045. Incompatible Files, Slipping, Office 97 RTM
“SINOFSKY! SINOFSKY! WHAT DO YOU MEAN YOU ARE CHANGING THE FILE FORMATS?? WE CAN’T DO THAT!” –Steve Ballmer yelling across the building 17 cafeteria
Back to 044. Our First Big M&A Deal (Beating Netscape)
Please keep the feedback rolling in. This post concludes with shipping Office 97. It represents the end of the first era of the PC, where the focus was on features, retail consumers, tech enthusiasts, and mostly just getting stuff to work and shipping. The next couple of chapters represent a major shift in the PC as the focus turns to the enterprise business and the primary customer the business itself and professional IT.
Office96 was quickly becoming the biggest slip to an Office or Apps release in years, and that was extraordinarily disappointing. We reached our Zero Bug Bounce milestone (ZBB, because for a brief moment the product would bounce around a mythical zero bug count) in July 1996, and that was great. We were also behind our original schedule by months, and we knew there was no way to catch up. We needed more time. We felt like shit.
I did not always go to lunch in the building 17 cafeteria but happened to go one summer day. I was waiting to pay, holding my two slices of Marriott cheese pizza, when I heard, “SINOFSKY! SINOFSKY! WHAT DO YOU MEAN YOU ARE CHANGING THE FILE FORMATS?? WE CAN’T DO THAT!”
SteveB was yelling across the cafeteria at me, and all of a sudden, I felt like a few hundred people got quiet and were watching to see how I reacted. His voice seemed to come from everywhere, so it took me a moment to locate him. A 10-minute conversation followed that started in earnest as he made his way toward me about the reality that Office 97, as it had been officially named, had a whole new architecture with loads of new features for document creation like drawing and fancy Word tables, animations in PowerPoint, and new charts with top-notch graphics in Excel. Word and Excel changed the file formats to support these features. A changed file format meant a customer file such as FOO.DOC or FOO.XLS on disk was only compatible with the version of the app that created the file, or a newer version. Trying to open FOO.DOC created with Word 97 on a Word 6.0 setup would give an inscrutable error message. This message would often terrify the user into thinking they had a corrupt file or the computer had eaten all their work. This horrible experience was perfectly normal for state-of-the-art software. I kid you not.
Well, it turned out that the strategy of not changing the file formats in Office 95 (at least for Word and Excel) was a huge hit with the field sales force. So much so that they thought it was our new strategy to maintain the same file formats forever. When, in fact, we thought of 95 as the exception and only way to sync up with Windows 95—we had spent a ton of creative energy arriving at sellable features that did not require a file format change but were severely constrained in doing so.
I never really thought of mentioning this because it was so obvious, or so I thought. Steve thought we were totally screwed and had misled his sales team.
The file format changed probably within the first weeks of the project and was standard industry practice. The success of Office 95’s unchanged file formats caused a bit of a crisis as we were working on Office96. It was obviously both too late and impossible, regardless, to reverse course. HeikkiK coordinated what could best be described as a crisis mitigation plan involving format converters for existing applications and Save As capability within the new versions to produce down-level files (with the risk of losing new features). This was a reasonable solution but would require organizations to install software on every PC, whether it was just a converter for existing computers or the whole new Office 97 on new PCs. In other words what we thought was reasonable was a huge pain. SteveB was right.
The Apps culture was tuned to avoiding crisis moments, usually by rewarding prevention, and a low tolerance for avoidable drama. When a crisis did hit, self-inflicted or otherwise, the Apps approach aimed for cool, calm, collected. ChrisP often described this like listening to the cockpit recordings of Naval aviators landing jets in difficult circumstances. That calm was accompanied by an executive team that did not view a crisis as an opportunity to micro-manage or upend existing processes.
As a result, we thought this through in excruciating detail. All of the test managers and PMs from each team had developed a complete plan to support the new file formats in old products and deliver product updates to existing products, even using the novel approach of internet downloads. This did not solve the physics problem that we could not make the new features of Office96 show up in Office 95 or Office 4.3, but we could ease the pain and prevent people from being unable to open files they received due to the growing use of email.
