Softwar

by Matthew Symonds

  Ellison had no illusions about how critical the order management system was: “The order management system is the heart of the sell side of our e-business suite. Order management is the funnel that captures the information about the sale. The better your order management system, the more information it captures. Our new system captures all the details about the products sold, contractual obligations, payment terms, everything. It manages marketing promotions, vendor rebates, and tracks what salespeople should be compensated on the deal. The flexible pricing engine handles complex and arbitrary rules—like buy two and get the third one free. And the automatic configurator keeps people from buying the wrong combination of products—like computer components that won’t work together. It’s a great system. But when it first came out, it had a lot of bugs.”

  Ellison was not acting on his own, however. Ron Wohl fully supported his view that putting in the new order management system was a risk worth taking. Wohl says, “Technically, we really had to include it to fulfill the objective of the E-Business Suite. If it had the flexibility to meet modern e-business practices, you’re in great shape, if it doesn’t, you don’t really have an E-Business Suite.” What’s less clear is whether Wohl was giving Ellison an accurate picture of the state the work on the order entry system was in. Mark Barrenechea complains that although Ellison had instructed both his and Wohl’s teams to standardize all their outward-facing applications on a completely rewritten customer master called Trading Community Architecture (TCA), “the ERP organization had kind of missed it along the way.” Critically, the new order management system wasn’t using the TCA. A close colleague of Wohl told me, “With most people, Ron comes over as cautious to a fault. But not with Larry. He always wants to look good in front of Larry, and he often ends up telling him what he wants hear.” When Wohl told Ellison that he believed that everything was ready for prime time, Ellison wasn’t going to argue.3

  Another mistake in the testing process was that a handful of chosen customers were trying a beta version of 11i under artificial conditions. Instead of the software running in their own data centers, it was running on Oracle’s own computers. Consequently, the severe upgrade problems that some customers would experience after the official launch were masked. Mark Jarvis says, “We were completely in control. We were solving their problems for them. We were doing seventy to eighty percent of what the customer should have been doing himself. It was a mistake.” Some good did come out of this botched testing program: it encouraged thinking about delivering software as a Web-based service, allowing the evolution of the earlier Business Online initiative, aimed at small businesses, into the current E-Business Suite Outsourcing.4

  Even with Oracle’s vaunted rapid implementation procedures for installing 11i, it would be at least three or four months before any significant “go-lives”—the moment when a business switches its operations to new software. Oracle itself would be moving onto 11i only in stages, between the fall and the end of the year. When I met Ellison for lunch in mid-October at his home in Atherton, he appeared not to have a care in the world. It was the calm before the storm. Clad only in a Sayonara T-shirt and shorts and enjoying the sunshine, he oozed confidence about every aspect of the business. The new version of the database that was nine months away was “unbelievable,” while 11i was building unstoppable momentum. Even with the clouds gathering over Silicon Valley as the dot coms imploded one by one and the hype about Internet-driven B2B faded, Ellison was sanguine. Just as in 1990, he seemed to think that any economic slowdown might pass Oracle by. His argument was that the E-Business Suite was such an extraordinary “cost-cutting engine” that customers were likely to need it all the more to preserve their profits in difficult selling conditions.

  After lunch, Ellison showed me a draft of the chairman’s letter to shareholders that he was writing for the annual report. “Tell me what you think when you’ve had a chance to read it,” he said. It was a five-thousand-word account of Oracle’s e-business transformation and a marvelously clear exposition of the philosophy of computing that had driven it. One passage went to the heart of what he was trying to do: “To eliminate inefficiency, we had to make information easier to find and easier to share. But how? The solution was quite simple. If we put all our information in one place—a global database on the Internet—then everyone would know where to look to find the information they needed. While conceptually simple, this single unified database approach required fundamental changes to our application software. It turned out to be a massive engineering effort involving thousands of computer programmers. But when we finished the Oracle E-Business Suite, it was the first and only set of applications to work with a single global database. We also developed several new applications, so that the E-Business Suite would be complete. Today, the suite includes every application you need to run your business—marketing, sales, supply chain, manufacturing, customer service, accounting, human resources—everything. It works in every country and in every language. The E-Business Suite is the first and only complete set of applications ever to have been built.” Ellison signed off with this: “The more we know, the more rational our decisions. Oracle’s getting smarter. Oh, the possibilities. . . .”

  If Ellison had known what was coming, he might have restrained himself a little more. In October, the first reports from the field described early 11i implementations afflicted with more than usual bugginess. At the Oracle Applications Users Group conference in Honolulu at the end of October, the complaints were legion. Attendees griped about a lack of reliable information concerning the status of applications, malfunctioning modules—especially CRM and order management—a constant stream of bug-fix patches, and poor support from the customer service organization, which, according to some, hadn’t been properly trained on the workings of 11i. In particular, customers that were upgrading from previous versions of Oracle applications were finding that theirs was not a favorable position to be in. During the conference, word got around that Oracle had released around five thousand patches for 11i. That was an oversimplification—many of the patches were to improve functionality rather than to fix bugs and many fixed multiple bugs, but the five thousand figure stuck in the memories of journalists and analysts, who would gleefully include it in almost every piece on Oracle for the best part of a year. Mark Jarvis says, “The five thousand bugs haunted us at least until midsummer of 2001.”

