Softwar

by Matthew Symonds

  At a strategic level, GE was already convinced that going down the integrated suite route made more sense than best of breed, and anyway, time wouldn’t allow for anything else. Nor was there much doubt that Oracle would get the nod—Release 11 had already been installed elsewhere in GE, for example, in medical systems, and was doing all right. In July, a month after the bulldozers started clearing earth at Veresegyház, the big discussion was whether to take the chance of going with 11i only weeks after its release. Kipp asks, “Did we really want to take the risk? Putting up a new plant in Central Europe and having to hire a workforce from scratch is risky enough. In the end, what swung it was the superior upgrade path. Because it was all Internet, Oracle convinced us that we could seamlessly get new versions and new features as they became available. The attitude here is that everything’s going to the Internet, so we just decided on making a leap of faith. The whole thing had to be installed in just a few months, and we kind of figured that Oracle couldn’t afford to let us down.” It was a shrewd calculation.

  The contract was signed in late October with a clear understanding that if Oracle delivered at Veresegyház, GE would start rolling out 11i across its thirty or so other turbine plants. Within a month, the project was in trouble. Kipp was using a local IT consulting firm named Ejiva, which had done work for GE in Hungary before but was now struggling. Worried, Kipp told Nardelli that at this rate he would never hit his milestones. Nardelli and Rice decided to call Larry Ellison. Kipp says, “Within a week we had thirty to forty Oracle consultants. It was like a MASH unit arriving.” Kipp was working from hastily converted offices on a run-down industrial estate that he christened “the bunker.” “There were lots of things going on: hiring people, riding the builders, working with local government. Until December, I thought things were under control. On the other side of town we had all these people from GE and Oracle—there were about eighty of them in an office designed for forty. They were working through the night, mapping processes and figuring out what they wanted the software to do.”

  When the consultants invited Kipp over for what is known in the business as a “conference room pilot” (CRP), the first of a series of trials that the software has to be put through before it is deemed to be ready, he realized that things were far from “under control.” Basically, a CRP involves running the business in a simulation mode—putting the systems that will run the manufacturing and financial operations through their paces. Within the first couple of hours, as he moved from room to room, Kipp felt increasingly uncomfortable. “It just dawned on me that a twenty-six-year-old recent graduate from Budapest Technical University and three guys from Oracle in San Jose were designing how to run my business. I realized I’d been focusing on the wrong thing. Pouring concrete may be important, but what was going to determine on a daily basis sixty to eighty percent of what my people would be doing was being written into the software. And I thought, shit, have we got the right people making the right decisions?”2

  What convinced Kipp that he was in trouble was a simulation that involved picking up hypothetical parts from the stockroom to assemble on one of the turbine units. By the time they had taken all the parts out on the shop floor, they realized they were working on the wrong job. Kipp says, “The software was designed that way because people from my finance team were putting in very strict requirements for tracking costs. As I learned more about it, I said this is crazy, if this were to go live we’ll just crash the whole business. So we had a meeting in the hallway that night at eleven P.M. There were people hanging off the doors. I’ll never forget it. I said, ‘What are you trying to do? We don’t need this level of sophistication and complexity to run this business.’ Because to have this very structured financial cost system means that everything has to be done differently on the manufacturing side.”3

  Kipp knew that if they continued down this track there was no way they’d meet the promised “go-live” date of March 7. His message was to make it simpler. The problem was that GE’s finance and manufacturing specialists from America had believed they were expected to add value by getting Oracle to customize the software to deliver all the features Kipp might possibly want. Unfortunately, they hadn’t thought to ask him, and the Oracle consultants were trained never to say no to a customer.4

  Kipp says that what had happened was the key to understanding why the installation of business software systems so often goes wrong. “You decide on buying a piece of software that pretty much has what you need to run your business. But then the consultants ask, ‘What would you really like it to do, because we want to optimize your business processes.’ So the guy says, ‘Well, what I’d really like is for it to do the following . . .’ and they have the same conversation with the finance team and the manufacturing team and the sales guys. And of course, the consultant doing his job writes down all the little other things that would be great to have. So Oracle has this great product that will do eighty percent of what you want just by throwing the right switches in the software—no special code or anything—and it all falls down because someone asks that innocent little question: ‘What else would you like it to do?’ ”

  Kipp says that if they hadn’t been sucked down the path of customization, they might have gotten the software installed in three rather than five months. The key was the commitment of his own time—as much as half of each working day for at least a couple months—and that of his dozen best managers. He says that it’s no good expecting junior people to make the right decisions about how the business should be run, and it’s crazy to ask them to do it. He admits that the strain of it nearly killed them, but that you have to use the people “who’ve really got that mental expertise to configure the system right from a process standpoint.”

