Worried about page views dying? Don’t be.

I found myself thinking, “Are we really having this conversation?” today after reading Steve Rubel’s post today on “What will replace the allmighty page view?” where Mr. Rubel commented:

“The page view is on life support. It fails to capture all of the myriad of ways consumers engage in online activities without ever leaving a web page.“

Okaaaaaaaaaaaaay.

I suppose Steve is coming at this from a different perspective than anyone who works in the web analytics field, more-or-less looking at page views as a basis for comparing the relative value of one advertising opportunity to another. If that’s the case then yeah, page views are becoming increasingly limited in their utility.

But damn, as a web analytics professional, doesn’t all this talk about page views going the way of the Dodo bird just make your stomach feel all funny? Like, you know there are problems with the metric, but A) when compared to the other problems web site operators have vis-a-vis counting (cookie deletion, cookie blocking, poor implementations, caching, robots, lack of understanding, lack of interest) and B) when put in the context of the number of sites that still rely on good old fashioned HTML, don’t these proclamations seem a bit premature?

Is it just me? Maybe it’s just me …

Anyway, we can stop worrying about dying pages and dying page views now since the answer has been with us the whole time. It’s not unique visitors … too many problems with how unique visitors are counted, what with cookie deletion and some of the inaccuracies ascribed to panel-based services. It’s not time spent on site … the problems with this metric as the basis of comparison are many (connection speed, amount of content, quality of content, bathroom breaks, etc.)

It’s sessions.

Yep, sessions. Good old “start ‘em with the first page view and stop ‘em after 30 minutes of inactivity” sessions. And while they don’t necessarily solve the problem of how many impressions a site can serve (you need old fashioned web analytics for that), they provide a stable basis for comparison across sites:

Sessions are defined by a widely-used and widely-understood standard, the 30 minute timeout between subsequent page views. Heck, in the web analytics industry, it’s pretty much the only standard we have …
Sessions are counted once and only once when a visitor goes to a web site in a single web browser and are thusly not subject to inflation due to crappy web design or RIAs. No more complaints about MySpace!
Sessions are time independent, except for the session timeout. You can click away all day and you’ll still only count one session, unless you walk away for 30 minutes and one second …
Sessions mitigate out issues associated with error pages and the such, because again, the number of pages viewed is irrelevant after the visitor views the first page. Again, no more complaints about MySpace …
Sessions are not affected by cookie deletion and are not always affected by cookie blocking. Whoopie! We can stop bugging out about cookie deletion …
Sessions are not affected by users visiting sites from multiple web browsers, since regardless of location (home, work, etc.) the session is counted. Hurrah! No more massive over-counting of unique visitors during Fantasy Football season …
Sessions can be counted even when the visitor is not on your web site, depending on what tracking technology you’re using and how it’s deployed. For example, a session can be counted when someone reads a post in their RSS reader …
Sessions are easily tied back to relevant referring sources, such as advertising units, RSS feeds, search terms, etc. Yippie! Not only do we get more accurate counts, we know from where the sessions are originating …

Yep, good old fashioned sessions … who’da thunk it? You can call them “visits” if you’d like!

What’s better is that the reporting networks should just as easily be able to report on sessions as they do unique visitors. If they can report on “unique searches” and “time per person” and “page views” and all that, nothing should theoretically stop them from using “sessions” as the basis for reporting.

Clint Ivy pointed out to me that Hitwise uses sessions as the basis for their reporting platform, only they report however on percent market share and not the actual number of sessions which is almost certainly what advertisers would prefer to see. Neither of us were sure why they don’t give raw session counts, do any of you?

Just think of all the problems we can solve by using sessions to compare the popularity of web sites! No more complaints about newspaper sites reporting more unique visitors than live in the entire state. No more complaints about huge differences in reported numbers ascribed to cookie deletion. No more freaking out about inanimate objects dying …

What do you think? Am I crazy? Is it just me? As always, I welcome your comments.

Posted Thursday, February 15th, 2007 | 26 responses | Share, Save or Email

Ian

Hi Eric,

The one clarification I feel really needs to be made every time this conversation rears its head is the context of the claim. Or more specifically what use of the Page View metric is really dying.

For the most part, the story of the Page View’s demise is told from the perspective of Marketing. The claim is largely based on the use of the metric for comparative reporting on how one site measures against another.

For this purpose I don’t disagree that the Page View isn’t (and probably never has been) the best one to use. I also agree with all of your statements for why sessions are the way to go in the context of this use.

But what about other uses? The application of Pages Views when used only to measure a real Page (the complete refresh of the content in the browser) is still valuable to analysis. Especially when used in comparison to something else that is perhaps the refresh of something smaller within the page.

There is also the fact that not everything on the web will be ajaxified/flexified. Pages will continue to exist for some time to come and in some uses, such as site optimization, understanding the impact of more or less of them to the user experience is very much important to the context.