  Over the next couple of months, customer complaints poured in. Instead of working on the planned features and functions to flesh out the suite’s capabilities as planned, development was increasingly engaged in a feverish attempt to produce patches that would fix the bugs that seemed to be flying at them from all sides. It was all made much worse by the sheer number of implementations that were under way. Yet, there was little air of crisis at Redwood Shores. Ellison says, “11i was a huge new product, so I knew there would be a lot of bugs, but there was no way to know how many there were or how long it would take us to fix them. I thought we could fix most of the bugs pretty quickly, during the controlled release phase. I was way too optimistic. It just kind of got away from us, and that became a very big problem.”

  The software business likes to play the blame game when products don’t work as billed. At about the same time that users started beating up Oracle, i2, a specialist in supply-chain planning, was involved in a fierce “debate” with one of its most high profile customers, the sports goods firm Nike. According to Nike, the malfunctioning of the i2 order entry software it had bought to power its demand and inventory management system had been the cause of an earnings shortfall. Not so, said i2; Nike had overcustomized. It was not unusual for Oracle’s senior executives to assume at first that customers’ problems might be the result of shoddy work by integration partners. They also had plenty of anecdotal evidence that Ellison’s well-publicized arguments against excessive customization and his pleas to put in Oracle’s software as “plain vanilla” were being ignored by some customers and their integrators.

&nbs
p; Out in the field, however, George Roberts had a rather different take on where problems lay: “Larry is very fact-based, very logical with decisions and processes. Customers saying they are having issues is one thing, but being able to track down and point directly at the issue is another thing. It’s a new product. Does the customer understand it? Is there a third party implementing it who’s been trained on it? The problem was that Larry really thought that his development leaders had it done. He listens to them closely, and that’s what they had been telling him. You know, the baby’s never ugly. We should have done two things: installed it in-house first and engaged with the early implementations at an executive development level. I think if either of those things had happened, we would have rapidly found out that it wasn’t done. As it was, it took us until November to realize we had issues here and to start to mobilize.”

  By this time, people from Ron Wohl’s and Mark Barrennechea’s organizations were being parachuted into troubled implementations—“customer escalations”—and were learning about “issues” firsthand. But what rammed the message home was that Oracle itself was, at last, beginning to go live on one part of the suite after another. By December, Oracle was running financials, some HR, field sales automation, and telesales on 11i. Barrenechea says, “The next really big push was to get in contracting and order management over New Year’s. What happened was one of the most dramatic moments I’d seen during my five years at Oracle. Everyone except Ron knew that we could not place an order with the new system.5 And Larry, knowing we can’t place an order, not knowing when we would be able to place an order, and knowing that once we turned off the old bespoke system that there was no going back, said, ‘We’re going live anyway. I need a forcing function, so off we go.’ So we upgraded, and Oracle went down. We were down for fourteen days. For fourteen days that January, Oracle could not place a single order. It was frightening. But we got it to work. Larry was right. It was a great forcing function. At that point I became acutely aware of the customer dissatisfaction and immediately started drilling down into the implementations we were doing.”

  Ellison’s view is less dramatic: “We knew that we were going to go down for some time, so we scheduled the installation over the Christmas/New Year’s holiday. I had decided on a massive big-bang implementation: we were going to upgrade all our software at once. I never, ever recommend that to our customers. But I wanted to get all our latest application software installed, up, and running at Oracle, and I wanted to do it fast. We should have done it before we released 11i to customers, but we didn’t. Okay, that was my mistake. Now I wanted to remedy that mistake as fast as possible. That meant installing 11i at Oracle the first opportunity we had. The fact that we got everything up and running in a couple of weeks was a major accomplishment. We uncovered a number of problems in our software, and we fixed a number of problems in our software. By the time we had finished the Oracle implementation, we had fixed most of the order management problems.”

  A revised release (11.5.3), which incorporated everything Oracle’s developers had painfully learned, was rushed out to customers in early February. It was the third revision since May and by far the most reliable. Barrenechea boasted that 11i was now at least as solid as any rival enterprise software products, most of which had been around much longer. Cap Gemini Ernst & Young, one of the biggest consulting and integration houses, declared in March that “Oracle had made significant improvements and fixes.” As a result, the suite was now sufficiently stable for Oracle to begin using the CRM modules in its own practice.

