bugs than any of us believed was possible.”
The decision to ship Oracle 6 unfinished was the beginning of a sea of troubles that came close to overwhelming Ellison and destroying the company.
* * *
1. LE writes: No longer young, I am still looking for answers. Today the science I find offering the most insight into our world is molecular biology.
2. LE writes: Virtually all of these so called high-risk decisions, from building the first commercial relational database to building the first complete and integrated E-Business Suite, involved innovation. Whenever you travel an untraveled path there’s risk.
3. LE writes: It turns out you never really “finish” a software product. Programmers never run out of good ideas. Customers never stop asking for more features. We’ve been working on the Oracle database for twenty-five years, and we’re still not “finished.”
4. LE writes: I have no ambivalence about good salespeople—the ones that understand our technology and how to apply it to solve customer problems. Those salespeople are among our most precious assets. I admire those salespeople. I try to be one of them myself. The other type of salesman—the hired gun, the too-creative deal maker—not only do I not admire them, I think they’re dangerous.
5. LE writes: I replaced Mike Seashols with Gary Kennedy for the very simple reason that Gary’s numbers were better than Mike’s. Mike’s sales results were good. Gary’s were astounding. I have never met a more driven person than Gary Kennedy. Gary would rather die than lose. For years that extreme drive worked to Oracle’s benefit. But in the end it proved to be a two-edged sword.
6. LE writes: Rewriting the database software for Version 6 was a much bigger effort than the Version 3 rewrite. But in both cases the rewrite led to software quality problems, and it took some time before the software stabilized. The only other project of similar scope and complexity to the Version 6 database rewrite was the Release 11i applications rewrite, which gave birth to the E-Business Suite. These rewrites are periodically necessary if you want to make major advances in your technology. Invariably there is a lot of pain before the payoff, but the payoff—for both Oracle and our customers—has been huge.
5
TO THE LIMIT
To what extent was Ellison responsible for releasing an unfinished product because of a treadmill of his own creation? In 1985, with sales running at $24 million, Ellison had declared that from that point on Oracle would double its revenues each year. Amazingly, for the next five years, Oracle did precisely that, hitting sales of $916 million in 1990. But Ellison argues that the sales projections were not his overriding concern. “We were small, and we had to grow fast or die. But the most effective way to grow sales is to make the product better. That’s why I spend most of my time with our engineering teams.”1
Maybe. But the notion that Ellison was measuring success purely in terms of Oracle’s technological prowess is far-fetched. The original idea of Oracle’s being a vehicle to provide him with a nice life, a reasonable amount of money, and plenty of time to goof off whenever he felt like had quickly evaporated. In his early forties, Ellison was willing to say and do whatever was necessary to defeat Oracle’s rivals. Apart from engineering genius, brashness and aggression were the most prized qualities at Oracle in the late 1980s. The only thing that seemed to matter was market share, and increasingly Ellison didn’t seem to care how reckless Oracle became to get it.2
However, the strain of achieving such phenomenal growth and the increasingly dangerous business practices needed to underpin it nearly killed the company. When Oracle went public with a market value of $270 million on March 12, 1986 (by an extraordinary coincidence, just twenty-four hours before Microsoft’s IPO), Ellison, with 39 percent of the stock, was suddenly wealthy beyond his wildest expectations. But if anything, the fact that he now had something to lose intensified the demands he was prepared to make on himself and his company. Shipping Oracle 6 before it was ready was a risk, Ellison calculated, that had to be run.
Despite its bugginess and lack of the key features that Sybase boasted about, Oracle 6 introduced many new features, including row-level locking, which meant that a transaction performing writes would lock only the affected rows and not the entire table, thus greatly improving the throughput of the system with a large number of users. That helped to make Version 6 the first scalable database that could grow as a customer’s business grew. Row-level locking, combined with another Version 6 feature, multitasking, enabled Oracle to perform extremely well on the symmetrical multiprocessor (SMP) computers, which were just starting to become important and are now the norm in enterprise computing, some high-end boxes today having as many as sixty-four CPUs. Unfortunately, the market was not yet crying out for that
kind of scalability, and the instability of Version 6 disqualified it as the enterprise-strength system it aspired to be. The launch was itself a bad omen. In grand style, Oracle chartered Concorde to fly the press and analysts to Bermuda for a formal briefing on the wonderful new database, but because of a technical fault the plane had to turn back.