This is why I whole heartedly agree we should stop worrying about dying page views and instead focus our attention on what all of the other stuff is.

-Ian
S.Hamel

I totally agree with you Eric. We had to explain why “hits” were no good, we now have to explain why “unique visitors” are not always that good either… but good old visits remains visits…

The only thing is… I prefer 20 min session timeout instead of 30, or maybe 15? Or why not 45? *grin* :)
http://www.webanalyticsdemystified.com Eric

Ian: Yeah, that is a pretty big problem. Some people have covered these proclamations as if “page views” were on the way out for we web analytics practitioners which is obviously and clearly not the case. I love what you’re doing with your framework (I’ll blog it later!) but I suspect the dominant event in your model will be the “page view” for years to come.

Stephane: My friend, why would you take the ONE THING we can all agree on and mess with it? ;-) Someone (Clint?!) should have a poll that asks people what session timeout they use is, I bet we’d all say “30 mins” don’t you think?

Thanks to both of you guys for your comments!
Anonymous

…I bet we’d all say “30 mins” don’t you think?

Actually, I bet a lot of people would say “what’s a session timeout?” ;)
June Li

Hi Eric,
About sessions – isn’t tabbed browsing a complication?

The Page View Metric is not going to die. Yes, it’s declining in importance because it doesn’t measure every website activity appropriately, but that isn’t really new, is it? Ignoring measuring some “events” has always been a problem (PDFs, flash, email clickthroughs and other conversion events were often ignored because they’re a bit of a hassle to accommodate.) Ajax, videos, podcasts, etc just make this gap more evident because of their prevalence.

Those most concerned about page views are folks whose compensation depends not on conversion, but on page views increasing year over year (i.e. this is an organizational problem, and as Ian points out, mostly in Marketing).

Take a look at history. When television was launched, radio’s demise was widely predicted.

And not long ago, the 30-second TV ad spot was said to be dead — hmm, is the cost of a Super Bowl ad declining?

Page view metrics won’t die. Their use will evolve. They’ll remain useful when relevant, typically in a targeted, narrower KPI than overall site performance.

Great post!
http://www.webanalyticsdemystified.com Eric

June: Tabbed browsing … well, I haven’t done an exhaustive review but let’s think about this. If someone has multiple tabs open and they “tab away” from my site for more than 30 minutes, isn’t it safe to say that they have disengaged with my site and thusly when they return they’ll need to re-orient, etc. at which time it’s probably okay to count another visit?

Put another way, we have to be able to draw a line somewhere. To me, 30 minutes feels a lot more reasonable than using a metric that is becoming less relevant for comparison’s sake (the page view) or another subject to discrepancies due to normal user behaviors like cookie deletion (the unique visitor).

I loved this comment from you: “Page view metrics won’t die. Their use will evolve.” This is perhaps the most true thing I’ve read yet in this interesting conversation. Thanks for your comments!
http://www.linkedin.com/in/danielwaisberg Daniel Waisberg

Hi Eric,

regarding your Hitwise question: “Neither of us were sure why they don’t give raw session counts, do any of you?”

I believe they don’t show raw numbers because they don’t want people to know what is the exact size of their sample. It might give users a reason to doubt their usefulness. For example, if I have 1,000 visits on my site and they say I have 5, I will think it does not represent reality (even though it might be statistically significant). Now, if they say I have 10% market share, I will keep on paying with pleasure without worrying about statistics ;-)

That’s my theory…
Daniel Waisberg
Judah

Hola ETP,

Page views aren’t going anywhere.

It’s all a matter of perspective based on your measurement goals. For calc’ing engagement, page views are suboptimal, but for counting “beans” based on cp* models, they have financial utility.

Try telling an editor that the page view doesn’t matter.

PV’s are a type of Web 3.14159 major event, and provide context for minor events.

Judah
Anil Batra

First of all I want to point me to an article on my blog that explains why I am not a big fan of “Time Spent on Site”.

I also agree that 30 mins is the standard so let’s all agree and move forward.

Page views were not the right metrics to compare web properties to begin with. Why? Because they can be manipulated very easily. Say it t takes 2+ pages on site A to do anything compared to 1 page on site B, is site A really doing better than site B? Additionally you can split your content in as many pages as you want, there is no min standard page size, thus inflating page views.)

As far as using unique users and sessions, I am not so sure if sessions make sense.
Yes, Sessions are relatively stable metrics compared to Unique Users.
I agree that Unique users have issues but those issues affect every web property, most likely in similar fashion. For example, if I delete my cookies, most likely I will delete for both myspace and yahoo. So why can’t we accept this fact and use Unique Users to compare web properties?

Here is an example to make my point:
I go to myspace and read 2 pages in 1 min, wait 31 mins and then go back and read 2 more pages in 1 min. So here is what the web analytics reports will look like

2 sessions (visits)
1 unique users
4 page views
2 mins.

