Answering my own questions since 2001

Page Views, Visitors, Apples and Oranges

Something else kind of stuck in my brain from Stephen Hume’s column. His claim that the Vancouver Sun had received 10 million page views in February, 2009 seemed unusually high.

Warning: This post gets pretty web-analytics-geeky very quickly, so bail out now if that doesn’t interest.

I checked out the Sun’s online advertising site. According to their downloadable PDF, these were the traffic numbers for May, 2008:

vancouversun.com
7.2 million monthly page views to vancouversun.com
522,000 unique visitors in May 2008 on vancouversun.com

theprovince.com
3.7 million monthly page views to theprovince.com
391,000 unique visitors in May 2008 on theprovince.com

There’s some fine print at the bottom of the page which indicates that the page view numbers come from (the links are mine) “Source: Omniture SiteCatalyst, Avg. May 2008″ and the visitor numbers come from “Source: comScore Media Metrix, Total Canada, Home & Work, May 2008″.

Analytics and Panels

I take an interest in those sources because Omniture SiteCatalyst provides a more accurate visitor total than comScore. SiteCatalyst is an analytics-based tool like Google Analytics, and if it’s counting page views, then it’s counting visitors, too. Like any such ‘web-bug’ system, VancouverSun.com has code on every page that enables them to capture and report on behaviour for each of their site visitors. I grabbed a screenshot of that code from a page on the Sun’s website.

ComScore, on the other hand, is panel-based. They use a sampling of over two million online consumers who have opted-in to share their online attention data with comScore. It’s a reasonable approach, but unquestionably less accurate than on-site analytics.

So why isn’t CanWest reporting the visitor number from their analytics tool? If their page view number is accurate, and I have no reason to doubt it, then the number of visitors from their analytics is surely higher than 522,000.

Why do I think that? First, there’s a correlation between “unique visitors” and “visitors”. Looking at the stats for a number of medium to high traffic sites, the ratio of unique visitors to plain old visitors (meaning those who may have returned multiple times) is about 1 to 1.25. So, working with the comScore number, let’s say the Sun has about 650,000 visitors. I do this extra bit of analysis because I want to consider the common metric “page views per visitor”.

Too Many Page Views or Too Few Visitors?

Assuming 650,000 visitors and 7.2 million page views, that means the average visitor views 11 different pages on the site per visit. Anybody who works in SEO or analytics will tell you that that number is extremely high. I’ve never seen stats for a site where the number of page views per visitor is higher than 5. Have you? Most are between 1.5 and 4. This site, for example, averaged (a very lousy) 1.5 page views per visit last month. The math for TheProvince.com is a little better, at about 7.5 page views per visitor.

So what gives? I don’t know. If that page view number is accurate, then CanWest ought to be promoting a visitor number that’s significantly higher than 522,000. More visitors means more popularity, which in turn makes the newspaper sites more attractive to advertisers. I don’t mean to imply any conspiracy–the likeliest explanation is just an oversight.

As an interesting side note, Kirk LaPointe, the Sun’s deputy editor, points out that “more than half of our traffic at vancouversun.com arrives from search and a large percentage of that is non-local traffic.” This is true for the stats of every established site I’ve ever seen. That fact presents an interesting online advertising challenge. Do they play up the local chunk of their traffic, reach further abroad with the non-local portion, or both for different audiences? I emailed CanWest and the Sun late last week to ask for a comment on this story, but haven’t heard back.

Any thoughts? Or is my math wonky?

15 Responses to “Page Views, Visitors, Apples and Oranges”

  1. Robert

    There is little doubt that panel based metrics, while they have their place, are much less accurate than empiric analytics. The problem is really rooted in the advertising world. Ad agencies are used to dealing with TV. TV is all panel based, because metrics aren’t an easily constructed part of television as a medium.

    I don’t understand, based on what the Web supports as a medium, why people still even use panels.

  2. D'Arcy Norman

    my photoblog gets 5.15 pageviews per visit at the moment. meaning that one meaningless metric, divided by a second meaningless metric, generates a third, also meaningless metric.

    pageviews, visits, etc… are all just seo circlejerking.

    darren Reply:

    They’re less circlejerking when you’re in marketing. Obviously visitors only matter in as much as you can get them to ‘convert’–to do what you want–but that’s how business on the web works, right?

  3. D'Arcy Norman

    and this really points to the problem with advertising supported journalism – first, there is an incentive to artificially inflate “readership” using any metric that looks most impressive. Second, there’s an incentive to write articles to maximize that “readership” – boosting page views and visits in order to turn around and charge for advertising.

    the problem is that this makes the newspaper not an objective source of reporting – they have a vested interest in boosting numbers to prop up falling advertising revenue.

    darren Reply:

    I generally agree. I read a great Slate piece a couple of weeks ago about how, in fact, newspapers have often been the less profitable arm of a larger corporate interest. Hence, the way forward may be to recognize them as non-profit foundations instead of money-making enterprises. That’s the most reasonable suggestion I’ve seen recently.

    pavlicko Reply:

    D’arcy –

    I definitely agree. If I were a newspaper or ad-sponsored site, I’d be loading up my pages with ajaxy crap everywhere, just to artificially inflate page views.

    Of course I don’t have a solution either. Am I going to pay a subscription price to read online news?

  4. Laurence Miall

    Personal and professional question here… Would the reason that your own average page view is only 1.5 be because it’s a blog, and hence, I can scroll down and read lots of content without clicking on separate pages?

