The big Web ratings company Nielsen//NetRatings announced today that instead of ranking Web sites according to the number of "page views" they receive, it will, instead, order sites by how long people spend there. The move is meant to deal with modern "rich" Web applications. Sites that use newfangled technologies like AJAX (think Google Maps) or Flash (think YouTube) don't need to refresh HTML pages very often, meaning that counting their "page views" may make them look less popular than they actually are. Counting how many minutes people spend on the site provides "a common denominator for user behavior that is independent of site design," Nielsen//NetRatings said in a press release.
Before delving further into this let's go over some nomenclature. In the beginning people coveted hits. This was circa 1995. A hit is a single request from a server for a bit of data -- a text file, a picture, a sound file. But because a Web page contains many bits of data, hits don't exactly correspond with popularity; a page that has a lot going on -- say, Salon's home page -- will generate many hits for each visit, but a sparse page -- like Google.com -- will get only a few.
At some point people realized this and began to look at page views. A Web page view is generated each time you load or refresh a site; go to nytimes.com and you're giving the paper one page view, click on an article and you're giving it another, click on Page 2 and you're giving it one more, and so on. (BTW: Juicing views is the main reason newspaper and magazine Web sites -- including Salon -- split up articles into multiple pages. Jason Kottke assessed this practice best: "Blech and BOOOOOOOOOOOOO!!!!")
Page view-based rankings benefit sites that demand a lot of interaction. Search engines, social-networking sites and blogs are page view factories; you keep clicking, reloading, searching for something new, and each time, you're giving them a new view. But take a look at Google Maps. The first time you visit, it loads up a map -- and then it never refreshes the page. Type in "1600 penn ave dc" and it shows you the White House, but does not generate a new page view. Now click Find Businesses, and then type in "pizza." The map adjusts to show you pizza houses closest to the seat of the executive branch, but does not generate a page view. You could go on like this all day or all week and never tally up a view for Google.
So that's why Nielsen decided that it's the time you spend on a site, rather than the specific technical way in which you interact with it, that should determine its ranking. But does this make sense? Is a site that demands your attention longer really a better indicator of "engagement" and utility than one you spend just a few seconds on? The one obvious counterexample is Google's search engine, which is designed to give you results as quickly as possible -- designed, in other words, to minimize how long you spend there.
When Google works well, you're on the page for next to zero seconds; indeed, when you're confident it'll work very well, you can press the "I'm feeling lucky" button and skip by the results page altogether. When Google doesn't work well you're there more than you feel you can bear -- and the more time you spend, the less useful it becomes. The same is true for most reference applications: You want to spend less time, not more, on a weather site, or price-comparison site, or a page of headlines. The Drudge Report is one of the biggest sites on the Web -- but nobody lingers at Drudge. The page is not the point; its links are.
Consequently the new time-based rankings downgrade Google and puff up other properties. By page views in May 2007, Google was the No. 3 "Web brand" in the country; but by time it ranks fifth, with users tallying up 7.4 billion minutes on the site. Ranking higher is Fox Interactive (which owns MySpace; 7.8 billion minutes in May), MSN (10.6 billion minutes), Yahoo (19.6 billion) and AOL (25.0 billion minutes). Why do Yahoo and AOL do so much better than all the others? Because Nielsen counts the time people spend on their instant-messaging applications, a huge time-suck.
The numbers make no sense. The Nielsen ratings are aimed at advertisers -- they're designed to guide ad buyers to the hottest Web properties. Google runs the biggest ad business on the Web, and execs at Yahoo and AOL would kill to get a fraction of it. So any ranking that puts these firms so far ahead of Google raises a basic question: Are all Google's advertisers being duped -- should they move to AOL or Yahoo's more "engaged" audience? Or is Nielsen's time-spent metric meaningless?
Note that Google's search-engine ad model thrives on feedback. People buy ads based on clicks -- regardless of how many times Google shows your ad to Web searchers, you only pay Google when someone clicks on it. Clicking is a far better measure of "engagement" than time; if a Web searcher spends just three seconds on a Google results page but clicks on your listing, it's far better for you than if she spends 10 minutes and doesn't click.
The new Nielsen numbers, then, really only highlight the uselessness of Web rankings. As Saul Hansell points out in the New York Times' tech blog, the time-spent measure may give AOL and Yahoo bragging rights -- but bragging rights aren't dollars. And in the online ad sweepstakes, all those dollars rolling in to Google are what really count.