One bit of good news was that the biggest motivator for changing the file format was to support the new world-wide standard for representing text in computers, called UNICODE, that Microsoft had contributed a huge amount to. Up until then, PCs had a difficult time representing many of the world’s languages and mixing several languages at once was almost impossible. UNICODE was a major step forward and for multinational corporations, the ones complaining loudly to SteveB, easing the pain of multilingual documents was a significant win. A decade later, UNICODE would become familiar to everyone as it supported Emoji.
This was a rare crisis for the team and a good learning experience to the degree that foisting incompatible files across a corporation could be called a learning experience. As an organization this did bring us into the modern, Office-focused world in two ways. First, the crisis created a new muscle, which was driving change to core behavior across all the products. That had to happen. It was the first of what could be called suite-first problem-solving and coordination.
Second, this was such a pain for every team that it served to be an early dose of the enterprise business and what would be different for the team. The needs of enterprise customers became paramount as we moved to future releases, and the solutions to those needs were addressed across the entire team.
Still, this was my first of many, many lessons in the difference between consumer products and enterprise products, between selling at retail and selling through account managers. Because every product always changed file formats, this should have been viewed as routine. But the world had changed—seemingly overnight documents were being emailed around, and all of a sudden it was all too common to get files that could not be opened. In a sense this was really a crisis for the Office business and one that would cause a sea change in how we viewed files. It also happened to be the strength of the internet and the World Wide Web—one single format that could be read by any browser, and soon one language that could be used to write a program once and run it anywhere.
Beyond that it was a lesson in how Microsoft’s customers were changing. The technology enthusiasts who embraced changes and had a high threshold for the pain that random changes caused were being replaced by system administrators and IT professionals who were not only change averse, but often barriers to change. The file format incident was a warning sign for what was to come, not just for Microsoft but for me personally as I had to reconcile innovation with a customer base that was becoming less interested in technology changes, which we thought of as cool innovation, and more focused on the costs of those changes, which we thought of as simply necessary.
Strategy changes don’t always come with a moment, but it was clearly this moment when it became clear we would move Office files to be open and to use the format of the internet, HTML. This would prove to be enormously controversial with BillG as the very origin of Apps, specifically Word, was the invention of the Word .DOC file format, which was a key, and proprietary, competitive advantage. It was also a technology Bill knew well.
By the Fall, Office96 was approaching the finish line. The team was tired. We were nine months late off our original schedule. We felt it. Nine months late was kind of inexcusable, but relative to the rest of Microsoft and the industry still well within norms. More than the feeling of exhaustion, we were feeling that as much as we accomplished, we had also not executed as hoped.
We were hard on ourselves for the slip. The routine post-mortem that would follow was filled with genuine discussions of trying to figure out how not to repeat this slip, even if it was also the last release we would talk about Office versus the Apps.
Even with exhaustion there was much to be proud of. The product cycle was difficult, but we did not experience any sort of death march. Yes, many people put in long hours. While many teams would do a “bug bash” one night in a week, we were not catering food every night as was common practice in other parts of Microsoft. Babies were born. People were married. There was all the life experience that happens within a group of 1,000 people over three years. Relative to the rest of the company, we ran the project with a sense of balance and normalcy. The core tenet of our process was that we arrived at the schedule and dates from bottom-up estimates, and when those estimates were wrong the individuals worked extra or we scaled back features—both of those were viewed as acceptable. We did not seek out heroes. We were proud of that. And we would get better at estimates, accountability, and balance each with release going forward.
The release to manufacturing, which came about six weeks before launch, was on a cold and rainy November 16, 1996. A few of us were at COMDEX in Las Vegas doing press when HeikkiK led the team through the final ship room meeting and signoff. He was kind enough to call in and I spoke from my flip phone but might as well have been in outer space—the connection was horrible. Still, the sense of accomplishment was as real as the relief. From afar I wished everyone well and knew there would be one heck of a party on campus. And there were pictures to prove it.
Marketing was gearing up for a launch and did a fantastic job creating awareness and a retail presence, and for the first time gearing up an ever-growing enterprise sales force. The industry had grown so much creating a massive demand for press coverage unlike any we had experienced. Every paper in every town in every country seemed to have a tech section. The summaries of press coverage were growing ever longer, and the skills the Office marketing had to get the word out in a consistent and clear way were growing with demand.