At about the same time, Oracle was getting serious about the emerging market for the packaged applications to automate businesses that became collectively known as enterprise resource planning (ERP) software. Late in 1985, Ellison had persuaded Jeff Walker to come to Oracle and set up an applications division—even then some people were predicting the imminent saturation of the database market, and Ellison was determined that Oracle’s growth shouldn’t depend on one product. Walker was about the same age as Ellison and had founded and run his own software business, Walker Interactive. But unlike Ellison he had taken money from venture capitalists, and when his firm had badly missed its numbers during a brief technology spending slowdown his investors had concluded that Walker, although technically brilliant, was not enough of a businessman to manage the company. He says, “I left because my venture people would have driven me out.” Pretty soon after, he had a conversation with Ellison. “Larry said that his strategy for getting into the applications business would have been to have hired somebody from Walker Interactive because it was the emerging technology leader. He had never thought he could get me [at the time Walker Interactive was slightly bigger than Oracle]. I told him that I thought I was unmanageable and I didn’t really want to work for anybody. He assured me that I would not really be working for him, that we would be merely collaborators.”
When Walker arrived, there was just one other person in the “applications division.” True to his word, Ellison gave Walker a free hand to work out a strategy, hire people, and design the initial applications. Walker, who now runs another troubled applications company called TenFold, has a very high opinion of his own abilities. He claims that in the six years to his departure from the company in 1991, “Oracle came from being a nonplayer to being the industry leader in applications. SAP hadn’t really entered the market at that point with their R/3 client/server product. They were selling R/2, their hosted product, but they weren’t really as successful as Oracle.” It’s an interesting point of view. However, just running Oracle’s applications operation was not a big enough job for the multitalented Walker. After a spell, when he both had his development responsibilities and had become head of marketing, in late 1986 Ellison asked him to take on the job of CFO while staying as head of apps. Walker says, “Larry quickly realized that I had management strength and experience that he needed on his team. It was not something that I had done before, but I think that I did a credible job as CFO.”
Don Lucas remembers, “Larry just announced to me that Jeff was going to be CFO. Jeff’s a nice person. I like Jeff. But I spent hours, probably an average of an hour a day, on the telephone trying to talk through finance with him . . . But he’s an engineer,3 and, to be honest, it was a personality issue, he never got it. Larry’s approach was to delegate massively, completely, almost totally. That’s the good news—as long as he gets good people, it really works. But when he gets a person like Jeff, who is a very smart guy but not a brilliant CFO . . .”
Walker’s appointment wasn’t the strangest thing that happened at Oracle during that time. Gary Kennedy was under appalling pressure to meet the Herculean targets that were implied by Ellison’s commitment to continually double revenues. Kennedy had become convinced that the corporate legal department was actively preventing deals by taking too long to vet sales contracts. The answer, he argued, was to give U.S. sales its own contracts and finance organization. Ellison acquiesced. “Gary told me he couldn’t get the job done because the legal department didn’t report to sales and they were slow in processing deals. I said, okay, you can have the legal department. What kind of CEO lets the salespeople write their own contracts? I just didn’t know any better.”4
The buzzword at Oracle became “decentralization”—another manifestation of the extreme form of delegation that Ellison appeared to favor. Noosheen Hashemi, an intense young Iranian who was running the financial administration for U.S. sales, was asked to write something called “The Decentralization White Paper” that was meant to define the checks and balances that would be needed in Kennedy’s brave new world. She says, “Decentralization occurred, and none of the things mentioned in the white paper were executed. Not one. All the regions implemented their own processes and started doing their own thing. Decentralization happened because U.S. sales didn’t want corporate to be watching over their shoulder. They didn’t want to have to justify what they were doing. It was like, here’s the deal, book it, let’s move on.”
