by • 6 Mar, 2011 • TechnologyComments (9)3632

Does a semantic network improve the quality of museum website visits? Some stats from the INNL network

Visitor Pattern
Photo by Dan Brickley on Flickr (CC BY-NC-SA)

It’s been more than three months since we at the Museum of National History launched our new website. And, apart from the usual glitches, it pretty much does the trick we expected it to do. Besides the website, we also launched the INNL network, a semantic network of historical websites. At the moment it connects over 200,000 stories, photos, people and events from different websites in a – hopefully – meaningful way.

The INNL network is built on the premises that providing your data in a meaningful way to a larger network of websites provides 1) more context for your online information and 2) a wider reach for your information. For a website owner this should mean more high-quality visits and less hours of editing. (I wrote a post about how the network works earlier.)

So, were these assumptions correct? Does a semantic network improve the quality of museum website visits? It’s a bit early to give a definite answer, but with 3 months of stats I feel I should be able to say something about the “semantic visitor” (the visitor coming to our website through the INNL network). The table below shows some indicators I believe are related to the quality of a visit:

Visit depth Visit duration Percentage non-flirts*
Normal visits 2.58 1:44 26.9
Semantic visits 2.51 2:41 34.8
Twitter visits 2.52 1:39 22.8

* Non-flirts is a term I borrowed from Avinash Kaushik and adjusted to mean visits with either depth equal to or greater than 3 or a duration equal to or greater than 2 minutes. These visits are more than merely ‘flirts’ (hits from Google or Twitter, quick checks to see if anything changed) but potentially interested visits. I’ve added Twitter to compare statistics.

The first thing that strikes me in the table above, is that although the normal and the semantic visitor both visit about the same number of pages, the semantic visitor spends almost 55% more time on the website. Among the many things this might mean, is that s/he actually spends more time reading/studying the information on the website.

The second thing is the difference in the percentage of non-flirts. Not surprisingly, semantic visitors flirt less with the website. Although not necessarily, this seems to support the earlier assumption that semantic visitors spend more time reading/studying the website. More interestingly it becomes, however,  when we have a closer look at the behaviour of non-flirts:

Visit depth Visit duration
Normal non-flirts 6.42 6:13
Semantic non-flirts 5.12 7:31

Semantic non-flirts visit less pages, and spend more time on the website than normal non-flirts. This might mean their visit is more focused and determined. I think this might be because they come from a related website in the INNL network and enter our website through a meaningful link in the context of the information they were already discovering.

Looking at the numbers above I feel it’s safe to say the semantic INNL network directs high-quality visitors to the participating websites. If we can increase the number of meaningful relations between the participating websites (and the number of participating websites, which we will), this should mean more visitors will discover more and better information in the network. Of course, it’s also possible more relations to other websites might mean the depth and duration of visits decreases, as visitors more easily find their way to other partners in the INNL network.

(All numbers are averages over the months December, January and February.)

Share on Facebook9Tweet about this on TwitterShare on LinkedIn0Email this to someone

Related Posts

  • Anonymous


    I’d recommend doing a qualitative look at the EXIT PAGES and see what sort of content the non-flirts are leaving on. Remember that the time spent on the exit page is not counted in total time on site! (which has big implications if the exit page is very content rich and/or contains media files)

  • Johan Oomen

    Hi Jasper, how do you determine “visit depth”? Also, hoe do you define the INNL network: can content from the network also be shared on external sites? (or is AnyMeta a precondition)?

  • Hi Johan, for visit depth I used the GA stat pages/visit. The numbers are the average, of course. There are technical preconditions to becoming part of the INNL network, and we’re now working on connecting other than anyMeta websites to the network. An “open website” should be enough to become a part of the network.

  • Thanks. A quick look at the exit pages shows most of these are “dead ends” on the website, not necessarily media-rich pages. One more thing I see (although there’s really too little data to say anything definite about this) is that some exit pages are rich in links to other INNL websites. Could be that these visitors are comfortable with clicking these links.

  • Johan Oomen

    Thanks for your swift reply …are you also going to publish the datasets as Linked Open Data?

  • I guess that’s not my department, but we’re usually definitely open to making as much as possible accessible and usable by others. Definitely something worth talking about and working on in the near future. We know where to find you;-)

  • Interesting stats Jasper! I would love to measure the percentage of actual ‘learning’ and increase of (longlesting) knowledge, but I guess that would mean mindcontrol… Johan, can’t we start with monitoring brain metadata? ;-)

  • Isn’t there a CATCH programme doing something like that? There should be… Facebook or Google must be working on that as well :-)

  • Pingback: The Museum of the Future » Growth of mobile visitors and the phones they use on