The launch was subdued relative to Windows 95, and that was by design. Late 1996 was mostly about the enterprise and servers so Office focused on those elements of the product.
While Office 95 was viewed as a significant release, the realities of a 200-page Office 97 reviewers guide detailing all the features of the product really sent the message that Microsoft was all-in on suites and had unparalleled depth and breadth to offer. The plethora of features in the product and breadth of tools kept the technology press busy and excited. The phrase shock-and-awe was routinely used to describe the overwhelming depth and breadth of the newest Office suite.
The reviews were extremely positive. Office 97 garnered nearly all the major awards across magazines, which at the time were a big deal. BYTE, PC World, PC Magazine, and more each recognized the suite and individual applications as editor’s choice or world class.
Clippy became the most controversial feature, even to this day, in productivity software. Even the reviews were controversial, with key mainstream reviewers citing both innovation and acknowledging the problem needed addressing. The late Steve Wildstrom, the widely read and deeply respected columnist at BusinessWeek, wrote, “I was deeply skeptical about these omnipresent artificial-intelligence devices. But to my surprise, I found the animated assistants useful—and a feature that sets the new Office apart from competing software suites.” Quite a few reviews were initially skeptical, and then using the product turned them around. For example, PC Week said, “Office Assistant exceeds my expectations: It's not only visually effective, it's also more than superficial in the help that it offers to even a veteran nerd.” These comments gave us hope, gave me hope.
Still, there was a love/hate relationship with our assistant friend. Clippy immediately became the stuff of legend when it came to the reception among the core technical customers in IT and tech enthusiasts—the kind of users that know all the keyboard shortcuts and don’t want anything to get “in the way” of their work.
I learned something in how this type of customer chooses to express distaste for a feature. Rather than directly say, “I do not like it” or “I will not use it,” these customers generally stepped up and claimed to speak broadly for “typical end-users,” often talking about their mothers or grandmothers (always female, to represent someone who is not fully versed in the product). In the case of Clippy, we would often hear that this is not the right feature to help “those users like my mother or grandmother needing help.” This projecting of product concerns rather than owning them directly would be a valuable lesson and something I would carry with me when bringing new features to market that must make it through these gatekeepers.
The negative reviews of Clippy were fairly brutal. I think everyone used the cuteness of Clippy as an invitation to spice up the language used to insult the feature. Stephen Manes in the New York Times wrote, “ . . .help is presented solely in the form of dialogue balloons attached to one of eight cartoon characters, most of which make irrelevant, distracting movements and sounds until you turn them off. . . . But these toon-zombies are as insistent on popping up again as Wile E. Coyote.”
That really hurt.
Most unfortunate, though, was how the feature could be construed as a symbol of Microsoft losing touch with customers, especially in this era of heightened scrutiny on the legal front. We were wrong, but not because we had lost touch—we were wrong because we went overboard (and in the wrong direction) trying to make computers easy to use for a much larger audience.
Still, we had pushed through a final design change to enable the assistant to be turned off. The change was easy technically, but difficult emotionally. Our goal was for Clippy to be an ever-present but unobtrusive assistant (Agent was the term at the time, today that would be Bot), thus turning Clippy off was an admission of failure. Making this possible was quite hard on the team given how far we had come from Bob through “tfc” and beyond. That is all the techie crowd needed, and with that most of the business deployments of Office 97 had no ever-present Assistant.
These reviews mattered. Reviews of products might seem old school in a world of instant hot takes in social networking and especially with products where bugs or errors can be fixed on the fly. In an era when software could not be so easily changed, publications assigned several people over the course of weeks or more to dig deep into a product, and our own marketing touted excerpts. These reviews really carried weight. We had a reviews team in marketing, 3 or more full-time people plus a reviews team at our PR agency working broadly with all the outlets, and this was repeated in most major subsidiaries, especially Japan, Korea, Germany, and UK. At the peak (over the next product cycle) there were easily 100 reviews being executed in the US. The marketing team was on a plane for two months meeting in person with all of these. I would visit dozens myself. There was a constant stream of communications fielding questions, dealing with bugs and compatibility, and offering support for demos and more. There was a growing parallel effort working with the rising importance of IT industry analysts. Analysts such as Gartner Group would soon eclipse the traditional tech reviews in importance to our business.
