Analytics/Key metrics explained in laymen's terms

From mediawiki.org

Based on 2008 mail, needs to be actualized.

Site hits, page views, visits and unique visitors are all different but related concepts.

Here is an example:

Joanna browses Wikipedia and looks at page http://en.wikipedia.org/wiki/Rembrandt_van_Rijn This generates at least 77 *server hits*:

  • 1 the html text
  • +/-55 pictures: (50 related to the article, 5 logo's, one of those in the address bar)
  • 10 stylesheet files (layout info)
  • 11 javascript files (browser macro's)

Each file is a separate request to the server and counts as a hit.

She then clicks http://en.wikipedia.org/wiki/Dutch_Golden_Age Again so many page hits, and a second page view

Now Joanna moves on to http://en.wikipedia.org/wiki/Eighty_Years%27_War A third page view, but two hundred server hits already.

Joanna decides she knows enough. She shuts down the computer.

After half an hour she decides to rent a movie, and goes back to Wikipedia to check the name of the latest Oscar winning director. This is registered as her second visit to the site today (because of the gap between this and her previous page view exceeds a predefined time limit, say 15 minutes). Note that the arbitrary time limit makes this count less unifom between web metrics collectors. Some may count this as the fourth page view in her first visit.

Still all of these page views are ideally attributed to one unique visitor, namely Joanna (known to the site owner by IP address). However here things start to become complicated. This last metric is much harder to determine.

First scenario: we assume Joanna's companion needs to check something as well, takes hold of her keyboard, and browses some articles on Wikipedia about Hawai in quick succession. Still *Joanna's* second visit (?!). Same IP address, so we see one unique visitor.

Second scenario: no companion, just Joanna. This time she returns to Wikipedia more than 24 hours later. Depending on our definition of a unique visitor (yet to be determined) we might attribute these and earlier page requests to two unique visitors. We might chose a threshold of 48 hours, or 7 days, but usually this threshold is rather short, as so many computers are shared between people especially in the office or in public places.

Third scenario: again no companion. But in this scenario Joanna's provider gives her a new IP address every time she starts the computer, to safeguard her identity (or to reuse a limited set of addresses over a large group of clients). Now her requests are registered as being from two unique visitors despite the short time interval (after all they come from two IP addresses, because of the computer restart), unless …

…unless the web site (Wikipedia) recognizes that it is still Joanna's PC (and presumably still her), because the same cookie file (site specific local settings file stored on your PC) is found in the second session.

So we should use cookie files to improve our metrics right? Not so sure, it intrudes on peoples' privacy, and some users make a point of deleting cookies automatically after each session. So to recap, there is ample room for ambiguity.


Wikimedia does not measure visits yet, let alone unique visitors.

For now we rely on comScore at least for unique visitors. They try hard to be consistent in this: There is white paper at http://www.comscore.com/method/tech.asp [dead link]

(alternative reference)  that explains their method. comScore claims "cookie deletion can lead to a 2.5x overstatement of unique visitors"

Wikimedia does measure server hits and page views (actually page requests because people may ask for a non existing page and still be counted, but this is nitpicking)

Reference:

The following Wikipedia definitions can help to make things clear:

Page hits: http://en.wikipedia.org/wiki/Hit_(internet)

Page views: http://en.wikipedia.org/wiki/Page_view

These and more terms also at: (but this is may be a bit overwhelming, so many terms and nuances) http://en.wikipedia.org/wiki/Web_analytics#Key_definitions

A more succint explanation at: http://www.opentracker.net/en/articles/hits-visitors-pageviews.jsp

Note some people use /page hits/ as /page views/, to add to the confusion. Some small web sites reputedly benefit from this confusion to inflate their numbers, by publishing server hits and give the implicit impression these are page views.