068. The XP eXPerience

The last months of a product development cycle the scale and length of Office10 are moments of calm punctuated by moments of terror. The calm comes from the lack of code changes as thousands of people test the product while hoping not to find anything requiring code changes. The terror comes from issues that arise when you have little time to change them because we’re still dealing with the physical world. Something that seems so trivial like a product name requires a month of more of work to get right, then rolled out around the world. Office10 did not have a name just three months before we would finish. Office10 was also one of the first product releases for Microsoft where when the product made it into customer’s hands there also needed to be live and working web sites that extended the product, for over one hundred million people on day one. Other moments of terror come just after the product ships and those first reviews—it is not just if they are good or bad, but will there be a recall class issue, be it a design choice or product defect. Such an issue could tarnish the entire product cycle. What if our marquee feature—the one in all the printed brochures and print ads—ends up part of a scandal rapidly spreading to Windows XP?

This post brings you inside the last months of a project that was the first Office product to ever finish on time, and perhaps one of the first Microsoft products ever to do so. As it would turn out 2001 might be viewed as peak execution for the middle-age of the PC journey. The Windows release under development which would become Windows XP would also ship on time, a first for Windows. It makes sense that the major products would learn to ship on time just as the world was moving on to the internet.

Back to 067: MYR-CDG: Product Meets Sales

In early 2000, branding and naming the next release of Windows was all gummed up and that slowed down Office, which needed a name in order to release on time in March. Whistler, the code name for Windows, was scheduled to ship about six months after Office10. An August ship date for Windows was achievable, and no one wanted to mess that up. The name, perhaps to the surprise of many readers today, was a long pole in the release as it touched code in many places and could impact localization, manufacturing, printing, and packaging. We needed a name. How hard could that be?

Windows trying to settle on some variant of “experience” because the main thrust of the release was to finally bring the Windows 95 experience of peripherals and consumer software to the enterprise-focused NT code base of Windows 2000. In addition, we collectively concluded the year names were not working (hence Windows Me) because of the difficulty of keeping ship schedules and customer confusion over what products worked with each other, both issues were entirely predictable and frequently discussed at the time. The branding team was thinking of something like Windows EXP or Windows XPR, and finally Windows XP. My concern in an endless email chain was not this product but the next one. Would the next name have a roman numeral, go back to adding version numbers, a superscript, or add descriptor Edition branding? At one point, someone mocked up Windows XP ² which made me wonder if I was being made fun of, especially in the Spanish release.

The corporate teams settled on and cleared the name Windows XP, and what immediately followed was the question as to whether Office10 should be named be Office XP. The corporate branding team and sales teams were in favor. The Windows team, however, did not intend for other products to add an XP suffix. This seemed to be another .NET in the making, which was still nowhere close to being resolved. The branding powers demanded evidence in the product that supported co-branding.

We held many meetings during development about aligning the releases of Whistler and Office10, but the products were on wildly different schedules. From our perspective, Whistler was about finishing Windows 2000 without much new for Office. Besides, Office was going to run perfectly well on the new corporate desktop of Windows 2000 anyway. Windows 2000 had been years late, so it was not unreasonable early in the process to doubt the next Windows schedule. Whistler was making a ton of progress and we were self-hosted, so any worst thoughts were not going to happen. This was a first for Windows! Still, there was no significant release synergy. Office ran on Windows XP, and realistically not even as well as it ran on Windows 2000 PCs with the same amount of memory.

At one of these meetings the topic of naming them the same became heated, if not awkward. BillG wanted to know what was interesting in Office10 and what was unique when it ran on Whistler. Baked into this assault was the belief that Office was not exciting on its own and more importantly that Office failed at exploiting the latest Windows release. This was the view that came from the perspective that innovation from Microsoft emerged first from Windows.