It was not without controversy that we focused so much on reviewers. This was an era where they could be characterized as a form of gatekeeper. It was no doubt limiting as any one reviewer represented a narrow perspective compared to all customers. There were no real alternatives. In many ways the internet would come to save us and enable a product as broad as Office to reach many more customers directly with more tailored messages, content, and calls to action. The changes in how we reached customers coincided with the maturing PC and PC industry.
Two reviews really stood out as defining, not defining the product in market as the product went on to great success but defining for me personally as a “product person” and also a leader. Stephen Manes, who had obliterated Clippy, was by most descriptions a curmudgeon and it would be difficult for me to say something he liked. We once debated whether a specific Sony laptop (the 505) was good compared to an Apple PowerBook in a session that went on for most of a press event at CES. His complete review, however, of Office 97 had a headline (keying off his description of the product) “An Upgraded Leviathan Sets Sail.”1 This was a brutal shot. It characterized the product as, in a sense, just another release while also describing it as, well, a leviathan. The text of the review is painful for me to read even today.
I will say for the record that one part of his review I often talked about was that he called out the addition of word and character count in the status bar of Word. While it surprises many professional writers, most people using Word never used the word count feature because most writing is not tracked by that metric. We did, however, add it to the status bar by default for the press who lived in Word. We have no shame about that. They were an important constituency. True to form, I think most reviews mentioned the feature.
While I hung that review on my Office door as I did many others, I also went to the office supply store and purchased a portfolio case—a plastic folder with a dozen or so clear sheet protectors in it. The first thing I put in the first page was the leviathan review. After that I put a few pages that I would carry around all the time and update as needed—the list of teams, senior managers, total headcount, ship schedule and milestones, data sheets on competitors, top support calls, and so on. It was this review and the next one below that I would talk about and hold up at team meetings when describing our challenges as a business, especially as we planned future products.
Much more problematic was Outlook’s reception as it was not a single feature to turn off after making fun of it—but rather a marquee addition to the suite, a new “puzzle piece” (the Office logo was a puzzle with each app taking on a different color). The absence of support for internet standards as well as the overall complexity of a version 1.0 product led to a series of fairly brutal reviews. Perhaps the one that stung the most came from the Wall Street Journal with the title “Microsoft Introduces Personal Organizer That’s Unorganized” and zingers such as:
The combination is so tempting that Microsoft incorporated Outlook in its showcase $190 Office 97 suite of software, and rejected the commonly used PIM [personal information manager] label for it, calling Outlook a “desktop information manager,” or DIM. Sadly, that unfortunate acronym is apt. Outlook 97 doesn’t live up to its potential. It’s a great idea, poorly executed.
What was an internal joke between the now rival Exchange team and the Outlook team was a review. The review concluded, “Microsoft has a history of doing a poor job on the first version of new products, and Outlook fits that pattern.”
The reporter, Walt Mossberg, had a broadly read and widely discussed column, Personal Technology, aimed at taking on the techies of the world and makers of overly complex products. He was the clear leader in the point of view that, in his words, “Personal computers are just too hard to use, and it isn’t your fault.” He would challenge us, consistently, repeatedly, and objectively. He played no favorites and was perfectly straight in dealing with us and any vendor. There was no way to exert undue influence over him, no special treatment, nothing. We showed him the product, answered his questions, tried to fix any issues, and waited for the review, then he returned the product and loaner laptop to us promptly.
Walt’s contribution to encouraging or pushing Office do a better job is something I have shared with him many times. I have enormous respect for him and what he accomplished in his column (and later the conference he cocreated and then a media and technology publication site). His keen reviews profoundly influenced buyers and makers alike.
The complexity of Outlook and Walt’s experience ironically deepened our ties. Walt began sending me questions and comments from his readers and I would often answer the myriad “how-to” or “why” questions he received. I never resented or tired of receiving these, and later felt nostalgic about them.
Walt was right about Outlook, and we had a lot of work to do.
I can’t exactly compare what our team went through to a battle or a traumatic experience, but we did have a shared experience that changed the team dynamic. It also came at the right time as the industry was changing dramatically right before our eyes. Teams are built by going through a journey together. When I reflect on Office 97 I am convinced that the journey to build the product is what created the team and culture that have been so enduring, even today.
