H. The Success of Microsoft's Effort to Maximize Internet Explorer's Usage Share at Navigator's Expense
358. Microsoft's efforts to maximize Internet Explorer's share of browser usage at Navigator's expense have done just that. The period since 1996 has witnessed a large increase in the usage of Microsoft's browsing technologies and a concomitant decline in Navigator's share. This reversal of fortune might not have occurred had Microsoft not improved the quality of Internet Explorer, and some part of the reversal is undoubtedly attributable to Microsoft's decision to distribute Internet Explorer with Windows at no additional charge. The relative shares would not have changed nearly as much as they did, however, had Microsoft not devoted its monopoly power and monopoly profits to precisely that end.
359. A developer of network-centric applications wants as many consumers as possible to acquire and use its products. It knows that only consumers running a browser that exposes the requisite APIs will be able to use network-centric applications that rely on those APIs. So in deciding whether to concentrate its development work on APIs exposed by Netscape's Web browsing software or Microsoft's, one of the questions a developer will ask is how much Navigator is being used in relation to Internet Explorer. Dividing the total usage of each browser product by the total usage of all browsing software (i.e., usage of the installed base) answers this question, for it reveals the proportion of total usage accounted for by each product. The relative attractiveness to developers of Navigator and Internet Explorer thus depends to a large extent on their relative shares of all browser usage.
360. According to estimates that Microsoft executives cited to support their testimony in this trial, and those on which Microsoft relied in the course of its business planning, the shares of all browser usage enjoyed by Navigator and Internet Explorer changed dramatically in favor of Internet Explorer after Microsoft began its campaign to protect the applications barrier to entry. These estimates show that Navigator's share fell from above eighty percent in January 1996 to fifty-five percent in November 1997, and that Internet Explorer's share rose from around five percent to thirty-six percent over the same period. In April 1998, Microsoft relied on measurements for internal planning purposes that placed Internet Explorer's share of all browser usage above forty-five percent. These figures are broadly consistent with ones AOL relied on in evaluating its acquisition of Netscape: AOL determined that Navigator's share had fallen from around eighty percent at the end of 1996 to the "mid 50% range" in July 1998 and that Internet Explorer's share had climbed to between forty-five and fifty percent of the domestic market by late 1998.
361. Before a developer sinks costs into writing applications that rely on APIs exposed by Navigator or Internet Explorer, the developer will also want to know what share of browser usage each of the competing platforms will enjoy in the future, when the developer's applications will reach the marketplace, and even farther into the future, when the developer will try to sell updated versions of those applications. Dividing the new usage of each browser product by the new usage of all browsing software (i.e., incremental usage) helps to formulate a prediction. If a browser product's current share of all browser usage is fifty percent, and its share of incremental browser usage is thirty percent, the product's share of all browser usage will, assuming the share of incremental usage does not rise, gradually approach thirty percent, as the size of the population of browser users grows and current users update their PC systems. So Navigator's and Internet Explorer's relative attractiveness as platforms also depends greatly on their relative shares of incremental browser usage. Microsoft's tactics were focused on channels for the distribution of new browsing software. Moreover, excluding the installed base from the calculation heightens the sensitivity with which share of incremental browser usage reacts to contemporaneous forces. Microsoft was thus particularly interested in share of incremental browser usage, not only as an indication of Navigator's and Internet Explorer's relative attractiveness as platforms, but also as a sensitive reading of the impact that its actions were having.
362. According to data on which Microsoft relied in the course of its business, Internet Explorer was, by late 1997, capturing a larger share of incremental browser usage than Navigator. Specifically, data that the company then deemed reliable showed that fifty-seven percent of the new users of browsing software in the last six months of 1997 used Internet Explorer, while only thirty-nine percent used Navigator. By February 1998, Microsoft's data showed that sixty-two percent of the new Internet connections over the previous six months were using Internet Explorer, versus thirty-eight percent for Navigator. Since there is no indication that Navigator users as a group employ their browsers more than Internet Explorer users, these data indicate that Internet Explorer's share of incremental usage had exceeded Navigator's by late 1997. This meant that Internet Explorer's share of all browser usage was moving to surpass Navigator's. To Microsoft, these numbers not only marked a significant decline in Navigator's attractiveness as a platform, they also reflected the substantial impact of Microsoft's actions.
363. The "hit" data collected by AdKnowledge comport with the share estimates on which Microsoft and AOL relied internally. AdKnowledge is a company that markets Web advertising services. Once the proprietor of a Web site sells space on its pages to an advertiser, AdKnowledge stores the advertisements on its servers and delivers them to the appropriate pages when they are accessed by users. One day every month, AdKnowledge monitors the number of times that each of the advertisements appears on users' screens. Each appearance of an advertisement on a user's screen is called a "hit." As part of the hit data it collects, AdKnowledge logs the type of Web browsing software used to access the pages on which the particular advertisements appear. Thus, the AdKnowledge data can be used to calculate monthly snapshots of the shares of usage that particular types of Web browsing software attract from the population of users accessing the Web pages that AdKnowledge monitors. To the extent AdKnowledge can detect the IAPs through which individual users access the monitored sites, the data can also be used to calculate estimates of the usage shares that particular types of browsing software attract from the subscriber bases of particular IAPs.
