Resolve and Fortitude : Microsoft's ''SECRET POWER BROKER'' breaks his silence

by Joachim Kempin

  First released in ’93 and by now in version 4.0, NT had fully matured—most ready for prime time. Intel’s new CPU, introduced in November of ’95, called Pentium Pro, a powerful successor to the 80486, made it shine further. There were two flavors of NT in the market: a server version, which had evolved from the original LAN-Manager product, and one designed for workstations. These were the higher-end PCs used for graphic design and complex analytical and engineering tasks. My group focused solely on selling the workstations derivate.

  NT, being more capable than Windows 95, justified in my opinion a higher price. To determine its value, I studied competitive products as I had done for 95. One was OS/2. Its pricing was hard to determine. IBM had published a retail price of approximately $250. Yet that version was hardly selling. IBM’s OEM pricing was unknown to me. I estimated it to be between $60–100. We further knew that IBM gave lots of OS/2 copies away for free or sold them in bulk at extremely favorable conditions to enterprise customers. Therefore, the real price fluctuated probably between $0 and $250. The other competitive OS was the one running on the Apple Macs. It arrived installed without being priced out separately. Only retail upgrade pricing was obtainable. Last but not least, there were several UNIX systems in the market. One was Linux; it was given away for free and mainly used on servers—not what I was aiming for. UNIX versions, bundled with workstations, were not priced separately except their upgrades. SCO’s version had a retail price tag of around $400–$550 depending on the included libraries. With scant information and such huge price fluctuations, I asked Bill for advice.

  We started our discussion by investigating future OEM business aspects and immediately agreed that growing my business was tied to two main factors: overall PC sales and a steady increase in NT penetration, assuming it could demand a higher price than 95. The first depended on the economy, affordability, and the PC’s hopefully enduring usefulness—which we could only indirectly influence and which in hindsight took over ten years to be seriously challenged. Increasing NT penetration, on the other hand, was directly correlated to how much enterprise workstation demand we would be able to create. I had a job to do.

  Pricing NT higher than Windows 95 was a foregone conclusion for us. Just how much was the next issue to attack? Taking OS/2 and UNIX pricing into account and knowing NT contained most of the same goodies and was capable of running a more diverse number of applications, we eventually arrived at a $100 price; it sounded like a bargain to us. It sure is an easy one to get accustomed to if used as a from-the-hip multiplier, but could we really extract that much?

  Our so-called bargain turned out to be extremely hard to sell. For every hundred units of Windows in 1995/96, only two to three units of NT were sold. The reasons: a combination of price resistance by the OEMs and so-far tepid customer demand because UNIX was favored on workstations. What made switching enterprise customers even harder was that they were still suffering through the OS/2 disaster. After IBM’s late solo had failed, they weren’t willing to easily trust MS’s intrepid lead. Stuck, I asked my marketing guys to break the deadlock and create a program incenting OEMs to crack the enterprise workstation segment with NT in mind.

  With NT-powered systems capable of supporting higher-end solutions and their potential to increase revenue and profit margins, OEMs should have jumped on that opportunity a long time ago. Instead of helping us to actively convert UNIX users into NT fans, they waited for us to initiate the move. A typical reaction considering they often saw themselves as mere distributors of our OS—proving a point I often emphasized: OEMs do not create demand for any OS! Ultimately, applications do that only when running on attractive computer platforms, and the top-notch ones were indeed running on NT workstations. This gave us a fighting chance to succeed.

  My marketing group soon identified a list of OEMs who each had their own corporate sales forces and should be interested in our endeavor to increasingly push NT for workstation usage. One of them was IBM. It would have been nice to have her on board, but even with repairs on the way, the company was still vying for warp-crowned revival and immortality. Our frontline partner Compaq was another ideal affiliate, and we discovered a great deal of interest in Hewlett-Packard (HP) as well.

  Encouraged by OEM feedback the marketing folks created a special NT workstation, market development agreement, including huge incentives for NT-specific promotional activities. We decided to measure its success on the sales ratios between Windows 95 and NT. We stayed away from asking for firm quarterly unit commitments to avoid any conflict with the existing consent decree. When executed to its full potential, this program could lead to a short-term 40–50 percent NT royalty reduction for participating and well performing OEMs. It took less than three quarters to reveal the desired trend. To our surprise, HP was the best performing program participant followed by Compaq. Dell hardly made any effort and achieved only small rewards, though they managed to complain the loudest about the structure of the campaign. By mid ’97, NT workstation sales resulted in nearly 10 percent of all Windows sales—a nice revenue boost for the company. We were on our way to set yet another standard.

  In parallel to the high-end workstation market, a new challenge appeared for my group with the appearance of Windows CE in ’96/’97. CE—for Consumer Electronics—was designed to capture the unique facets of the rapidly evolving electronic device opportunities. They had undergone meteoric changes over the last four years as semiconductor components became increasingly affordable, tightly integrated, and miniaturized. A new clientele was waiting to be conquered. The producers of industrial controllers was one of them, mobile phones another, with handheld devices and car-computer manufacturers completing the picture.