Office 97 was the last release to be sold primarily to individuals at retail. It was the last release to be marketed significantly as app features. And it was also the last release to be built as independent applications with shared code versus a shared strategy.
PCs were the new onramp to the information superhighway. The internet had become a global phenomenon driving PCs into every home, as Bill and Paul had envisioned more than 15 years earlier. No longer did demand need to be created for PCs; it needed to be met. For Office, all those PCs getting on the internet were being used for schoolwork, homework, and work at home using Office.
PCs were also standard on every desktop in the business world, as Bill and Paul had hoped. The race was on to equip workers with all kinds with PCs, to get them email, and to provide them with the tools needed for the creation and dissemination of knowledge. Office was a standard part of this.
We needed to scale our product development approach to meet these needs and scale.
In terms of PCs, dollars, documents, and customers, millions would turn to hundreds of millions and hundreds of millions to billions.
We were at the “end of the beginning” of the PC Revolution—the first part of the journey dominated by hobbyists, tech enthusiasts, and early adopters. The PC was now front and center in a revolution taking place in business and Office was destined to be a foundational element of the computerization of work. Over the next twelve months, PCs would sell more than 100 million units and more than half of those were bought directly by businesses. The PC was no longer a hobby or a luxury, but an essential element of business.
For me personally, Office 97 marked both an end and a new beginning.
On to 046. Prioritizing a New Type of Customer [Ch. VIII]
leviathan. [ li-vahy-uh-thuhn ] 1. (often initial capital letter) a sea monster. 2. any huge marine animal, as the whale. 3. anything of immense size and power, as a huge, oceangoing ship.
Steven, since you brought up Unicode in this post, it’s a good place to mention another important innovation in Windows from the hero of an earlier chapter.
For background, the tools and platform teams had come up with a series of C macros and #ifdefs that would allow a developer to create either a “ANSI” or “Unicode” Win32 EXE from his source code. At build time, the developer would pick whether his strings for Windows API calls would be ANSI or Unicode, and the appropriate binary would be built. It was a little clunky in some ways, but it mostly worked, with a “gotcha.”
The “gotcha” was that Microsoft had two Win32 operating systems, and only one of them would run Unicode binaries. And the one that ran Unicode binaries wasn’t the one that had been flying off retail shelves, the one bundled with most new PC’s, the one advertised on TV with a Rolling Stones tune. No, Unicode may have been the way of the future, but Office needed to make money selling to customers in the present.
Had Office 97 required Windows NT to run, no one would have been complaining about new file formats.
The rest of the story came to me through the grapevine.
The TrueType fonts were already Unicode, but all the Win95 API’s were ANSI, and the Chicago Master Plan said they were all going to stay that way.
I heard that someone in Systems decided to do a little skunkworks project in Win95 GDI, adding Unicode versions of just the API’s needed to display and print Unicode characters and to deliver font metrics to applications. He did this without a getting much buy-in from the management chain. He did it without getting SDET’s to write test suites to verify that all the edge and corner cases worked correctly, so there were a few bugs. (I recall one involving East Asian characters.) But Win95 shipped with enough Unicode support in it for an application to compose, display and print Unicode text. Mind you, these API’s were not undocumented. Any ISV was able to use them. But at the time, only Office 97 did.
Office 97 was a very attractive product in the EU because it could produce documents that contained English, French, German, Spanish AND Greek. You could do Western European languages without Unicode, but Greek was a different character set. And Greece was an EU member.
The way I heard the story of the skunkworks project, the guy who went into Win95 GDI and added those Unicode API’s “under the cover of darkness” was David Weise. (I later worked in the same group with David, but I never thought to ask him about it.)
Some short while after we'd shipped O97, I ran into Mr. Sinofsky in the break room, and he semi-guiltily explained how an unexpected thing had happened, regarding the file format change. In companies all over the world, a corporate exec would get sent an O97 docx, and couldn't open it. Instead of demanding his people use SaveAs for doc, rather he'd yell at his IT team that he, senior veep, ought to have the latest version dammit. And that, of course, would translate to buying the upgrade for the entire corporation. Profits ensued.