My answers were not satisfying as I restated the series of meetings, lack of APIs, and our system requirements. I chose to highlight the work we did to align with Windows Server for SharePoint, totally irrelevant to the conversation in this context. BillG expressed frustration at “another release of Office that relies on Windows to make it exciting,” which honestly didn’t make much sense and bordered on insulting. By and large we were not looking to the Windows desktop to innovate for Office. In reality, many innovations in user-interface flowed from Office to Windows for years already. That wasn’t interesting to Bill in this moment.

Windows really wanted the XP name to be unique to Windows. Office really didn’t want to look like it required or was matched to Windows XP. We learned that lesson when Office 97 arrived and many consumers at retail were wondering where the matching Windows 97 was or even if Office 97 ran on Windows 95. After more time back and forth, Office marketing and corporate branding agreed to, no surprise, a compromise of a name—Office XP with a “Version 2002” prominently displayed on the box. The apps were called Word 2002, Excel 2002, etc., not Word XP or Excel XP because branding didn’t want to overuse the XP moniker. Windows did not have such a version on the box or in the software. Also, the boxes didn’t look at all similar. This meant customers calling product support or using the web site would need to search for “2002” and not XP, unless they used Google which got it right. I can’t make this stuff up, but just wait until the next release of Office.

We had a name. We were on track for March 2 and we had a date for launch (and boxes for Germany). We signed off on 3/2/01 just like we planned. It was magical to do that. We beat the Excel record from more than a dozen years ago and hit our planned ship date. I know this sounds ridiculous—product development schedules are supposed to work, but it simply did not happen with software projects. We were so happy.

For launch, marketing planned one main US event, which we expected to be covered globally. BillG headlined the event at New York City’s Manhattan Center Ballroom aimed at getting broad media coverage, while around the country and world there were hundreds of local events targeting enterprise IT professionals. There was no effort to get people lined up at stores or any sort of midnight madness, though local offices around the world did some of that and we did have BillG greet the early buyers in New York. A fixed launch date is a great forcing function (there’s that phrase again) to get everything ready for press tours, reviewer workshops, and enterprise product information.

For enterprise customers the main features of the release going back to the original product plan were collaboration and integration with the just-shipped Exchange 2000. SharePoint Portal Server 2002 (there’s that naming again) released a few weeks after Office XP which included SharePoint Team Services and also served as a collaboration platform. Our collaboration story was as complex as predicted in our vision statement, a result of targeting the same scenarios to two different back-end infrastructure products.

For system administrators, we enabled new scenarios, such as installing Office XP from a website and ever-more controls and customizations for deployment. There was a lot there and it was all new for enterprise customers. Industry analysts were having a great time digging into the idea of Office shipping servers.

Through some incredible outreach efforts by marketing and the field, more than 500,000 enterprise customers used the product in pre-release. The internet made it easy to distribute the product, and because of Watson we not only knew (anonymously) that people were using the product, but we were fixing bugs based on their usage and knew the release was high quality. We really knew that, not just hoped. Shipping software had radically changed using Watson. The availability of data forever changed what we worried about when shipping.

These are the Office XP consumer data sheets. These were used at retail outlets, tradeshows, and by sales people making calls on potential resellers.

The traditional tech press and mainstream media continued to struggle with explaining, or making broadly interesting, heavy enterprise features like collaboration. I was still smarting from the reviews of Office 2000 and was determined for us to get credit for the personal productivity features. We built a great set of capabilities that worked with or without servers, and as was common at the time with or without a broadband internet connection, though by that point most every customer was connected.

Office XP introduced several novel features we believed gained notice including the first features that worked seamlessly using an internet connection from within the apps. Building such features was a fascinating lesson in team transformation. With many stories from the early days of the internet about traditional print-based offerings unable to transition to the web, we set out to do something new and innovative with the thousands of pages of training materials and the vast library of content we shipped on CDROM. Jeff Olund (JeffO) began his career in “user education”, creating the written materials that accompany a product. He became a leader in building reference and training materials for Office and then led the worldwide localization team based in Dublin, Ireland. With JeffO’s leadership, content went from a cost center to an asset for the business. The Dublin team went from taking almost a year to localize Office into a dozen languages to localizing Office into 100 different languages in just a month or so.

