Are pageviews obsolete?
Wed, Sep 6 2006

By Jonathan Oxer, IVT Technical Director

One of the things I find really interesting about websites is examining the statistics that can be generated showing user activity and behaviour. Whenever you visit a website a heap of information may be recorded by the server, including:

* Which ISP you are using.
* Your geographic location down to suburb-level.
* What type of web browser you are using.
* Which pages you looked at.
* What site you visited prior to the current site.
* How big your computer screen is.

Using the Internet might feel anonymous but it definitely isn't! Whenever you go online you leave behind a huge digital trail that can be read by anyone with access to the information left behind on websites you visit.

That information can then be processed using a logfile analysis program to generate both general reports such as "How many different people visited our site last month?" as well as user-specific reports such as "What pages did user X visit on our site, and how long did they spend looking at each page?".

Trawling through traffic reports is a fascinating thing to do, but not many people actually have the specialist vocabulary required to make much sense out of them.

Most people still use the term "hit" as a unit of measure when talking about their website traffic. They have a vague idea that a "hit" somehow equates to visitors to a site, and that bigger numbers are better. However, as I explained in chapter 25 of How To Build A Website And Stay Sane, the term "hit" doesn't actually mean what many people think it does. In the field of web analytics, the term "hit" actually means "any request for a file from a web server". A web page consisting of the HTML page itself plus 10 images will constitute 11 "hits" when one person views the page. Of course, if that page had 100 images on it then there would be 101 "hits" recorded for one person viewing the page, so a graphic-heavy site will record far more hits than a simple site even if they attract the exact same amount of user activity. That makes raw hit figures pretty much useless when it comes to understanding site traffic.

So for about the last decade web analytics experts have been working hard to get everyone using the term "pageview" when discussing general site traffic levels. A "pageview" (also often called "page view") can be defined as a request to load a single page. It doesn't matter how many elements make up the page: it could be plain text or it could have 100 images on it. What matters is that it's one user loading one page. Unlike hits it's quite valid to compare page views between websites, and it forms the basis of many common metrics such as "pageviews per visit".

But now, just as we're starting to make headway and people are actually understanding that the term "hits" must die and that "pageviews" is probably what they actually mean, there's a big spanner being thrown in the works.

The term "pageviews" is in the process of becoming meaningless, and we don't have a term to replace it yet.

So what's going on? Why can't we keep using pageviews?

The problem is with the changing nature of the web. Back in the dim dark past the vast majority of websites were built on the "page" paradigm that the web was originally designed around. A website would consist of "pages" of information, and you'd click on links to move around from page to page. Each page would have distinct content and purpose, and by examining which pages your visitors access you can see which topics they're interested in and how they interact with your site. Old-style websites were like a "choose your own adventure" book that you could wander through in your own way.

But the page-based nature of the web was blown wide open with the widespread uptake of RSS, AJAX, and other related techniques, where interacting with a website is no longer like flipping through pages of a book but instead is much more free-form and includes many more "micro-interactions" as well as interactions that don't involve a web browser at all.

Think about the scenario of looking through an online store. A typical sequence of events might be to:

1) View a list of categories but not see what you want.
2) Enter a search term in a search box and click "search".
3) View a product list generated by the search box.
4) Click a product and view the product details.
5) Click "add to cart" and be taken to your shopping cart.
6) Click "continue shopping" to return to the product list.

That's a distinct series of pageviews to find and add an item to your cart.

Now consider the exact same scenario in a store that heavily utilises the latest AJAX techniques:

1) View a list of categories but not see what you want.
2) Start typing your search term into a search box. There's no "search" button to click, because as you type the page automatically updates with a "live search" list of matching products that becomes more specific the more you type.
3) See the product you want and click on it to view the product details.
4) Add the product to your shopping cart by clicking and dragging it onto a "shopping cart" icon in the corner of the screen. Your shopping cart updates automatically and displays the newly-added product in a little summary list without taking you to another page.

The second scenario may not involve *any* pageviews at all, other than loading the initial page. But there sure was a lot of activity going on! As each letter was typed into the search box the results were updated automatically with the browser talking back to the server to get a newly refined list. When the item is dropped on the shopping cart you don't go to a new page, but the browser sends an "add item to cart" event back to the server which responds with updated information such as the current list of items in the cart and the total order value. You may not see it, but the second scenario is much more "chatty" in terms of software activity: even though you stay on one page your browser and the server may be constantly sending little snippets of information backwards and forwards as you interact with the page.

In fact it's not really even very meaningful to call them "pages" anymore. What is a web page anyway? In the new world order it's not a distinct object or collection of information. Modern web interfaces are much more amorphous - more like using a piece of software and having it respond to our needs rather than moving between predefined chunks of information.

That's great from the user's point of view, but it's going to make tracking user activity extremely difficult indeed! A "pageview" is a useless measure when we're no longer dealing with pages.

So what are we going to use when measuring user activity? How will we report it? What vocabulary will we need when discussing site performance? These are the sorts of questions that web analytics experts are now scratching their heads over, and the answers aren't at all clear.

Personally I think we'll see much more focus on three things: visits, time, and events.

Visits are something we commonly measure already. It's an indication of how many times you have people's attention online, showing how many different people have come to your site and how many times each person comes. Metrics such as "unique visitors" and "visits per user per month" will become much more important when we can no longer measure pageviews.

Time spent on a site is something that we can measure very imperfectly at present but which is much easier to measure with techniques like AJAX. Not only will we care about how many times our site was visited, but metrics such as "visit duration" will also become more important (and accurate) than they were with the old web.

Events include the things people do on your site. An event could be "placed item in shopping cart", or "viewed product details", or "read news item", or "used search box". These are all the sorts of things that used to be distinguished as unique pageviews but which may now take place within the context of a single page. Tracking and reporting events is incredibly powerful and provides site owners with even more information than could be obtained from traditional page-view-based server logs. Rather than creating reports that show a sequence of cryptic page views which need to be interpreted to be understood, event logging makes it possible to report in a much more natural-language manner and with a finer level of detail.

The next version of SiteBuilder, SB4, has been in development for a while now and heavily uses AJAX techniques to provide users with a smoother and more interactive experience. So to give site owners the information they need to understand what users do on their site we've put a lot of work into building a next-generation user activity logging system that records the multitude of events that take place as users interact with sites running on SB4. The new logging system is designed around the concept of events and provides far more information than old-style, page-based logging systems.

So after many years of educating people to use the term "pageviews" when talking about website traffic it gives me both great pleasure and great pain to say that it's time for pageviews to die.

The new, more interactive, more personal web is here. It's time we updated our analytics methodologies to match.