Nicolás López Coombs, Utrecht University
Sophie de Groot, Utrecht University
As a subset of the Digital Methods for Number Critique group, we focused on tracking trackers over history. Trackers, both on the front and back end, are present across the Internet, where they influence and record our behavior.
In order to explore this facet of Internet life, we chose the Slate.com website as our object of study. Other work being done in the Number Critique group looked at trackers on news websites, so this seemed a useful way to expand that corpus. Slate.com nonetheless stands apart from many other news outlets as it is only available online. It also has a longer history than many of its current popular peers, having been created in 1996 (Kinsley 2006) and remaining relevant as the 189th most visited site in the United States as of January 20, 2016 (Alexa 2016).
We decided to analyze the presence of trackers and widgets present on Slate.com over time, represented not only in the data we digitally captured, but in numerical representations and social or other widgets visible to the site visitor.
Initial Data Sets
For this project, we would need access to no less than one archived Slate.com page per month. Our main focus was the homepage, so by looking at the Internet Archive's Wayback Machine we found that that data was available as far back as 2005. After we began gathering data, however, we found little tracker activity occurring before 2008. Therefore our final dataset was comprised of the trackers and widgets on Slate.com’s homepage captured once per month for each year from 2008 to 2015.
For the visual analysis, our limited time meant that our data set was conducted yearly for 2011 to 2015 on both the homepage and one article page.
How does the presence of trackers and widgets change by month on the Slate.com homepage from 2008 to 2015?
How does the visual presence of widgets change by year on the Slate.com homepage and article pages from 2011 to 2015?
We began by using the Internet Archive Wayback Machine Link Ripper to put together a list of available pages for the Slate.com homepage. We then cleaned up the list by removing anything from before 2008 or after 2015. We further reduced the dataset to include only one instance per month (with a date as close to the 20th as possible).
We then inputted the url list into Tracker Tracker, which exported a list of which widgets, analytics, ads and trackers were present on those urls. From this list, we reduced the set to the items defined as “widgets” or “trackers” by Ghostery's library. From this set, we were able to make a comprehensive list of the widgets and trackers present on Slate.com from 2008 to 2015 according to our data sample.
We then put together a spreadsheet that listed the possible widgets and trackers then marked their appearance from month to month. We used Infogr.am to produce a visualization of the number of widgets and trackers available from month to month. Then we manually composed a GIF that combined that graph with a chart to identify which widgets and trackers were present in a given month.
(You’ll note some distortion noticeable in the visualization that we did not have time to correct but we feel it does not materially affect the representation.)
To better understand the phenomena present in our dataset, we performed a visual analysis of the Slate.com homepage to identify the visual presence of widgets. Due to time constraints, this was only possible for one instance per year from 2011 to 2015. We also performed the same visual analysis on yearly article page captures. We used the Grab Them All Firefox plug-in to capture these page snapshots, then put together videos the visually indicate where the widgets are present on the different pages.
We found that trackers first peaked in September 2010 at 3 before going back down to 1 in December 2011. They gradually increased again, eventually reaching the second peak of 6 in November 2015. Widgets reached their first peak of 7 in November 2011 before descending all the way to 2 by June 2014, yet they returned to a second peak of 7 in May 2015. May 2015 was also the combined peak of 12 for widgets and trackers together. As recently as January 2015, however, the combined number was only 5, down from the previous peak of 9 in May 2012.
Video Slate.com Homepage (2011-2015): screencap_homepage.mov
Video: Slate.com Article pages (2011-2015): screencap_articles.mov
Only one tracker (Omniture) remained present throughout our sample. The closest analog for widgets was Brightcove, which irregularly appeared from November 2009 to July 2015. Also of note is that the second rise of widgets was not a return of widgets that had previously been removed, but rather the appearance of new widgets (trueAnthem, Tinypass, Livefyre and Janrain) except for Facebook Connect.
By visually analyzing the pages and connecting that to our other data, we found that from 2011 to 2013 there was a visual focus on numbers/widgets in the interface with a corresponding peak on the backend with widgets/trackers. From 2014 to 2015, the style shifted to a more minimalist interface with the numbers in the background. At this time, trackers also dropped but repeaked in 2015.
We had expected to find a gradual and more or less steady increase of both widgets and trackers over time. It seemed logical that the relative invisiblity of trackers means website owners have little incentive to reduce them. However, we found that there were significant variations in the number of trackers, notably a drop and rise between 2013 and 2015. As for widgets, we expected that increases in social networks (Facebook, LinkedIn
, Instagram) would prompt increases in widgets present on Slate.com. Yet widgets as well dropped and rose between 2013 and 2015. Our visual analysis of Slate.com provides some explanation, as at this time the website opted for a cleaner look with fewer distracting items and reducing the widgets to small icons at the bottom. Nonetheless, the research as it stands now poses more question than it answers. Most of all it urges the researcher to consider what other factors may be involved in the adoption and implementation of trackers and widgets on a webpage.
Our findings suggest that there are variations in the use of trackers and widgets more complex than we had envisioned. Though we surmise that trends in page design are an important factor, we do not have enough data to be certain of the reasons.
We would recommend further research to do a more detailed analysis of the visual aspects of the website (month-by-month) as well as interviews with Slate.com to find out what intentional motives they had (if any) for increasing and reducing trackers and widgets or if they were actually prompted not by Slate.com but by tracker and widget companies. Further research into the continuing use of these same trackers and widgets elsewhere would also provide useful information (for example, certain trackers may have been discontinued.) A more developed visualization of the presence of individual trackers over time would also provide potential insights.
- Alexa. 2016. “Slate.com Site Overview.” Alexa. January 20.
- Digital Methods Initiative. n.d. Internet Archive Wayback Machine Link Ripper. Amsterdam. https://tools.digitalmethods.net/beta/internetArchiveWaybackMachineLinkRipper/.
- Digital Methods Initiative. n.d. Tracker Tracker. Amsterdam. https://tools.digitalmethods.net/beta/trackerTracker/.
- Ghostery, Inc. 2015. Ghostery (version 5.4.10). https://addons.mozilla.org/en-US/firefox/addon/ghostery/.
- Infogram Software, Inc. 2015. “Infogr.am.” https://infogr.am/.
- Kinsley, Michael. 2006. “My History of Slate.” Slate, June 18. http://www.slate.com/articles/news_and_politics/slates_10th_anniversary/2006/06/my_history_of_slate.html.
- rzelazko. 2014. Grab Them All (screenshots of Many Pages). https://addons.mozilla.org/en-US/firefox/addon/grab-them-all/.