The next step in content to help customers was to use the internet to provide endlessly growing features such as adding images or clip art to a document, templates, how-to, and more. As an example, before Google pioneered image search, it was luck that enabled an author to find suitable clip art. Especially in large corporations, most of the clip art we shipped with the product was eliminated to save disk space. In Office XP we introduced an Insert ClipArt feature that used an ever-expanding and easily searchable collection of internet images much larger than anything on a CD. The idea of tens of thousands of images available for free was quite cool for business users of PowerPoint tired of angry person and idea lightbulb over head person.

In hindsight, a product having a website seems obviously trivial. But at the time, we were deeply concerned about adding a website to a product used (and liked) by hundreds of millions of people. The web was still flakey (and slow) and while novel, sites routinely did not work. We did so much work in quality, the thought of having toolbar buttons or menu commands leading to strange errors from unavailable sites was horrifying. Websites needed to be new and fresh all the time, and always work.

The help system was no longer limited to what was written and shipped with the product but could also search a large and growing library of how-to articles on the web, including an under-the-radar hit called Crabby Office Lady who offered tips with an irreverent tone (“advice with attitude”), authored by a professional writer on the team Annik Stahl (AnnikS). Annik’s column was wildly popular with hundreds of millions of views. She was interviewed by local newspaper tech columns (often writing similar content) and even had television appearances (as herself, not the character). The character and personality Annik created even drew some inquiries from the VP of Human Resources over concerns of stereotyping. Annik’s goal was to take aim at the perceptions of the Office product, not the Office customer and she had full control over the project she initiated. We saw this approach used to great effect with a series of books such as Word Annoyances, that had become one of the most popular books on using Office products. So popular was the Crabby column on the site that Annik also went on to write a book and was part of several behind the scenes stories.

We created a new online services team under JeffO’s leadership. AndrewK moved from OPU to lead program management to help focus on content working for JeffO, who reported to me. Seasoned managers Mike Kelly (MikeKell) and Randall Boseman (RandallB) had to create new processes and tools to go from zero to a one hundred million web visitors seemingly overnight. The Content team developed schedules for creating, localizing, and releasing content at regular intervals, something we had only previously done on multi-year release boundaries.

Between online content, Crabby Office Lady, and the introduction of a new sidepane user-interface, plus the ongoing ridicule aimed at the Assistant we also made some big changes to Clippy. The plan for the product cycle was to provide even more options and administrator controls to reduce Clippy’s visibility, including making sure that by default Clippy no longer appeared, though it could be summoned if desired. Importantly, the message for Office XP was that it was so easy to use that “ . . . Clippy is no longer necessary, or useful.” That might have been spin, perhaps.

Lisa Gurry (LisaGu) in marketing thought up a clever idea to make the most of this change, and to embrace the opportunity to be self-deprecating in the process. She planned a formal retirement for Clippy. Instead of tacking the feature change on to the press tour in April, she planned a web-based celebration. Gilbert Gottfried, comedian and voice of dozens of animated characters, was enlisted in a set of internet Flash videos (these were all the rage then) of Clippy trying to insert himself into the daily lives of people. On retirement day, Lisa issued a press release and dressed another member of the team in a giant Clippy suit as we strolled Union Square in San Francisco, complete with a cable-car ride.

While this was the least corporate approach to PR, it set a high-water mark for intentionally cool viral marketing. To this day, the “retirement” still gets a laugh.

Laughs, however, were not part of the marquee feature, known as Smart Tags. Smart Tags were a set of buttons shared across Office applications appearing when and where the user needed them (such as when a user made an error in an Excel formula, when Word automatically corrected a user’s action, or when a user pasted some data), providing options to adjust the chosen action or fix an error. Smart Tags appeared when pasting text into Word, giving a choice of whether to match formatting or not, a common need with the rise of taking text from web pages. Smart Tags made it easy to undo (or never do again) when Office autocorrected something incorrectly, like convert a row of dashes to a typeset line.