  It’s exactly what Ellison’s been saying all along. Just after Veresegyház went live, John Rice called Kipp over to Orlando to share his experiences with some of GE’s other senior managers. Kipp says, “I told them: don’t let it happen to you. Just take the software, install it. If you simplify and standardize your business processes along the lines of the way the software’s configured, you’re going to get eighty percent of what you want. If you want to go back after the next year or three to get some special feature you need, you can still do that.” By way of example, he told Oracle that if he needed improved data on quality, the chances were that most of their other customers would too. Oracle’s response was that if he could wait six months they’d write it and give it to him in the next upgrade. Instead of saying that was too long and demanding a customized solution, Kipp said, “OK . . . the world’s not going to fall apart in six months. Plus, I’ve got supported software that’s in all their upgrades—not some special code that I have to take responsibility for.”5 It’s another example of what Ellison calls his “80 percent rule”: take 80 percent functionality now rather than risk never finishing by holding out for a 100 percent perfect solution that exists only in the imagination.

  If the lessons for GE are clear, they should also be for Oracle. When the project nearly came off the rails because the consultants were obliging requests by junior managers that were at odds with the approach laid down by John Rice, there was, in Kipp’s words, “no one on the Oracle side who was going to tell the customer, ‘No, you can’t have that.’ ” Kipp says that GE’s managers, culturally used to demanding “everything and now” needed to be told more clearly what’s in their best interests. “If they’re offered this feature, they’re not going to say, ‘Well, I don’t need that.’ But somebody with authority has to be able to say no.”

  * * *

  1. LE writes: Working up close and personal with GE was like going to a real-world business school. What separates GE from most companies is its willingness to force change after change into its business in a relentless, never-ending pursuit of efficiency.

  2. LE writes: This was the turning point of the project. Once senior management got involved—deeply involved—we started making real progress.

  3. LE wri
tes: Only senior management looks at and makes decisions for the business as a whole. Individual department managers optimize their internal departmental functions without much sympathy for or understanding of the burdens they may be placing on other departments. That’s why it’s treacherous to automate one department at a time. You’re better off taking a self-contained segment of the business, like the Hungary plant, and automating the end-to-end business flows that cross departmental boundaries.

  4. LE writes: Today, Oracle’s consultants are trained to say “no” when it comes to customizing the code. The less a customer changes our software, the more of it they’re likely to buy. We make more money selling software than consulting. When consulting is your moneymaker, it’s a lot harder to say “no.”

  5. LE writes: Our five thousand applications developers put lots of new features into the E-Business Suite based on customer priorities. This close collaboration was what made it possible for most of our customers to avoid the complexity and cost of custom code.

  13

  HILL BY HILL

  May 2001

  A year after the E-Business Suite was launched with such high hopes, the world has become a very different and, for Oracle, more hostile place. It would be tempting to describe Ellison as “beleaguered,” but that overworked word doesn’t really fit. He is still cracking jokes and generally enjoying life. At Oracle there’s no hint of crisis. There’s a good deal of grumbling about the evaporating value of stock options and the disappearance of bonuses, but at least Oracle isn’t laying off swaths of workers like many companies in Silicon Valley. There’s an atmosphere of sober determination. There are enough veterans around at Redwood Shores—guys who have been through far worse times than this—to steady the ship and calm the anxieties of young engineers who really did think, as Oracle’s advertising insisted, that the Internet had changed everything, including the abolition of the business cycle.

  We’re on the road again—or, more accurately, flying at 45,000 feet in Ellison’s Gulfstream—heading for more customer meetings and CEO roundtables. This time we’re going to Chicago and then on to Fisher Island, an exclusive resort near Miami. It’s a strange moment in the E-Business Suite campaign. Ellison is convinced that the bugginess and instability are a thing of the past. But the media and the analysts have gotten their teeth into that story and won’t let go. The only way to turn it around will be increasing numbers of references from happy customers. But the bad publicity and the economy have persuaded many to delay planned installations. For Ellison, it’s a waiting game.

  Nonetheless, he says, “We’ve had some very significant E-Business Suite ‘go-lives’ lately. Alcoa went live at a major location in less than five months, start to finish, without any modifications. Liberty Mutual went live. More and more customers are going live on 11i. That’s really important. People love the idea—the concept—of a complete and integrated E-Business Suite. But the question is ‘Does it really work?’ And if it does, ‘How do I get there from where I am today?’ ”

  I’m curious to know how many of these customers, if any, are actually using the whole suite. In Hungary, GE Power Systems was deploying the ERP part of the package—the manufacturing and financial apps plus the hot new order management system—but they had no need yet for the fancy CRM piece of the action. What about Alcoa? “Well, when I say that some company is using the E-Business Suite I mean they’re using several E-Business Suite applications, not all of them. Nobody, not even Oracle, uses all of the E-Business Suite. We don’t use manufacturing because we don’t manufacture anything. However, I still describe Oracle as an E-Business Suite user. If a company is using a number of the release 11i applications, I call it an E-Business Suite user. GE Power is running a lot of our applications but not all of them. The same is true of Alcoa. Both GE and Alcoa are industrial giants in the process of modernizing their businesses. They’re using E-Business Suite applications to modernize their supply-chain processes, their procurement processes, their manufacturing processes, and their accounting processes. Step by step, one location at a time, they’re modernizing their companies using the E-Business Suite.”