  CE intended to target them all. A functionally stripped-down cousin of Windows 95, it was designed to run in small memory footprints and on a variety of CPUs called ARM processors. ARM stood for Advanced RISC Machine, representing a low-cost CPU with reduced instruction sets and low power consumption. Combined with reduced memory prices, the appearance of cheap graphical displays—replacing text-based ones—and higher-integrated circuit boards, manufacturers were suddenly able to produce handheld devices with more computing power than the original IBM PC. Just fifteen years later!

  Compared to the other sales opportunities for CE, the industrial controller market was the most fragmented. It had its origins in replacing manual labor—first with reliable mechanical systems, then with electronic ones. Companies like Honeywell, Bosch, and Kontron come to mind when talking about key manufacturers. Beyond these larger companies, we encountered a huge array of smaller ones focused on highly specialized applications in particular in Germany. As early as the mid ’80s when digital technology had evolved enough to present fertile ground for new application opportunities, they all started to experiment. Unfortunately for us, the larger companies had meanwhile developed proprietary OS for their controllers—translating into an uphill battle for us. The smaller ones tried out Wind River software or DOS versions from DRI or us. Neither DRI’s nor ours were designed to address the needs of this environment adequately, yet they beat Wind River because they were less expensive. Windows CE was a much better fit. We got its feature set right in its second release and delivered a componentized OS, which industrial OEMs could pick apart and modularly adopt for a variety of applications.

  When it came to mobile phones, CE had intense competition from a company called PSION along with a parade of other OSs developed within the cell phone industry. In version 1.0, it lacked essential communication functionality, but when added later, CE eventually possessed enough compelling features to compete effectively. Handheld or pocket PCs were just emerging. The market was in an embryonic state, and Palm’s OS was our most riveting rival. Car computers were available already; they were expensive and mostly used for navigational aids that left us with no reason to assume such a narrow focus would not be expanded.

  In studying the opportunities for CE, I soon concluded that we needed people with di
fferent types of expertise for a successful run. Ideally, they would possess specific industry knowledge and act like design engineers. We delivered source code to our targeted customers when selling CE. In order to be successful, our new breed of sales reps needed sufficient tech knowledge about how to utilize its components for key industrial applications. Success further depended on the availability of a high-quality development kit. Without it, CE was rendered useless. Complicating matters further, its components had to be extremely stable and reliable to pass industrial strength muster.

  For the next two years, we experimented and struggled with selling what was a premature offspring of Windows. Until version 2.1 arrived, the software was not well-enough componentized or stable and its development kit not as capable as required—nothing new here. The other issue plaguing CE had to do with its initial performance. Never giving up, we overcame these obstacles. But to make matters worse, we had a tough time locating the right sales specialists, and we scrambled to select the right distribution partners to serve the smaller customers.

  As indicated above, our fiercest competition was often an in-house-developed OS. Our potential customers employed talented tech personnel intimately familiar with their own software. For them to change and use CE meant extensive retraining and often an internal turf war. Using our readily available system was judged as giving up hard-fought-for competitive advantages based on proprietary solutions. Trust seemed to be another factor not favoring us. With all the bad press and the antitrust saga raging on, companies had grown hesitant to buy core technology from us. They understood that we were not licensing a monolithic and unalterable OS as we did for PCs. Instead we were offered components for different devices and applications, which they could customize as they saw fit. Several prospects still remained reluctant to bet any part of their destiny on our solution—in particular the phone manufacturers. Growing the business was therefore harder than expected. Considered a breakthrough, one of the great success stories was an early deal we closed with my help with Bosch in Germany. It set a trend and made some of competitors follow her lead.

  As always, in good old MS style, we kept plugging away with unflinching perseverance and never gave up despite product-group management reshuffles, design changes, and notorious delays. Five years later, the CE business grew into a $150 million annual business with less than twenty dedicated sales engineers. I referred to it as a barely profitable hobby. We suffered from insufficient resources, not developing a consulting arm supporting customers sufficiently at the engineering level, and an underdeveloped distribution system. When I left my job in 2001, we had barely laid the foundation for a sustainable business. By then I had hired an industry expert in Philippe Swan, who applied a passionate dedicated effort, hired experienced people, revamped the distribution system, and managed to increase annual CE revenue more than threefold five years thereafter. A great accomplishment in itself after a slow start but, compared to MS’s core PC business, far less significant than I had originally expected and hoped for.

  RUN BY COMMITTEE

  In the summer of ’96, I relinquished my rest-of-the-world sales job and was full-time and happily back at the OEM wheel. I made senior VP that year. The current Office of the President was replaced by a newly created leadership group consisting of nine executives. I considered the arrangement a weird setup and never believed that a group of nine could effectively act as one. Bill, as the largest shareholder, with an amiable board on his side and Steve as collaborator, ultimately held the power and called all the shots. Committees at that level create confusion and nurture unwieldy politics. The reason for the new structure was explained to me by Steve as an opportunity to prepare the future leadership of the company for increased responsibilities. Why not send them to management training? I had a hard time investing much credibility in his explanation. He approached me after the announcement was already made. As we discussed the possibility of my participation with apparent frankness, I could not determine what Steve genuinely wanted. Was I too old or too controversial to be considered earlier? Perhaps he was presenting me the opportunity of self-determination. I made it easy for him. I just wanted to continue doing an excellent job in OEM.