Hashemi insists that Gary Kennedy never asked her personally to do anything unethical, but all hell was breaking loose. Essentially, U.S. sales was now responsible for oversight of its own deals, reviewing contracts and booking revenue without any interference from the corporate center. The consequences were myriad. Huge discounts were offered to induce customers to buy software they wouldn’t need for years; risks were taken in selling to companies that didn’t have the money to pay their bills; exotic barter deals were entered into. In one, Oracle was given a couple of jets by Israeli Aircraft Industries in exchange for software.5
To make matters worse, Kennedy wanted a huge expansion of the sales force and told Hashemi to organize the recruitment of 1,500 additional salespeople within nine months. Not surprisingly, few of those taken on had time to be properly trained or inducted before they were sent out into the field. Unless the new recruits started selling quickly and bringing in revenue, the cost of such breakneck hiring would be crippling.
It’s hard to say how much more Jeff Walker could have done, other than to threaten to resign as CFO. But he made things worse by supporting the extreme form of delegation that Ellison had carelessly encouraged. Amazingly, even today he defends it: “Larry had found a management team of very strong individual guys that could work together and yet autonomously run their areas. It was almost a feudal environment, in terms of feudal lords over whom Larry as king only had very weak controls. He chose to make them weak. I think it was a wonderful decision on his part because it allowed Oracle to come out of nowhere and become this huge market force.”
Walker made things still harder for himself by believing so passionately in the efficacy of his own financial applications that he thought he could operate with only a fraction of the people he really needed. Eventually, in response to Walker’s concerns and Hashemi’s whistle-blowing to Ellison, in 1989 Ellison asked her to go and work for Walker in an attempt to clean things up. Hashemi says of Walker, “There he was, building products to use internally and sell to the outside world. He wanted to shove whatever he made in applications down my throat. To prove how fabulous his applications were, he wants to fire two thirds of the hundred and fifty I have working for me in finance.” Walker was also to some extent conflicted in reining back the increasingly desperate sales antics of Kennedy’s people. His bonus was tied to the sales of applications. If Walker came down too hard on sales, the applications he was so proud of would appear less successful.
A good test of how well a sales organization and its financial department are working together is the efficiency with which money is collected from customers. By the end of fiscal 1989, although revenue had doubled to $584 million, trade receivables were running at $262 million. A year later, the corresponding figures were $971 million and $468 million. Hashemi says that Walker wasn’t prepared to face up to the problem: “Instead of acknowledging it, addressing it, going through every statement, Walker was still lying to himself.”
Kennedy had developed what he liked to call a “commitment” culture at Oracle. He was committed to achieving the revenue targets set by Ellison, and in turn his own team were committed to realizing the quotas set for them. Commitment meant that you did what you had undertaken to d
o. Period. It sounded good, but the consequences of Kennedy’s terrifying insistence on commitment were appalling.
If ever a company was heading for a crash, it was Oracle. And it duly came, almost exactly four years after the company’s IPO. Steve Imbler, Walker’s number two in finance and, unusually for Oracle, a qualified accountant, had become so anxious about the ballooning receivables that he had started to trawl through as many of the dubious-looking contracts as he could. By the time the third quarter of fiscal 1990 was due to close on February 28, he had found $15 million worth of deals in which the likelihood of payment was minimal. What followed was more than a devastating year of purgatory for both Oracle and Ellison.
Imbler managed to persuade Walker and Ellison to write down the missing $15 million when the results for the quarter were reported on March 27. Although revenue, at $236 million, was still up by 54 percent over the year before, net income was more or less unchanged, while earnings per share, at 18 cents, were little better than half the upper end of analysts’ forecasts. The following day, Oracle’s stock lost more than 30 percent of its value, falling to $17.50 at a time when such precipitous single-day drops were much less common than they are now. It should have been the warning that Ellison needed to put Oracle’s house in order. But far from being chastened and repentant, Ellison told investors that he fully expected to make up the lost ground in the final quarter and that Oracle would be back on what he called its “annual plan.” Eighteen months earlier, at the company’s annual strategy meeting, with just five slides, Ellison had made a presentation that he titled the “Five in Five” plan: “I was simply extrapolating the annual doubling of revenues. We would reach $5 billion in five years. I had absolutely no idea how absurd and naive the plan was.”
Softwar Page 11