|
<advertisement>
|
 |
 
|
|
media search: |
 |
web watch
Wanted: A Way of Counting That You Can Count On
by James Ledbetter
James Ledbetter writes for CJR
about the Internet and other new media, and their growing impact on journalism.
Ledbetter has a unique vantage point as New York bureau chief of The
Industry Standard (jim@thestandard.com), a magazine
that covers the Internet industry. He formerly was media critic for The
Village Voice and is the author of Made
Possible By - The Death of Public Broadcasting in the United States.
It’s
one of the most fundamental questions of the media business: How many people
see, hear, or read what you produce?
For decades, traditional media organizations have refined the ways they answer that question. At newspapers and magazines, the consensus has been the audit: a publisher makes a claim to reach a given number of readers and an outside agency (like the Audit Bureau of Circulations) checks certain guideposts and either accepts or refines the number. For radio and TV, companies like Nielsen and Arbitron provide ratings based on a sample of tightly monitored consumers.
Media insiders know that these numbers can be distorted or even misleading. Print circulation depends on too many factors to ever be measured exactly; a single poor-selling issue of a magazine can set back or exaggerate official circulation figures for a year or more. Moreover, magazines often claim — at varying levels of plausibility — that single copies are viewed by so many people (in, say, barber shops and libraries) that their "readership" is far higher than their paid circulation.
And for all the seeming precision of broadcast ratings ("Program X consistently receives a 1.7 rating"), the sampling method has come under repeated attack. Critics such as Jesse Jackson have maintained, for example, that too few minority households are included in TV audience samples to reflect the actual nationwide viewership.
The Internet was supposed to change all that. Because every impression on an Internet site is recorded, the people who run them have access to an array of data that is astonishingly specific. By simply examining the log of users to a given site, a Web master has almost pinpoint precision about how many people come to visit. He or she also knows which pages on a site are visited and which aren’t; how long visitors stay on individual pages; whether they got to the site via a search engine, bookmarked Web browser, or a banner ad.
And yet: try to get a single answer about how many visitors a Web site has and you will encounter dissembling and dissent. Do you want to determine whether more people read Slate or Salon? Good luck. Even knowing how many people visit a single company’s site can be difficult.
Part of the difficulty is that Web publishers guard their numbers closely. They are a valuable commodity for advertisers and marketers; indeed, for Web publishers who have no fixed assets — no real estate, no printing presses, etc. — user data is often the most valuable thing they own.
But there is a broader, industry-wide problem. The few paid sites, such as The Wall Street Journal Interactive and TheStreet.com, are exceptions, since they can measure subscribers who’ve coughed up money. But on the whole the Internet business has no single, agreed-upon method of defining a Web audience. There is no consensus.
In the early days, individual Web pages frequently referred to the number of "hits" they received, either cumulatively or during a particular period. That yardstick quickly proved inadequate. A person who logs in from the office and from home is counted as two hits. Indeed, a "hit" is recorded every time anyone, anywhere clicked to a site, even if the full Web page did not load on to a user’s screen.
As sites became more mature and incorporated multiple pages, the "page view" became a more common measurement, but soon proved to be only a small improvement over hits. For example, if someone visits a Web site’s home page, reads a story that takes up two pages, and returns to the home page, that counts as four "page views." If they repeat that process with a new two-page story every day for a week, that’s twenty-eight page views. But it is still only one person — a ratio that makes magazines’ "pass-around" figures look modest. In November, a press release arrived from ESPN Internet Ventures, claiming that their combined sites received a whopping 1.3 billion page views in October. That sure sounds impressive, except when you consider that there aren’t anywhere near that many people on earth who have Internet access. There is a huge gap, in other words, between page views and actual readers.
Over the last couple of years, the "unique visitor" has emerged as arguably the most reliable way of gauging audience. A unique visitor is, theoretically, a single person who visited a site within a specified period. The firm most identified with this standard is Media Metrix, a New York-based company that became the leader in its field after gobbling up its largest rival, Relevant Knowledge, in 1998.
Media Metrix uses a sample audience of 50,000 computer users both at home and at work. It releases data on a monthly basis; sites that refer to their audience in "monthly unique visitors" are generally using Media Metrix numbers. Last year, the Nielsen Company also began offering a formal Internet ratings system, using a sample audience but publishing its results weekly; it does not measure at-work Internet usage. A third service, PC Data Online, recruits a panel through advertising and telemarketing and then weights its findings to reflect the presumed demographics of the Internet population.
Media Metrix says its samples are the best way of measuring a site’s unduplicated audience. Yet many media executives complain bitterly about being undercounted by Media Metrix. A Wall Street Journal story in November, for example, quoted an official from Yahoo who said that the firm’s measurement of his site’s traffic was "just wrong." (Of course, many of these firms that complain about Media Metrix put out press releases when their Media Metrix numbers go up.)
Some complaints have merit: sites with large numbers of international readers are almost certainly undercounted. Doug McFarland, senior vice president at Media Metrix, concedes that "Our panels for years were based only in the U.S., and so of course we didn’t count international traffic." Now, with more than 10,000 sample users abroad, the company claims to measure 72 percent of users worldwide. (Rival Nielsen NetRatings says it will begin offering coverage in thirty countries by 2001.)
Another common complaint is that Media Metrix does not adequately measure visitors who come to Web sites using America Online. Perhaps capitalizing on this perception, a rival, PC Data Online, boasts of its ability to count AOL. By some measures, AOL handles nearly half of all Internet traffic.
McFarland insists the AOL charge is groundless. "It’s absolutely false," he says, explaining that the company’s patented meter molds to a computer in ways that can measure any Net activity accurately.
Such controversy is understandable, given how fiercely companies compete for Web visitors. There is at least one other possible way to count them: the logs to Web sites could be submitted to auditors, just as newspaper and magazine circulation reports are. Unfortunately, as sites get larger, the hardware and software required for that task becomes bulky and potentially expensive. As it is, Web companies often hire separate companies just to count their own numbers.
The main reason for knowing such figures precisely, of course, is to attract advertisers and charge them accordingly. There is another, potentially troubling use for them, however. For journalists, the notion of knowing exactly how many readers read which pieces might be useful, but it can be dangerous. At my own magazine, we get a top ten list once a month of the Web site’s most-read stories from the previous month. It’s fun, revealing, and inspires a little friendly competition.
If, however, editors began determining assignments based on the most-read topics, that could spell trouble. There is a thin line between meeting reader’s needs and pandering. For one thing, the temptation to put salacious headlines on articles could become severe. Writers at Salon say that their rule of thumb is that articles with "sex" in the headline get twice as many hits as average, articles with "oral sex" in the headline three times the average. It’s easy to see how in such an environment serious articles on topics "known" to be unpopular might simply be discouraged.
Even more than in traditional media, then, it becomes incumbent upon online editors to set priorities based on good journalistic judgment. There may be a certain inefficiency built into the way that old-line media measures its audience, but if the alternative precision leads to shying away from important stories for fear that the audience will turn away, then a little inefficiency looks pretty good.
