We have a great crowdsourced transcriptions site1 that has received a lot of publicity lately which corresponds to some significant usage peaks. This image from our Google Analytics shows use from 14 September 2014–13 November 2014.
The first peak (3,654) on 2 October corresponds to a BuzzFeed post on 1 October. I am uncertain about the 26 October small peak of 507. It appears to be from Facebook. The 5 November peak of 911 is due to beatonna’s “Hark! A Vagrant” Tumblr post and the 11 November peak of 1,923 is thanks to John Green’s Tumblr post. (The 16 October 2014 NBC news article did not bump our usage.) However, looking at Google Analytics to see the source for our traffic during this same period shows surprisingly low numbers from both beatonna.tumblr.com and fishingboatproceeds.tumblr.com (the two blogs which we know caused the spikes). In addition, the number of direct users also seems high.
In fact, if we were to just look at our social usage (8,185 sessions of 23,313, or 35% of all our usage), we would not think Tumblr was that important in directing people to our content. And yet we know those spikes correspond to our site appearing on two extremely popular Tumblr blogs.
We thought perhaps using a Tumblr app on a mobile device would result in what appears to be a direct link. The mobile use of the site is fairly substantial.
However, looking specifically at November 5–12 (the time period of the two peaks presumably from Tumblr), there is actually a higher percentage (82.55%) of desktop usage than for the month as a whole.
I need to explain the dashboard for those of you that don’t use Tumblr. You can link to a Tumblr post directly, which works well for cross-posting to Twitter or Facebook (and is the way people who are not active on Tumblr typically see posts). These URLs can be seen in our source data above. However, if you have a Tumblr account and follow various blogs, you will typically see content in the Tumblr dashboard (http://www.tumblr.com/dashboard). This domain does not appear as a source in our Google Analytics data.
Thursday evening I conducted tests going from Tumblr to our digital collections. I am using our Iowa City Past Tumblr and our CONTENTdm collections for testing because I could more easily isolate use. These are the items I went to in our collection and where I started the use.
- http://digital.lib.uiowa.edu/cdm/ref/collection/ictcs/id/978 – coming from another person’s Tumblr dashboard
- http://digital.lib.uiowa.edu/cdm/ref/collection/ictcs/id/9634 – using the Tumblr app on my Android phone, on a reblog of one of our items
- http://digital.lib.uiowa.edu/cdm/ref/collection/ictcs/id/15296 – using the Tumblr app on my Android tablet
- http://digital.lib.uiowa.edu/cdm/ref/collection/ictcs/id/18031 – coming from the Tumblr blog’s administrator’s dashboard
I expected the use from the Tumblr apps to be counted as direct in Google Analytics. Given our lack of Tumblr data for our other site, I thought the Tumblr dashboard might also be counted as direct. What I found simply puzzled me.
The following shows the Google Analytics data for 13 November 2014 of the specific landing pages I tested. The first item (which shows use from Tumblr) was not part of my specific testing. The other four lines were all tests coming from Tumblr.
|Landing Page||Source / Medium||Sessions||Pages / Session||Hour|
|/cdm/ref/collection/ictcs/id/978||iowacitypast.tumblr.com / referral||1||2.00||16|
|/cdm/ref/collection/ictcs/id/978||l.facebook.com / referral||1||7.00||17|
|/cdm/ref/collection/ictcs/id/15296||iowacitypast.tumblr.com / referral||1||5.00||18|
|/cdm/ref/collection/ictcs/id/18031||pinterest.com / referral||1||5.00||18|
|/cdm/ref/collection/ictcs/id/9634||(direct) / (none)||1||5.00||18|
My Android tablet test using the Tumblr app is the only one that correctly shows I came from Tumblr. So, not only is our Tumblr use under reported, in this case the Facebook and Pinterest use is over counted!
This is the social data for the ictcs collection on 13 November 2014.
Specifically, the Tumblr use is:
|Shared URL||Hour||Sessions||Pageviews||Source / Medium|
|digital.lib.uiowa.edu/cdm/ref/collection/ictcs/id/13545||14||1(14.29%)||6(15.00%)||tweitzelposts.tumblr.com / referral|
|digital.lib.uiowa.edu/cdm/search/collection/ictcs/searchterm/Photocollages%20(Photographic%20compositions)/field/typa/mode/exact/conn/and/display/200/order/title/ad/asc||15||1(14.29%)||1(2.50%)||iowacitypast.tumblr.com / referral|
|digital.lib.uiowa.edu/cdm/ref/collection/ictcs/id/978||16||1(14.29%)||2(5.00%)||iowacitypast.tumblr.com / referral|
|digital.lib.uiowa.edu/cdm/search/collection/wwiid||17||1(14.29%)||2(5.00%)||iowacitypast.tumblr.com / referral|
|digital.lib.uiowa.edu/cdm/ref/collection/ictcs/id/15296||18||1(14.29%)||5(12.50%)||iowacitypast.tumblr.com / referral|
|digital.lib.uiowa.edu/ictcs/||18||1(14.29%)||22(55.00%)||iowacitypast.tumblr.com / referral|
|digital.lib.uiowa.edu/cdm/search/collection/ictcs/searchterm/Military/field/all/mode/all/conn/and/order/title||19||1(14.29%)||2(5.00%)||iowacitypast.tumblr.com / referral|
We have Google Analytics code in our Tumblr blog, but we do not receive data from people viewing our content through the Tumblr dashboard. We only show 11 sessions for our Tumblr blog on 13 November, two of which came from Twitter, three from Google, six direct and none using mobile devices. Only one of the pages I used for a test appears as a landing page (which means the use is attributable to someone else). Essentially, this data is useful to see how people who are not regular Tumblr users see your content, but it misses the active Tumblr community.
Edited to add:
It occurred to me that the situation is akin to Google Analytics only showing Facebook as the source when people link from a public Facebook page and not when they link to content they find in their news feed.
1 I should have originally made it more clear that the “we” in the first word of the post refers to my fantastic colleagues and that I have not been part of the project, other than to cheer them on.