  Alcoa will be an enormous Oracle installation, involving all 362 of its plants worldwide. “I started participating in regular conference calls at GE Power after we ran into some problems a couple of months into the project. We quickly solved those problems and stayed on top of new ones as they arose. During every call we go through major product and implementation issues in excruciating detail. I’ve learned a lot in the process. I’m going to try to spend more time with Alcoa because it’s a huge project that automates all the different Alcoa businesses, from making and selling boxcars of aluminum ingots to consumer-packaged aluminum foil. Another big project I’m watching closely is at POSCO, the world’s largest and most profitable steel company. They’re doing a big-bang go-live in July. They’re switching over the entire company all at once. Not many companies take that approach because of the risks involved. But if you pull it off, it’s the by far the fastest way to go.1 It takes iron balls to go that route, but hey, they’re a steel company so. . . . Anyway, I always recommend a safe, step-by-step implementation process: define a set of standard processes, bring up one location or one division using those processes, then start adding more locations or more divisions. That’s what we’re doing at GE, that’s what we’re doing at Alcoa, that’s what we did at Oracle. But you must automate one business at a time, not one department at a time. One-department-at-a-time automation, best-of-breed systems that don’t work across departmental boundaries, that’s how companies got in the mess they’re in today. Companies need a road map to move from their detached departmental systems with their disconnected processes and fragmented data to a modern integrated system built on a single global database.”

  While the campaign for 11i continues, Ellison’s main focus is about to switch back to his first love, the database, and the launch of 9i, the latest and most significant new version of Oracle’s flagship product for many years. “I’m back to spending about half my time on the database. 9i is a very big deal because for the first time our database clustering works with real applications—Oracle applications, SAP applications, Siebel applications, custom applications, everything. We’ve been trying to do this for more than a decade. It’s not an easy problem. Microsoft’s and IBM’s clustering technology works fine for performance benchmarks, but it’s useless for real applications. They run clustering tests and then run an ad bragging about their database performance. That’s good marketing, but marketing can only carry you so far. Their database clusters can’t run real-world applications, so their clustering performance tests are meaningless. Eventually people will notice.

  “But our clustering isn’t just a marketing scam. Our clustering works with real-world applications, and that fundamentally changes the database business forever. But it doesn’t stop there: clustering changes the entire computer industry. The computer industry is based on the principle of scale: a big application runs on a big, expensive computer. But if you can make a bunch of cheap little computers ‘look’ like a big computer, then you’ve altered the economics of the industry. And that’s exactly what we’ve done with 9i clusters. We’ve been able to take a group of cheap Windows or UNIX computers, cluster them together, and run SAP applications, PeopleSoft applications, Siebel applications, and everything else. We can do it. Nobody else can. Nobody.

  “Before, if you wanted to run faster, you bought a faster machine. Once you bought the fastest machine there is, you’re done, you can’t run any faster. With Oracle 9i clustering, those limits no longer apply. You can have multiple database computers running your SAP application. To add performance, you add another computer; to add reliability, you add another computer. Rather than having one expensive computer, you have four inexpensive computers, you run faster, you run more reliably, and you run much more cheaply using an array of small machines. We broke the code. We made it work. That’s not good news for people who buil
d huge expensive machines. IBM says, ‘Gee, our database software costs less per processor than yours.’ We say, true enough, but you have to buy software for twice as many processors and then spend five times more for hardware to run IBM’s database. IBM would have to pay you to run their database for it to make economic sense. Even if they gave you all the software and hardware for free, it still wouldn’t be nearly as fast or reliable as an Oracle 9i cluster. So Oracle9i is the ‘last database.’ This is game, set, and match in the database business. Oracle 9i clusters make it very difficult for IBM and Microsoft to compete on performance, reliability, or cost. We’ve got to do a good job getting the clustering message out there, but I think within three to six months it will become obvious that Oracle 9i really is the last database”2

  The trouble is that people think of databases as mature technology and not as sexy as applications. How will Ellison communicate his excitement? “Well, it’s pretty interesting if you can show eight Compaq PC servers running faster and more reliably than the largest IBM mainframe. And those PC servers could be running Linux. People say Linux isn’t reliable enough for enterprise computing. Now, it doesn’t have to be reliable. Not if you use lots of them. If one fails, who cares? Not fast enough? Use more of them. 9i is great computer science, not just another software product. It’s, I hate the expression, a ‘breakthrough.’ Bill’s been talking a lot about Microsoft’s horizontal scalability [using a cluster of little machines to get performance rather than one big machine] strategy. But they can’t make it work. We’ve got horizontal scalability for every application ever written; they have horizontal scalability for one and only one application—their custom version of the TPC-C benchmark [an industry standard database performance test]! It’s amazing. And, it will take IBM and Microsoft a long, long time to catch up. Maybe ten years, but probably never.”