Now I go to yahoo spend 2 mins reading 4 pages in 1 session. Here is what the web analytics report will look like
1 session
1 unique users
4 page views
2 mins

What about the following scenario

Which property is number 1? Aren’t they both the same? If you use Session myspace appear to be number 1. But if you look at Unique Users and rest everything too, they both are equal.

What about the following scenario

I go to myspace and read 2 pages in 1 min, wait 31 mins and then go back and read 2 more pages in 1 min then come back after 2 hours and read 2 more pages for 1 min. So here is what the web analytics report will look like

3 sessions
1 unique users
6 page views
3 mins.

Now I go to yahoo spend 2 mins reading 4 pages in 1 session. My Friend goes and reads 2 pages in 1 mins in one session. Here is what report will look like

2 sessions
2 unique users
6 page views
3 mins

Which property is number 1? Session will say myspace, while unique users show yahoo.

Maybe it is time to find a new metrics combining some (or all) of these above metrics.
Thoughts?
http://blog.instantcognition.com Clint

Daniel,
If memory serves, Hitwise claims a panel of around 10 million for the US, so I don’t think it might be a suspect panel…

-Clint
http://www.webanalyticsdemystified.com Eric

Daniel and Clint: Thanks for the info about Hitwise. I asked a contact at comScore who indicated that they could report on sessions so I might push them on that a little more.

Judah: I have no idea what you’re talking about. What is “Web 3.14159″? Do I need to come to your and Ian’s Emetrics presentation to find out?

Anil: I’m with you on “time spent” … I very much prefer looking at distributions of time spent by visitor than any average. However, I disagree in your assessment that since cookies are deleted or not, why can’t we just move beyond that issue?

The problem is that unique visitor counts become contentious when you try and compare data in your web analytics system to a panel-based service like comScore or Nielsen. Both systems refer to “users” and “unique visitors” but they’re calculated in dramatically different fashions and have dramatically different sources of error. I guess my thesis is that by using a well- and widely-understood metric such as “sessions” we could move beyond the issues of definition and margin of error.

Regarding your example scenarios, I think you’re making my point for me. If we could simply decide that the “right” metric to assess the popularity of a property, at least when we’re trying to compare properties, was sessions then we wouldn’t have to wonder which property was “number one” … it would be the site that generated the largest number of sessions regardless of the number of visitors, page views, or events.

Anyway, I’m super impressed with all the commentary you folks are providing. This is a hugely important issue but one that has the potential to change the way we think about our sites. When someone says, “I get 1,000,000 unique visitors to my site” we wouldn’t have to wonder whether they were talking about the sum of Daily Unique Visitors (ack!), visitors derived from log files that haven’t had robots and spiders removed, the size of the panel or panel service, etc. We could just say, “Whatever, how many sessions did you generate last month?” and be done.
Judah

June: Your excellent point on folks whose compensation are based on page views could also apply to business models that charge by or are judged on page view volume. If I spent time engaging in events on the same page, how does that affect Internet business, in the WA industry and beyond?

After all, we all know McD’s sells a lot of hamburgers. Doesn’t mean their burgers are good. :)

Eric: Yes, attendance is mandatory and resistance is futile.

I promise to talk about mushroom burgers. Mushroom veggie burgers.
http://www.webanalyticsdemystified.com Eric

Judah: Unfortunately your and Ian’s presentation is at the EXACT SAME TIME as my good friend Dylan Lewis so it’s very likely I’ll miss your talk. The upside is that you can make hippie jokes about me behind my back if you like. ;-)
http://www.jacqueswarren.com Jacques Warren

Hi Eric,

Nice, hey, to let the steam off, once in a while! Ah! Finally voicing all the frustration we face everyday with all those *funny* numbers, the innacuracy, the inconsistancy… To want to be perfect in this very imperfect world…

Well, back to sessions, I’m all for it (I think Jim Novo has been saying exactly that for 2 years now). But aren’t we letting ourselves give in to temptation? Are we really ready to let go of visitors? Of all the relationship metrics/analysis? Of engagement indexes?

I’m afraid we’re stuck with them: people. You’ll always want to know if you’re dealing with a million of them or a hundred thousand; do they come once or ten times a month? I know my clients will for still a long time.

Ok, then, here’s a new solid metric: Number of email addresses acquired/(Time period). Or phone number, or whatever it takes to start engaging in a dialog. Anyhting that could help us measure a *precise* response to an offer, to determine RFM, just plain to know who to mail again!

Whoa! Now it’s me who’s letting some steam off. I guess it’s because of the harsh Montreal winter…
http://www.webanalyticsdemystified.com Eric

Jacques: Yeah, it appears I have definitely struck a chord with folks. I think the takeaway message here is that we need to use the “right tool for the job” and that collectively we need to do a better job of messaging what the right tool is. Here are some examples:

If you want to compare two or more web sites, use sessions because of the reasons I outlined in my original post.

