Hits, page views and other garbage we pass off as audience metrics

This week, Nielsen/NetRatings announced that it will switch emphasis in audience measurement reports from “page views” to “time spent on a website.” The move reflects the changing nature of the Web experience over the past few years, but it’s just a toddler’s step toward the larger goal of cleaning up the mess that is online audience metrics.

The problem

Web veterans will remember when “hits” were the preferred term for describing a website’s popularity. (Please see OJR’s glossary of online news terms for an explanation of the difference between “hits” and “page views.”) Slick webmasters soon figured they could inflate their “hits” simply by loading their pages with dozens of graphic files, so “page views” have been since the late 1990s the preferred metric for describing how much content a website has served to its readers.

But readers don’t need to load a different page (i.e. a new URL) to see new content anymore. AJAX-powered webpages can serve up “page” after “page” of fresh content while on the same URL. And a single YouTube URL can keep a viewer watching for several minutes. Today, a huge “page view” count is as worthless as a huge “hit” count in confirming a site’s popularity.

But “time spent on a website” does not always accurately reflect a site’s worth to advertisers, either. What good is advertising on a site where readers will spend an hour, if none of those readers care about what you sell? Audience composition is, and always will be, the metric of primary importance to advertisers.

And different advertisers will care about different aspects of audience composition.

“As a local newspaper site, our focus is local… local… local,” wrote Chris Jennewein, Vice President, Internet Operations, at The San Diego Union-Tribune. “Local online penetration (% of local-market adults) and total unique users are the best statistics for us.”

On a niche topic website, however, the geographic location of readers matter little in comparison with their interest in the topic and their buying habits. On the niche sites that my wife and I run, we survey annually to measure our readers’ likelihood to purchase within the next 12 months the products and services associated with our niches.

Either way, a Web publisher must first be able to demonstrate that it is reaching the audience that its publication is designed to reach. Once that is established, sites can distinguish themselves from other such sites through metrics such as “time spent.”

Ultimately, a publisher ought to be able to show not only that it is delivering readers to an advertiser, but that it is positing the ad in a way that readers will see it, without offending readers to the point where they turn against the advertiser, or the site. (Remember the hated “X10″ pop-up ads, anyone?) “Time spent on the website” doesn’t deliver that assurance.

Web publishers are not the only ones dealing with ad layout efficacy issues, of course. Digital video recorders, such as Tivo, have made it easier than ever for TV viewers to skip commercials, forcing TV networks and producers to find new ways to embed advertising messages within shows. Eyetracking research has shown many publishers how to lay out static HTML pages to draw the most available eyeballs to their ads. Additional eyetracking research, coupled with fresh data from our television colleagues, should help guide Web designers in the AJAX/video era.

What you can do about it

In the meantime, however, Web publishers, large and small, can take immediate steps to clean up information they collect and publish about their traffic.

Quit using the word ‘Hits’

July’s typically a slow time of year in most newsrooms — the time of year when editors can think about things such as their organization’s style book. So why not suggest to the style gurus in your newsroom a ban on the term “hits” in reference to Web traffic — in all news stories.

While we’re at it, let’s educate the reporters in our news organizations to quit asking about “page views” as well. When journalists report a website’s popularity, ask publishers to provide the number of absolute unique human visitors to their site in a typical day, week or month. If a publisher wants to make a point about his or her readers’ loyalty, let him or her quote the site’s time spent or repeat visitor rates as well.

But under no circumstances should a professional news reports ever use the term “hits” in a story about a website again (unless it is a story about a dDOS attack). Consider yourself warned, as an “OJR Hall of Shame” might otherwise be forthcoming.

Use a third party to report your stats

At this point, I know of no good reason not to use a third-party reporting agency to collect and analyze your traffic data. It’s far easier than filtering your own log files for spiders and bots, and you can get high-quality traffic reporting for free.

Just because a publisher has accurate information about its traffic, however, does not guarantee that publisher will release it to advertisers, or other websites. That’s why companies such as Nielsen/NetRatings exist, to provide publishers and advertisers with a neutral third party to report individual websites’ traffic data.

Unfortunately, most of these third parties build their industry traffic rankings from surveys of a sample of Web readers or a sample of ISPs’ traffic logs. Those samples can be skewed if they do not reflect an accurate mix of home, work and school users. And even if their samples are not skewed, these rankings cannot report data for the millions of “long tail” niche and hyperlocal sites whose traffic does not rise above the rankings’ margins of error.

More ideal third-party ranking services would instead base their reports on data collected from tracking scripts on participating websites.

Coastsider publisher Barry Parr recommended Quantcast for such publishers: “I recently (last week) “Quantified” my site, adding the javascript for Quantcast to my pages in order to create a relatively independent public record of my unique users that can be compared to my competition. This is about as close to an ABC audit as small Web publishers are likely to get. I’d like to shame my competition into doing the same.”