364. The AdKnowledge data show that Internet Explorer's share of hits to the monitored Web sites rose from twenty percent in January 1997 to forty-nine percent in August 1998 and that Navigator's share fell from seventy-seven to forty-eight percent over the same period. Dividing the change in the respective numbers of Internet Explorer and Navigator hits from the first quarter of 1998 to the third quarter of 1998 by the change in the number of total hits over that same period yields a fifty-seven percent share of incremental browser usage for Internet Explorer and a forty percent share for Navigator. These figures are again consistent with the estimates on which Microsoft and AOL relied internally.
365. When a user accessing the Internet through AOL moves from one Web page to another, AOL temporarily stores, or "caches," the first Web page on a local server. When the subscriber seeks to return to the first page, AOL delivers it from the local server rather than returning to the Web for a refreshed version of the page. AdKnowledge only counts a hit when one of the monitored advertisements is served to a users' computer from the Web. Thus, AdKnowledge undercounts hits by AOL users. AdKnowledge's attempt to implement "cache-fooling" measures has not eliminated the effects of caching. Largely as a result of the restrictive terms Microsoft prevailed upon AOL to accept, Internet Explorer enjoys a very high share of browser usage by AOL subscribers. Consequently, Internet Explorer's share of all hits detected by AdKnowledge is lower than its actual share of all usage. Correcting for the effects of caching results in virtually no change to the AdKnowledge-based calculation of relative browser usage shares in early 1997; however, it raises by approximately five percent the figure representing Internet Explorer's share of browser usage in the third quarter of 1998.
366. Although AdKnowledge only monitors hits to commercial Web pages, there is no indication that certain types of Web browsing software are used more than others to access commercial, versus non-commercial Web sites. Furthermore, the same share trends reflected in the AdKnowledge data appear in data collected from a prominent academic site. The University of Illinois at Urbana-Champlain monitors, on a weekly basis, the browsing software accessing its popular engineering Web site. The resulting data, which AOL found important enough to rely on in evaluating the purchase of Netscape, yield virtually the same usage share figures as do the AdKnowledge data.
367. AdKnowledge does not undertake to collect data on the use of browsing software to navigate proprietary OLS content or intra-enterprise networks ("intranets"). This does not detract from the value of the AdKnowledge data as a measure of usage share for developers' purposes, however, for most developers of network-centric applications look to write applications that will run through Web sites, not through OLS proprietary content or pages on an intranet. Most developers will therefore pay most attention to estimates of the extent to which a particular type of browsing software is being used to browse the Web. Moreover, only a very small percentage of the copies of Web browsing software in operation are used exclusively to navigate intranets.
368. The advertisement banners on some Web sites alternate between different advertisements. Assuming that AdKnowledge delivers these advertisements, a single visit to a Web site could register with AdKnowledge as multiple hits as the advertisements "rotate" on the user's screen. This phenomenon does not spoil the essential reliability of the AdKnlowledge data as a reporter of browser usage share, though. In order for there to be a bias of significant proportions, users of either Internet Explorer or Navigator would have to exhibit a special propensity to keep pages open as the advertisements rotate. There is no reason to believe that this is the case.
369. Thus none of the characteristics of the AdKnowledge data invalidate it as a useful measure of browser usage share. It is understandable, therefore, that in evaluating the purchase of Netscape, AOL viewed AdKnowledge's hit data as one of the more reliable indicators of trends in the relative shares of all browser usage enjoyed by Navigator and Internet Explorer.
370. Microsoft's economic witness, Richard Schmalensee, testified survey data collected by Market Decisions Corporation ("MDC") provide a more accurate measure of the usage shares enjoyed by different brands of Web browsing software than AdKnowledge's hit data. The calculations that Schmalensee made using the MDC data lead to results that differ, in one main respect, from the results generated with hit data. Whereas the AdKnowledge data show Navigator's share falling from seventy-five to fifty-six percent from the first to the third quarter of 1997, the MDC data show Navigator's share holding steady at fifty-five or fifty-six percent over the same period. Although both sources show Internet Explorer's share gaining steadily throughout that period, the MDC data indicate that Internet Explorer's rise was coming not at Navigator's expense, but rather at the expense of other browser products, which, according to the MDC data, collectively enjoyed a substantial share into 1997. The AdKnowledge data, by contrast, indicate that the share of usage attributable to browsers other than Internet Explorer and Navigator has never been substantial and that Internet Explorer's rise has always been at Navigator's expense.