While there were many different features across the product, Smart Tags provided a single and consistent unified interface. From a marketing perspective, Smart Tags felt like toolbars in their ability to be a visual symbol of innovation in the release. The screen shot was used all over the place.

It was Office at its best. We thought.

In press tours and reviews, Smart Tags proved a novel interface approach and were demonstrably innovative. Surprisingly, it took us several years to achieve—this idea was in the works going back to the original Office interoperability work in 1994, but before we could execute new ideas we needed to clean up the old implementations of menus and toolbars, which we did.

Smart Tags were also extensible. A corporation could recognize an order number or a shipping code and make it easy to link directly to those websites or systems. Browsers, especially Internet Explorer, and tools like free email programs or websites, were not yet universally inserting automatic links by recognizing the text of http://, phone numbers, dates, or other common strings. The most well-known was the use of 1Z as a prefix for a United Parcel Service tracking code (tracking was a new thing for consumers), which with a Smart Tag enabled that text to act like a link to UPS, if Internet Explorer or Outlook used Smart Tags extensibility. Phone numbers, addresses, dates, and more were all candidates for Smart Tag actions. We computed the time saved by the use of Smart Tags to be an insane number of millions of hours—pure marketing.

We thought this such a clever idea and an opportunity to show off synergy with Internet Explorer that we created a Smart Tag add-in for IE that recognized many common potential links. It seemed useful and synergistic and was included in the Internet Explorer 6 beta test. The IE team thought this was clever too—this was before browsers could be customized with add-ins and web pages were mostly static HTML (except for the occasional blinking or marquee text).

It wasn’t long before JimAll (the leader of Windows) called to tell me that people were freaking out. By this he meant that the press tour for IE was not going well. Concerns about Smart Tags were expressed within the context of IE. While there was some hyperbole, the concerns boiled down to feedback expressed stridently by Walt Mossberg, who wrote in the Wall Street Journal, “In effect, Microsoft will be able, through the browser, to re-edit anybody’s site, without the owner’s knowledge or permission, in a way that tempts users to leave and go to a Microsoft-chosen site—whether or not that site offers better information.” The terminology isn’t aging well, but the idea was simply that links would appear on HTML pages that were not authored by the site owner. This in effect could be viewed as editing the site, where editing means adding new links to a page authored by a third party.

This notion of “re-edit” went beyond what we considered or even thought of the feature. There’s some irony in that today browsers and mail clients (and many products) do this for a whole host of special strings and even Apple computers provide reference lookup to Apple sources. At the time, it was routinely pointed out that the climate of 2001 was filled with concerns that Microsoft might leverage one part of Microsoft to benefit another poorly performing or just poor part of Microsoft. Smart Tags in IE surfaced those concerns. Perhaps exactly the right feature, at exactly the wrong time, from exactly the wrong company. Conspiracy theories felt real to many perfectly rational people in the industry.

To make that point, Mossberg concluded, “Microsoft’s Internet Explorer Smart Tags are something new and dangerous. They mean that the company that controls the Web browser is using that power to alter others’ Web sites to its own advantage. Microsoft has a perfect right to sell services. But by using its dominant software to do so, it will be tilting the playing field and threatening editorial integrity.”

We did not think through the potential abuse of the feature, though even in the worst case it was not precisely what was suggested by some. Sites were not being rewritten and links were not getting replaced. Neither was it benign nor free from potential exploitation. We did not consider how someone with bad intentions might develop a Smart Tag add-in that could at best be annoying and at worst recognize text and offer a link to a nefarious web site. As a result, IE removed the feature, which spared us the challenges. It appeared as capitulation, and the press was not shy about expressing that. As a broader point, the company, particularly under the new leadership in our corporate legal group, preferred and encouraged capitulation, especially for anything that might catch the attention of its constituents, the regulators.