Quantcast’s data for publishers who have not installed its tracking javascript are highly suspect, as they tend to fall far below comparable sites that do use Quantcast’s script. I’d like to believe Quantcast when it reports that one of my personal niche websites, which uses the Quantcast script, is more popular than Slashdot, which does not, but… c’mon, who are we kidding?

As Parr alluded to, an industry standard tracking service could finally provide accurate, apples-to-apples data with which publishers could see where they stand versus the competition… and with which advertisers could make a more informed decision on where their budgets should be spent.

Show, don’t tell

Maybe those j-school classes did teach us something about running a business after all. Over the long run, third-party traffic data, survey results and advertiser testimonials can help show potential advertisers a more accurate picture of our websites’ ability to attract readers and deliver them to an advertiser than another salesperson’s spin will.

Tell it like it is. And if you’re not happy with that, think about how you ought to change your site to make it a better vehicle for your readers and your advertisers. That’s a better investment of your time that wasting it trying to game your stats to make it look like you’re already there.

About Robert Niles

Robert Niles is the former editor of OJR, and no longer associated with the site. You may find him now at http://www.sensibletalk.com.


  1. “When journalists report a website’s popularity, ask publishers to provide the number of absolute unique human visitors to their site in a typical day, week or month.”

    Given all the non-cookie-enabled browsers in the world, how would we go about doing this?

    UV measures aren’t perfect but they’re as good as anything the author presents as an alternative for many Web sites. Total time spent on a site is pretty meaningless if the user isn’t engaged. PVs give a decent measure of engagement for many sites.

    One big problem in determining a standard measure is that Web pages aren’t used in the same way. For a site like mine (WLKY.com) PVs are a pretty solid measure given that each of our articles, videos, indices and pics has its own URL. What’s more, it’s a TV site, so the articles, built from scripts [EDIT: By “scripts,” we mean TV scripts, not code that generates the articles], are generally pretty brief, which limits time spent on the site even for very active, engaged visitors.

    But UV measures might not work well for a site that offered multiple things on the same page, or videos of extended length. Then time on the page would be far more important. If every video you offer is 15 minutes long, and people average 30 seconds on your site racking up 15 PVs, your audience is probably not engaged.

    Anyway, long story short, I think PVs are fine for some sites and bad for others. It’s not TV or the newspaper — media which are essentially accessed the same way wherever they’re presented — and, thus, it’s very difficult to find a standard measure.

    Frankly, this isn’t such a bad thing. I think big ad firms are used to looking at very simple metrics and just plugging them into a campaign.

    What’s more, the old, homogenous way of doing things may or may not be effective for the advertiser, depending on if the apples are being forcibly compared to oranges or not.

  2. While page views can retain value for internal use, to allow people within a news organization to track the relative popularity of individual articles, etc., they hold no value in making comparisons between websites, as some websites use a traditional HTML page architecture and many others do not.

    So while I rail against PVs as an industry metric, I do not mean to say that they retain no value as an internal metric. (Heck, internally even hits can be a useful term in determining server load.)

  3. Patrick Slavenburg says:

    How does Quantcast differ from Alexa and Compete.com (it looks like compete). Same type of guesstimates.

    I am totally not impressed by their user demographics. You see something similar done by MSN in their developer platform.. and it is SO NOT correct.

    I have the impression they use panels to benchmark my visitor demographics. IF this is the case.. how on earth could you recommend this as so much better compared to pageviews (with their own set of flaws ofcourse).

    Income level analysis ? Geez I wish it were THAT simple.. use Qauntcast and I do not have to worry about it anymore.

    I wonder why Google Analytics has not offered it as an add-on-for pay service ? They already have their javascript loaded in tens of millions of pages, AND it’s an independent analytics. I am sure many would pay for such reliable stats on user demographics, income levels etc.

    Perhaps there is a reason why they didn’t ??

  4. Patrick Slavenburg says:

    It all really depends. I agree with the first comment. It just *depends*. If you have advertizers who like simple metrics.. hey I know sectors where they still look at # of visitors per month, that is it. Basically CPM. Fine.. other advertizers or sponsors may want a more in depth analysis from you.

    It doesn’t really matter if it comes from your own logfiles (and there is software to weed through this) or some javascript code that may or may not be read by browsers. Heck.. so many bots pretend to be humans and scrape your site it doesn’t even matter anymore. It also depends where you place your javascript: top or bottom of page.. is it fully loaded or not ? So once again… it all depends.

    Return visitors yes, and for journalism/blogs it’s a pretty decent metric.. for other areas it is not. And so you can continue.

    So here’s the thing: if you want one or two industry metrics, they’re all going to be flawed and really one is not much more superior than the other.

    (After all there is a reason advertizers do not pay much for 1000cpm’s on social networks with their inflated pageviews). And advertizers know it.. but may want it for internal political reasons.

    And yes.. Time spent on site is going to look better and better with AJAX interfaces, but an intelligent advertizer, would want to know why.

    If not.. and you get an extensive discussion and they are willing to listen, I think you can make pretty clear what is happening on your site and why. Once again I do not see any added value in replacing one lousy process by another (quantcast)