371. The MDC estimates of the shares attributable to Navigator and other non-Microosft browser products in 1996 differ markedly from those on which Microsoft and AOL relied in the course of making business judgments. Notably, in August 1996, four months after it commissioned the first MDC survey, Microsoft continued to estimate Navigator's share as exceeding eighty percent. In fact, the senior Microsoft executives who testified in this trial still believed at the time of their testimony that Navigator's usage share in late 1995 and early 1996 had exceeded eighty percent. To the extent the MDC estimates differ from those which Microsoft and AOL used internally, and which senior Microsoft executives still embrace, the Court is inclined to trust the latter estimates. More broadly, the sets of questions contained in the MDC surveys and the internally inconsistent responses they evoked reveal that a substantial percentage of the respondents misunderstood some of the patently ambiguous questions they were asked, and that a large number responded to questions when they were unsure of, or even clearly misinformed regarding, the answers. The Court accordingly gives no weight to any of the conclusions that Microsoft draws from MDC survey data.
372. In summary, the estimates on which Microsoft and AOL relied and the measurements made by AdKnowledge and the University of Illinois provide an adequate basis for two findings: First, from early 1996 to the late summer of 1998, Navigator's share of all browser usage fell from above seventy percent to around fifty percent, while Internet Explorer's share rose from about five percent to around fifty percent; second, by 1998, Navigator's share of incremental browser usage had fallen below forty percent while Internet Explorer's share had risen above sixty percent. All signs point to the fact that Internet Explorer's share has continued to rise -- and Navigator's has continued to decline -- since the late summer of 1998. It is safe to conclude, then, that Internet Explorer's share of all browser usage now exceeds fifty percent, and that Navigator's share has fallen below that mark.
373. These trends will continue. In February 1998, Kumar Mehta, the Microsoft employee responsible for tracking browser share, told Brad Chase that Microsoft's best model projected that Internet Explorer's usage share in early 2001 would stand between sixty and sixty-eight percent. This comports with the forecast on which AOL relied in deciding to purchase Netscape: The report presented to AOL's board of directors prior to their vote on the transaction predicted that Navigator's usage share would fall to between thirty-five and forty percent by late 2000. The most reasonable prediction, then, is that by January 2001, Internet Explorer's usage share will exceed sixty percent while Navigator's share will have fallen below forty percent.
374. Navigator's large and continuing decline in usage share has demonstrated to developers the product's failure to mature as the standard software used to browse the Web. Internet Explorer's success in gaining usage share, together with the lack of contenders other than Navigator, has simultaneously sent the clear message to developers that no platform for network-centric applications can compete for ubiquity with the 32-bit Windows API set.
375. The changes in usage share described above would likely not have occurred had Microsoft not improved its browsing software to the point that, by late 1996, the average user could not discern a significant difference in quality and features between the latest versions of Internet Explorer and Navigator. As Microsoft's top executives predicted, however, Internet Explorer's quality and features have never surpassed Navigator's to such a degree as to compel a significant part of Navigator's installed base to switch to Internet Explorer. An internal Microsoft presentation concluded in February 1998 that "[m]any customers see MS and NS as parity products; no strong reason to switch," and another internal review three months later reported, "IE4 is fundamentally not compelling" and "[n]ot differentiated from Netscape v[ersion]4 -- seen as a commodity." For a time, even among new users, Navigator was likely to win most choices between comparable browser software, because most people associated the Internet and cutting-edge browsing technology with Netscape rather than with Microsoft. So, if Microsoft had taken no action other than improving the quality and features of its browser, Internet Explorer's share of usage would have risen far less and far more slowly than it actually did. While Internet Explorer's increase in usage share accelerated and began to cut deeply into Navigator's share after Microsoft released the first version of Internet Explorer (3.0) to offer quality and features approaching those of Navigator, the acceleration occurred months before Microsoft released the first version of Internet Explorer (4.0) to win a significant number of head-to-head product reviews against Navigator. This indicates that superior quality was not responsible for the dramatic rise Internet Explorer's usage share.
376. Including Internet Explorer with Windows at no additional charge likely helped the usage share of Microsoft's browsing software. It did not, however, prevent OEMs from meeting demand for Navigator, which remained higher than demand for Internet Explorer well into 1998. Moreover, bundling Internet Explorer with Windows had no effect on the distribution and promotion of browsing software by IAPs or through any of the other channels that Microsoft sought to pre-empt by other means. Had Microsoft not offered distribution licenses for Internet Explorer -- and other things of great value -- to other firms at no charge; had it not prevented OEMs from removing the prominent means of accessing Internet Explorer and limited their ability to feature Navigator; and had Microsoft not taken all the other measures it used to maximize Internet Explorer's usage share at Navigator's expense, its browsing software would not have weaned such a large amount of usage share from Navigator, much less overtaken Navigator in three years.
Microsoft Findings Home
Findings of Judge Thomas Penfield Jackson, Nov. 5, 1999