If you’re interested in the number of people coming to one web site (presumably yours), use de-duplicated unique visitors but be mindful of cookie deletion.

If you’re interested in the activity of people on your web site, and if you have a “Web 1.0″ web site, use page views but be mindful of issues like code coverage, proxies, robots, etc.

If you’re interested in the activity of people on your web site, and if you have a “Web 2.0″ web site built around RIAs, etc., use some form of event model (perhaps see Ian Houston’s posts at Visioactive for more on this.)

Does that make sense? I’ll try to better summarize in another post based on all the ** excellent ** feedback you folks have given me but thanks again for your comments and your time.
Pingback: Marketing Productivity Blog » Blog Archive » More Trouble for Unique Visitors
Pingback: Edmonton Web Design, Edmonton SEO,Bed and Breakfast Web Design,Restaurant Web Design,Alberta Web Designer,Edmonton Web Development,Edmonton Website Designer,Real Estate Web Design,Edmonton Web Site Design » All about SESSIONS
Alex Destino

Great article, Eric. I have forwarded it to a few people in my company as I have been making the case for sessions for a while now. I do think it should be said that sessions is not without its issues, but it does have less issues than visitors and page views. A session is a collection of page views for a single visitor within a certain time period. The fact that the time period of the session is relatively small compared to your reporting period mitigates the error caused by cookie deletion…but it still obviously exists. If a visitor deletes cookies and then continues a session, not only is the visitor count inflated, but so is the session count. From my experience this session error is about 1/5 the size of the visitor error. So if there is a 7% inflation of visitors, there is about a 1.4% inflation of sessions. I can measure this due to the fact that my site requires authentication and I can compare authenicated visitors and sessions to normal visitors and sessions.

Another point I wanted to make is that the authenticated session is the most accurate metric out there for the same reasons that you stated, but obviously not too many sites require authenication and thus do not have the luxury of using this metric.
eric

Alex: Awesome comment, thanks! I kind of doubt there are many sessions in which a visitor deletes cookies mid-session but who knows, stranger things have been reported. I agree about authenticated sessions but you’re right, too few sites have this option available to them. And, it doesn’t really give us any better metric to compare across multiple sites.
Kristin

Hi Eric,

today, I had a quite interesting conversation with one of my sales colleagues who he explained to me why unique visitors are so important and why visits cannot really replace that measure: unique visitors (for the whole website or just a specific part) are *the* figure to tell advertisers and customers about your reach (=how many different users are coming to your website). The number of visits can not really give that information.

Example (let’s ignore the problem of measuring unique visitors for a moment): A user may visit a news site every day (-> 30 visits per month) but visit a more special site only one a week (-> 4 visits per month). Nevertheless, it’s one unique visitor for the news site and one unique visitor for the special site.

So I suppose we will stick to the unique visitor for still some time (besides other figures like visits, visits per unique visitor, page impressions etc., of course).

.Kristin
eric

Kristin: Absolutely agree with you, but I don’t think you can actually “ignore the problem of measuring unique visitors” as you put it. What if your unique visitor who visits the news site every day deleted his or her cookies every evening? Then you’d be reporting 30 true visits but also 30 unique visitors where you only really had one?

The reason I advocate for sessions is that, for the most part, sessions is protected from issues associated with cookie deletion and the changing definition of page view.

Don’t get me wrong, I know that advertisers ** want ** to know about unique visitors, I’m just saying that given issues of inaccuracy and dramatically different strategies for measuring UVs (cookies, panels, etc.) that perhaps energy is better spent educating advertisers instead of evangelizing for UVs in some situations. Especially when advertisers want UV counts that they can compare across multiple sites (the real crux of this problem), these different strategies are less robust than most people would like.
Pingback: links for 2007-03-20 | The Marketing Technology Blog
Pingback: Best of Feeds - 27 links - blogging, google, tips, development, alist, howto « //engtech
Marco

Hello everybody… I’m quite a newbie in web analytics, but maybe one thought concerning unique visitors: IMHO the interesting point is not to know that a visitor is unique (i’d be still happy if one unique visitor made 100 visits with a purchase each…) The most interesting and challenging thing is to recognize him, and learn why he’s coming back and how i could make him come back again, if i wanted to…
http://www.webanalyticsdemystified.com eric

Marco: It sounds like you’re trying to understand how visitors are engaged on your site. If this is the case, you might want to check out the posts I have done on visitor engagement. Have a look at:

http://blog.webanalyticsdemystified.com/weblog/category/engagement/

Thanks for commenting!
Pingback: Web Analytics Demystified » Blog Archive » On NetRatings and time spent on site