    I’m wondering to what extent the architecture of a site will prompt difference page view results.

    darren Reply:

    I’d imagine that there are lots of different reasons. I’ve never tried to improve the average page views, but here are a couple of thoughts:

    * As you mention, regular readers just hit the front page.

    * A lot of my traffic comes from search, and given the diversity of the content on my site, it’s natural that the average searcher wouldn’t browse around very much. Maybe I should stick a related posts widget high in a sidebar of individual pages, and see if that helps?

    * The same goes for visitors coming from external links. They’ve probably been pointed at a article, as opposed to the site as a whole.

    Other factors that contribute to average page views: page layout, the type of site (a client with a community site has a very high average page view count), whether articles are paginated (as they are on the Sun’s site) and so forth.

  5. Laurence Miall

    That’s a very useful insight into it. Thanks. Clearly the way the blog is set up right now is working. I resist the idea of fragmenting online content into tiny little niches — especially if it’s a personal blog.

    However, the idea of a related posts seems a really good one. I have personally wondered whether allowing related posts to get buried way back in the chronologically archives is the best way of bringing back people who may only be looking for posts on one particular topic.

  6. Boris Mann

    Site architecture has a TON to do with many of these metrics. For off site traffic, or deep linked search result traffic, how do you engage a visitor?

    I’m a big fan of “related articles”. For blogs like this, I put them in at least two places. Once at the end of the article, before comments start, and repeated again at the end of comments. I find them much more useful than next / previous links, which are usually not very related other than temporally.

    My total guess on this, Darren, is that they’re somehow conflating pages with hits … or something. Because there is no way that they are getting 11 page views per visitor.

  7. darren

    All rightee, I’ve added a ‘Related Posts’ widget to the right sidebar, replacing the ‘Recent Posts’ widget that was there. We’ll see in a couple of weeks if that keeps people on the site any longer.

  8. Light & Dark

    Darren, the tourism site I manage definitely has avg. pageviews/visit over 5. Closer to 7 if just organic traffic is considered. (PPC traffic for my type of site is generally accepted to generate lower engagement numbers.) We run 8 to 8.5 for traffic coming from certain countries, so source of the majority of the traffic can have a real impact too. Just another data point

    As far as the use of a panel-based metric, the only real use I see for them any more is to get an idea of traffic for sites where the site owners are unwilling to share their own Analytics stats. Or where they’re untrustworthy.

    And I think that may be where this case comes from – the paper is probably used to reporting visitor numbers from an “independently verifiable source” like they do with circulation. (As Robert points out.)

    Given how differently different tools capture and report data, I’m always very careful comparing analytics from one site to another, especially if using different tools. Hell, moving your Analytics code from the bottom of your pages to the top can easily skew your stats by over 10%.

    I had one client who’s log-based analytics was counting search spiders’ visits to his robots.txt file as visits and pageviews. Nice high pageviews/visit when the “visitor” is actually hitting every page on your site!

    I think your point about site architecture and pagination of articles is very telling. If their articles are split to 4 or 5 pages each, that 11 pageviews/visit doesn’t look nearly as impressive, but certainly allows for more ad real estate.

    As I keep telling my team and clients – web stats are great as a barometer, but unreliable as a thermometer.

    Be very curious to see if you hear anything back.

    (Congrats on your Globe & Mail mention too, by the way!)

    Paul.

  9. dave

    I’m not so sure that their numbers are all that questionable. I agree with your general analytics observations but this might be a little unique.

    I think the important thing to keep in mind is that the Sun is a local newspaper site. A significant part of the readership is going to be more accustomed to the newspaper format. I know a few people who visit their local newspaper site every day. This type of visitor would skew the numbers – if they read 3 stories a day, their page view count at the end of the month is going to be near 100. Even if you have 9 other unique visitors who only view a single page, that’s still going to give you an average of 10 page views per visitor.

    As for your related posts block, I’d consider placing that in the same area that you have the tags in full page view – at the bottom of the entries before the comments, or even embedded in the text. Not sure if this is easy to accomplish with WordPress.

  10. kgrandia

    An average of 11 pageviews does seem extremely high, I normally assume a 3x’s unique visitors for pageviews as a starting point. While maybe the Sun site is a little “stickier” than your average site, I would say that even that wouldn’t get them to an 11 pageview average.

    Having fumbled around the Sun’s poorly organized site, trying to find an article, I would suggest that this might be the reason – usually takes me about 5 clicks to find what I want.

  11. Kevin

    Advertisers/marketers want to know now many unique viewers, not page views. Unique viewers provide a look into the demographics (age, gender etc) that analytics can’t provide. This is important when advertising because you need to know if your “target audience” is actually viewing a web site.

    Analytics are based on 3rd third-party cookies and security settings on browsers often have this turned off. It’s also important to filter out crawlers/bots because they are often included in web site numbers. The ideal use of analytics is for analyzing behaviour while on the site.

    Most “reputable” websites use comScore or Neilson for the unique visitors. Btw: comScore annouced a measurement which incorporates panel and analytics. See http://www.comscore.com/Press_Events/Press_Releases/2009/5/comScore_Announced_Media_Metrix_360

    Lastly, most web sites report “total” hits, page views etc. The thing to remember about comScore, is that when they say “Total Canada” then this is Canadian unique visitors only to the web site. This could be the reason for the descrepency with the vancouversun.com. They may have not filtered out only canadian page views!

Comments are closed.