The new round of Internet-browser wars is accompanied by a battle over browser statistics. At stake: which Web browser, Microsoft Corp.'s or Google Inc.'s, is the most popular around the world.
Last month, a company that tracks browser market share declared that Google's Chrome had eclipsed Microsoft's Internet Explorer as the pre-eminent browser, 33% to 32%. The company, Dublin-based StatCounter, had shown Chrome on the ascendancy for some time. Microsoft disputes StatCounters' numbers, though, and cites a browsing-counting competitor that finds Internet Explorer remains the choice of over half the world's Internet users, leading Chrome, 54% to 20%.
Earlier this year, a Microsoft executive made that argument in a blog post. Last week, StatCounter hit back with a 10-point response, posted online. But none of the back-and-forth has cleared up the bigger question: Which Web browser is used by how many people?
The answer matters to Microsoft and Google, who see browser choice as a crucial part of their strategies for influencing how people useâ€"and searchâ€"the Web. It also matters for Web developers who must decide which browser should be used as the basis for optimizing their sites. "Whoever owns the browser has a chance of tilting the playing field towards their ecosystem of products and services," says Al Hilwa, a software analyst with the research firm IDC.
That no one is sure which browser is on top highlights a persistent and puzzling characteristic of Web measurement: Internet usage automatically produces terabytes of trackable data, but interpreting them is complex, and getting more complicated all the time.
Among the difficulties comparing browsers is that Internet Explorer and Apple Inc.'s Safari browser are made by companies that also make operating systems and tie the browsers to them, often including them with new computers. Chrome and the Firefox browser, which is competitive with Chrome in usage by most comparisons, typically must be downloaded by choice.
Measuring Internet usage was easier during the prior round of browser wars in the 1990s, when Internet Explorer challenged and eventually overtook market leader Netscape, recalls Nathan Newman. Dr. Newman, who has a doctorate in sociology, studied browser share at the time for NetAction, a nonprofit, San Francisco-based Internet consumer group that is no longer active. "The Web was just a simpler place in the prehistoric times of the 1990s, when the Internet was the Internet and a browser sat on a computer," Dr. Newman says. "So it was pretty straightforward to track which browsers people were using."
No longer, as people access the Web from their phones or computers, directly or using proxies or RSS readers. It's not always humans accessing Web pages, either: The number of automated programs, such as search-engine crawlers that access pages, has grown rapidly. Browsers also have changed. Chrome, for instance, sometimes preloads pages in the expectation that users are likely to want to see those pages, based on their online activity. And Internet usage has grown around the world, with China now home to the most Internet users, by several estimates.
These issues complicate the work of the Web-browser bean counters. They rely on traffic logs of websites that use their Web-tracking tools to see which browsers are being used to access them. However, those sites might not be representative of the Internet as a wholeâ€"if, for instance, they aren't as popular with Chinese users as with American ones. Also, if programs, not people, are accessing the sitesâ€"either because of bots, or because browsers are automatically downloading pages that haven't been requestedâ€"data could be muddled in the process.
All of these factors come up in the current debate over browser share. In his March blog post, Roger Capriotti, Microsoft's director of Internet Explorer product marketing, criticized StatCounter for counting pages that Google had "prerendered,"or loaded in the background. StatCounter responded by removing those, and found it reduced Chrome's market share by just three-tenths of a percentage point.
These prerendered pages matter because site logs could signal extra Chrome activity that isn't the result of a person choosing to read those pages, thus inflating the numbers.
Net Applications, a StatCounter rival that Microsoft says is more reliable and which shows Internet Explorer to have more than twice the number of users as Chrome, says its methods do better at removing bots from data. Vincent Vizzaccaro, executive vice president of marketing and strategic relationships at Aliso Viejo, Calif.-based Net Applications, calls StatCounter's method "very susceptible to bot attacks." StatCounter founder and chief executive Aodhan Cullen says the company screens out known bots and finds others from unusual spikes in traffic from a single source.
The biggest reason for the discrepancy, though, is the different way the two companies measure numbers on global traffic, and how they crunch the numbers. Both measurement companies offer tools for Web publishers to assess traffic, then use the data from those tools to extract which browser readers of those Web pages use. But each takes a different tack in analyzing the data.
Net Applications acknowledges that visitors to its customers' websites likely aren't representative of all Internet users, and weights the data by the population of Internet users world-wide, according to the U.S. Central Intelligence Agency's World Factbook. So if, for instance, its numbers show China's share of world Internet users is half what the CIA's numbers show, Net Applications would count each Chinese browser in its data twice. StatCounter presents its data as is, and even though other counts of world-wide Internet users generally rank China first, the company's reading suggests that China ranks 22nd in the world in Web usage. StatCounter notes that Cisco Systems Inc. data indicate that the average Chinese user generates far less traffic than the average American user.
Mr. Cullen explains his company's approach in two ways. First, he says that the key measure of browser usage is actual browser activity, or pages viewed. Net Applications is measuring unique visitors. That's often the preferred metric for advertisers on a given website, as they don't want to keep paying each time the same ad is shown to the same person. But for the purposes of determining which browser is most used, and most important, usage makes the most sense to Mr. Cullen.A Microsoft spokesman declined to comment beyond a statement the company released last week reiterating that it preferred Net Applications' numbers because of the weighting issue. Google declined to comment.
Also, there are no reliable numbers on Internet users by country, Mr. Cullen says. Different groups have different estimates, and most don't separate mobile Internet users from those accessing the Web from a computer, even though the two groups face different choices of browsers. "There's no good weighting system that could be used that is accurate," says Mr. Cullen. "That's why we haven't applied any weights to our data."
Mr. Vizzaccaro says Net Applications' approach is meant "to provide a more accurate picture of market share showing the number of users of a technology instead of the number of clicks."
How StatCounter handles China makes a big difference in its findings. In its blog post the company said Internet Explorer's market share in May would have been 43% instead of 35% if StatCounter had used the same weighting technique as its rival, which would have been enough to propel Microsoft into a healthy lead. Puzzlingly, even with that change StatCounter still would be showing a decline in Internet Explorer usage in the past year, contrary to Net Applications.
Asa Dotzler, product manager for Firefox at Mozilla, which developed the browser, the No. 3 browser by StatCounter's estimate, said he can't confirm any particular numbers but called a competitive browser market one indicator that "we are moving toward our goals."
Mr. Cullen, who said his company doesn't support any particular browser, agrees. "It's fantastic having three or four major players," he says. "Nobody is king of the Web anymore."
â€"Learn more about this topic at WSJ.com/NumbersGuy. Email numbersguy@wsj.com.
No comments:
Post a Comment