  Sooner or later, Steve would run the company anyway. There was little room for anyone else at the top, especially considering Steve’s boisterously ambitious and turbo-bold nature. Why create false hope? Declining his invitation was one of the better decisions I ever made. Five years later, only two out of the nine selected execs were still with the company, including Steve. Preparing for which future?

  Bill’s announcement of a revised company direction that summer was again a result of his annual think week. This wasn’t the first time he had outlined something similar internally, though making his thoughts and conclusions public was a first. The new mission was to deliver a comprehensive set of products encompassing all communication and computing platforms. He wrote about seamless integration of the Internet with legacy systems such as mainframes and wide-area networks, which was probably the key reason why he still wanted to mend fences and team up with IBM again. His vision was to make the development of client-server applications easier—his indirect answer to Notes—and to implement a new navigation paradigm between computing environments. The message, loud and clear, was engineered to alert the outside world: MS was once again laying the foundation for broad-stroke global leadership. On the face of it, the propagated direction sounded fine, though for my taste it was not crisp enough and lacked finesse, an initial and troubling indication that the company was run by committee and nobody wanted to step on anyone else’s toes. An early sign that our leadership style was changing!

  As a result, the system-development organization got split into four main groups under Paul Maritz’s oversight. One group got focused on developing Internet Explorer. Another was working on the rest of Windows. The other two dealt with NT server and NT workstation technology. All were supposed to orchestrate together and implement the outlined strategy. For the two NT groups, things worked out well, while Windows gangs ran into trouble caused by a leadership crisis.

  Splitting up the Windows effort made sense considering that Internet technology then, still in its infancy, was moving along faster than the rest. The need for an ever-better browser to decisively challenge Netscape was obvious—meaning imperative. Tying IE releases solely to new Windows versions would have been competitive suicide. The IE group solidly reflected that burgeoning competitive zeal. On the surface, dividing the two worked well. Behind the scenes, things got messy. The IE group and the core Windows group soon engaged in turf wars unable to coordinate each other’s work. Most of the contention centered on the Windows desktop design. It had to be modified whenever a new version of IE was released. Who was responsible for changing the software IE needed to hook into? Unable to find common ground, the ensuing quarrels of their two leaders stretched even Paul Maritz’ rock-solid management capabilities. Both he—and Bill—had to intervene several times and mediate ego-driven disputes, though the rivalry raged on unabated. I refused to take sides and hoped the two prima donnas would bury their hatchets and turn their energies to timely deliverance of their creations.

  For an end-user, upgrading to a new version of IE felt that only the IE code itself was being changed. Therefore, many people believed IE and Windows could easily be separated. In reality—and invisible for the upgrader—our clever update technology always refreshed a host of related core Windows code in the process—the leading cause of the aforementioned conflict. Windows and IE were indeed joined at the hip as one.

  MS had therefore always shipped Windows and IE together, except for its first Windows 95 release, targeting the retail channel. For less software-knowledgeable people, this further strengthened the argument of IE and Windows being two independent and barely cobbled-together products. A total misconception. For that retail-specific variety, the development group had simply missed a production cutoff date—a minor hiccup. The Feds later drew the concl
usion that Windows and IE were in fact separable, and rashly and wrongly explored the incident mentioned to prove the tying claim Netscape had alleged all along—an erroneous judgment, with severe legal consequences to follow.

  COMPAQ GOING ASTRAY

  As explained earlier, OEMs, including Compaq, created extra revenue by charging for third-party-vendor icons placed on Windows’s desktop. In late May of ’96, Compaq decided to go one step further. The VP in charge of her consumer product line, Celeste Dunn, boldly eliminated the MSN and the IE icons from her consumer line of PCs, replacing them exclusively with the ones from AOL and Netscape. Both companies paid Compaq amply for their newly won exclusivity. We were shocked and immediately voiced three strong objections. As frontline partner, Compaq had promised to promote our solutions primarily and by removing our icons she reneged. Second, altering our Windows code violated her license agreement. Third, it violated our copyright. Compaq, as distributor for Windows, needed our consent for any and all modifications of the original!

  Celeste Dunn had simply presented us with a fait accompli. Upon confronting her, our account team leader was told she would look into it. Six months later, no effort had been made to remedy the situation, and what I considered bad news had not been escalated fast enough either. I heard about her blatant act by the end of that year and immediately contacted Compaq’s relationship manager. A Compaq old timer and a great guy, but in this case, he demonstrated zero influence. AOL’s and Netscape’s money continued to talk.