Tynt, a whole new set of Web metrics

Web Applications Alert By Mark Gibbs, Network World
December 22, 2009 12:08 AM ET

Mark Gibbs' Web site tips, plus network applications news headlines

Just over 10 years ago I had a good idea and thus was NetRatings born (the company went public in 2000 and eventually became a wholly owned subsidiary of The Nielsen Company). Ever since then I've been fascinated by techniques for measuring what people do online.

So it was that when I heard about what a relatively new company, Tynt, was doing in the Web metrics field I was intrigued. What I found was a novel and very clever approach to audience engagement measurement that adds a whole new insight into how your Web content is viewed and used.

To explain what Tynt does, consider the following:

---------------------------------------------------------

Related Content

Recipe for White Christmas:

Combine in a bowl: 1 cup of powdered milk, 1 cup of icing sugar, 1 cup of rice bubbles, 1 cup of coconut and 1 cup of mixed dried fruit. Melt 250 grams copha and mix with the dry ingredients. Press into a small slab tin and allow to set. Cut into fingers.

---------------------------------------------------------

I had copied the text from "Recipe for …" to "Cut into fingers." from the Web site Aussie Cooking. When I pasted the content into this document, lo and behold, the text from "Read more …" onwards was also on the clipboard, appended to my selection.

Tynt does this through a snippet of JavaScript added to your Web pages. This code detects when a copy action on text or graphics is performed and does two things: The copy event is logged by Tynt and a link to the page that the copy was made from is appended to the copied content.

Tynt also parses the copied content and from research the company has done they have discovered that copied names are very often used as search terms while names in things such as mailto: URLs usually aren't. These insights along with other heuristics allows Tynt to distinguish search-related copying with what they claim is about 80% accuracy.

Now it turns out that people copy Web page content much more frequently than you might expect: According to Tynt's metrics copying of Web page content is in the range of 2% to 6% of all page views and they currently monitor some 7.5 billion page loads per month.

The extent to which content is copied in terms of both frequency and the amount of content copied is dependent on the type of site. Thus, blog postings will tend to have 100 to 500 words copied while gossip site tend to have tens of words copies. Also, recipe sites usually garner long copies, sports sites in general get lots of images copied. Another interesting observation is that local sports news tends to get long copies while copying from national sports news tends to be short.

Another interesting metric that detecting and analyzing copying tells you are which Web site pages are truly interesting. Look at normal Web server log data and you might find that a Web page was loaded 15 seconds after the home page and then a third page loaded two minutes later that was viewed for 16 seconds. Most analyses would conclude that the second Web page was the most important but knowing that the third page was copied from tells you that despite the user's shorter page 'dwell' time that content was really important to them. Moreover, by knowing what they copied their interests can be discovered.

Mark Gibbs is a consultant, author, journalist, columnist and blogger.