Office XP and the press materials still had this marquee feature, using the name Smart Tag everywhere for consistency. Was the name tarnished, though? There wasn’t anything we could do other than downplay the name and issue Q&A and FAQ documents to the marketing teams around the world. This was an unfortunate side-effect of not having thought through the implications in the browser.

The Smart Tag incident took place at what was perceived to be the height of Microsoft’s power and potentially negative influence on the industry. The fact that we relatively quickly backed down has been viewed as a milestone moment. The incident was even portrayed in a widely read profile of Walt in WIRED, May 2004, titled “Kingmaker.” It was an example of “dozens of companies that have redesigned products in response to Walt’s unsparing criticism.” Walt was right. I viewed this incident as the system working—the reviewers acting as a check on poor product choices.

Finally, we always reminded people we made the product more stable each release. With Office XP, however, this was provable. We equipped the press and field with statistics and descriptions of the Watson curves and buckets so they could understand our new approach to improving product quality. Watson was industry-leading and pioneering. Every time I represented this work I am sure I beamed with pride, even though showing it off meant using a tool we wrote for demos that forced the product to crash (again).

The reviews of Office XP generally reflected the positives of the product for both consumers and businesses. The industry was going through a post-dotcom consolidation and, while tech was still enormously popular, the slow but certain decline of print journalism (and the dotcom bubble) affected reviews. First, the budgets in time, people, and hardware for the kind of testing we used to see at Byte and PC Magazine were reduced or gone. This led to a decline in reviews based on usage of new parts of Office, especially the hard-to-test parts like SharePoint. Second, the web itself favored shorter and faster takes on what was new. People wanted reviews at the time of release. While we gave ample lead-time and supported embargoes, other pressing demands made it difficult to spend time in research mode prior to a product release. With the PC deep in middle age, in-depth reviews of technology products were replaced by more instant analysis of meta topics, such as the impact of the product to the company or a take on whether the product should be deployed or not. (Hint, they always said to stick with a current version unless that was old.)

In that context, Office XP provided a glimpse of what was in store for the business. Every review evaluated the product in the context of the value of the upgrade. The assumption for Office was that everyone with a PC owned it and, while this was far from true, it might have been true among those reading tech press with their up-to-date PCs at work, but not all PCs by any measure. Enterprise customers already owned Office XP and were by all accounts only upgrading with a PC refresh cycle. Reviews almost always included benign comments about the product from IT managers along with a quote about not being clear if there was enough value to upgrade at the time, but certainly the future looked good.

The decline of awards like Editor’s Choice and in-depth 10-page reviews with benchmarks was, honestly, sad. It was as though the world cared less about what we did. The internet was the new focus for consumers and businesses. It was no longer the PC and new desktop software.

The enterprise field fully engaged on an XP desktop motion, meaning Windows XP plus Office XP, even with Windows months from finishing (as an aside, Windows XP was the first Windows release to have a plan and schedule that remained steady and the product finished on plan late summer). Office XP and Windows XP were better together for the enterprise. That reduced the sales surface area for the field from two products to one, so to speak, which was greatly preferred. In that sense, the timing of the two products together reinforced a core belief of BillG’s, that new applications provided excitement for a new OS, which created demand for new applications—a virtuous cycle.

Our glitzy launch event in New York at the end of May 2001 featured BillG leading an enterprise-focused event but with enough consumer demonstrations to capture headlines. The focus on productivity was illustrated by a broad though relatively unsupported statement, that by “making Office just 10 percent better we can save hundreds of millions of man-hours.”

We were joined by some special guests that day. Chief among them was the founder and CEO of a relatively new Seattle-based company called Amazon.com. Jeff Bezos joined BillG on stage to demonstrate searching for Office XP on Amazon and buying it with Amazon’s new One-Click ordering feature. A box even arrived on stage for Jeff and BillG to open.

But my mind was already deep in navigating what should come next for the product and the team.

On to 069: Mega-Scale, Mega-Complexity (Chapter X)

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit hardcoresoftware.learningbyshipping.com