  However, just as it took time for the problems with the E-Business Suite to surface, the fact that Oracle had now overcome many of the defects did little to stanch the flow of bad news. The technology press and the analysts were on the hunt for stories about 11i implementations that had run into trouble. Fairly typical was this story of woe in the March 12, 2001, issue of InformationWeek, a magazine aimed at information managers and integrators: “International courier DHL Worldwide Network NV in Brussels, Belgium, for example, was unable to complete transactions based on the euro currency using Oracle’s 11i financial applications. Problems included defective currency-to-currency payment functions and slow performance. The glitches put the courier three months behind on its plans to deploy the suite in 12 European locations. DHL came within two days of a drop-dead January 31 deadline before Oracle engineers delivered software patches that fixed the flaws, says Jeremy Young, DHL’s finance business-process manager. ‘No software is perfect in its early release, but it took seven months to resolve these problems,’ Young says.”

  Young also happened to be the president of the Oracle Applications Users Group, which been locked in a bitter feud with Oracle about who had the right to stage conferences—the users or the company. In 1995, Oracle had taken over the conferences, which had previously been run by the database user group, launching Oracle OpenWorld later that year. OpenWorld had been a huge marketing success. The number of attendees had gone from 2,500 or so to events that regularly pulled in 40,000 people, blowing the doors off some of the biggest venues in the world. Mark Jarvis reckoned that with the E-Business Suite soon to debut, Oracle needed to do something similar with applications.

  The issue was that Oracle was putting a lot of executive time and money into the OAUG’s conferences but not getting enough of what it wanted out of them. Jarvis says, “The OAUG ran two events a year in the U.S., and we were highly restricted in what we could do at them. The OAUG insisted that it was very independent. For example, they didn’t want Oracle logos and signage around—a bit odd. There were other peculiarities, but their primary goal was to have users talk to other users, while our goal was to get out there, market our products, and educate people on all of the new stuff. The third factor was that a lot of the guys in the user group are not always on the leading edge. They are more interested in running a version of our software from three years ago and they want to talk about that, whereas, of course, we wanted to always talk about the latest thing so that we can get people upgraded and sell them more stuff. We were at cross-purposes. In early 2000, we told the user group of our desire to run a conference and we asked them if they wanted to be part of it. They took it as a declaration of war.”

  The result was a stalemate. Oracle went ahead with its plans to hold its first AppsWorld conference early in the first quarter of 2001 in Paris and New Orleans. The OAUG continued to hold its conferences, albeit without the official presence of Oracle executives and no keynote speeches from Ellison. Jarvis says, “They also started a very active PR campaign against us, continually emphasizing their ‘independence.’ ” It proved very successful. Whenever industry analysts or press wanted to find out how Oracle customers were getting on with 11i, they started going to the user group for references rather than Oracle. “OAUG board members were suddenly being quoted in the press saying negative things, knocking whatever we were claiming, and attacking the company’s whole approach. They wanted to point out to everybody that they weren’t on 11i yet. We’d be saying in all our messages that people are starting to use 11i, and they’d be saying, ‘But we don’t know anybody who’s on it.’ They provided a counterpoint to everything we said.”

  If Oracle hadn’t been vulnerable to criticism because of the early 11i quality issues, the damage caused by OAUG’s revolt would have been easy to limit. But with plenty of people happy to take a swipe at “arrogant” Oracle for claiming more than it had been able to deliver, the user group was the perfect stick to beat it with. Even quite early on, Oracle could have pointed analysts in the direction of many pretty successful 11i implementations if they’d been interested, but it was much more fun to get the horror stories from the OAUG. Ellison maintains that the fight was never with the users as such but with the user group management: “Some of the officers in the users’ group were not customers or users of our applications at all—they were independent consultants who made their living running the Oracle Applications Users Group meetings. We were threaten
ing their livelihood by running our own meetings. It was unfortunate that at the same time we were having the quality problems with 11i, we were also having the spat with the user group. They bashed us every chance they got.”

  If Oracle had gone on making its numbers, the bad blood with the user group would have been no more than a minor irritant. Even the fallout from the buggy early versions of 11i would in all probability have been containable. But on the eve of the first AppsWorld show in Paris, the moment when Oracle had planned to deliver to the world a confident message about its winning applications strategy, a series of almost uniformly gloomy reports from Wall Street analysts thudded on the mat, including one by Morgan Stanley’s Chuck Phillips. He argued that Oracle’s forecast in December of 15 to 20 percent year-over-year revenue growth in its core database business, along with 75 percent for applications, would be extremely hard to achieve given the fact that no fewer than thirty of Oracle’s dot-com customers had gone bust and that around 12 percent of the firm’s total database license revenue had come from dot coms in the previous year. A prescient report by Bank of America analysts noted, “Though management seems confident in the progress Oracle has made thus far in the quarter, we remind investors that a good amount of the quarter’s business remains to be done. The next three weeks will tell the tale.” As shareholders chewed over the possibility that Ellison might have exaggerated the extent of Oracle’s immunity to the forces that were blighting other tech companies, Oracle’s stock price plunged by 13 percent, to $23.56. In response, Oracle reiterated that it was standing by its bullish forecast.