Year |
Month and date (if available) |
Event type |
Event
|
2007 |
December 9 and 10 |
Statistics availability |
The pagecounts-raw dump starts being created by Domas Mituzas for the English Wikipedia starting late in the day on December 9, 2007.[1][2] Data from this dump is used by stats.grok.se, managed by User:Henrik, so stats.grok.se data goes back to December 10, 2007. Note that although the dumps are hourly, stats.grok.se reports data at a daily granularity.[3] You can look at the page on Barack Obama for instance.[4] Desktop-site sitewide analytics for the Wikimedia projects are available via the legacy API for data starting this day.[5]
|
2008 |
February 1 |
Statistics availability |
Data for non-English sites can be accessed in stats.grok.se starting with this date. It's not clear if pagecounts-raw started including other languages from this date onward, or if stats.grok.se simply started processing other languages starting with this date. You can compare the French Barack Obama page counts in January 2008[6] and February 2008.[7]
|
2008 |
July 13 to July 31 |
Statistics availability |
stats.grok.se does not have pageview data for this time period.[8] This appears to be due to issues with stats.grok.se; the underlying pagecounts-raw dataset is available for these days.[9]
|
2009 |
June 30 |
User experience |
Wikipedia mobile site launched, and is the default for "supported mobile devices" (note that automatic redirecting of traffic from tablets to the mobile site only occurred much later).[10] The pagecounts-raw dump, and therefore the stats.grok.se data, do not include pageviews on mobile web, which account for the majority of mobile device pageviews, from this point onward.[11]
|
2009 |
November 3 |
Statistics availability, partnerships |
Wikimedia Foundation and web analytics and digital media specialist comScore announce a partnership, in order to help the Wikimedia Foundation gather more data on metrics such as the number of unique users from different geographical regions.[12] Note that comScore does not track all traffic to Wikimedia servers, as it excludes people below age 15 (age is known for people who signup to the comScore panel, from which behavior is extrapolated to the entire population).
|
2009–2010 |
November 2009 – March 2010 |
Statistics availability |
"[C]ounts [during] this period are 10% to 25% too low, due to message loss (server overload)".[13][14] This only affects desktop traffic because it is before the time when the earliest mobile traffic data are available.
|
2010 |
June 1 |
Statistics availability |
Total mobile pageviews available starting around this time, but per-page dumps are only available from September 23, 2014.[15][13]
|
2011 |
February |
Search engine use experience |
"40% Of SEOs Say Farmer/Panda Hurt Their Sites In Google"; unclear if Wikipedia was one of these sites.[16]
|
2011 |
October 4 |
Mobile assistants |
Apple launches Siri, its voice assistant, for its new iPhone, the iPhone 4S.[17] Digital assistants such as Siri have been cited as a potential reason for decline in Wikipedia pageviews, since people can quickly look up information by asking Siri without having to visit Wikipedia pages.[18]
|
2012 |
|
Observed trend (Wikipedia) |
Wikimedia global monthly(?) uniques pretty much constant between 2012–2013.[19] Meanwhile monthly uniques in North America seem to be going down rapidly after mid-2013.[20]
|
2012 |
March 6 |
Discussion group |
The Wikimedia Analytics mailing list is inaugurated with a welcome message from Diederik van Liere.[21]
|
2012 |
May 16 |
Search engine use experience |
Google starts rolling out Knowledge Graph, used by Google internally to store semantic relationships between objects. Google now begins displaying supplemental information about objects related to search queries on the side.[22][23][24][25][26] Knowledge Graph integration into Google has been cited as one of the potential reasons for the decline in Wikipedia pageviews.[18]
|
2012 |
May 26 onward |
User experience |
Wikipedia Zero is launched (starting with Malaysia on May 26), in cooperation with some mobile carriers. The service allows people to read Wikipedia pages in a format similar to the mobile page, but without images loading, and the carriers do not charge them for the data use costs for browsing Wikipedia this way.[27] For a full timeline of launch by region, see Wikipedia Zero#History. Note that, like mobile pageviews, Wikipedia Zero pageviews is not part of the stats.grok.se pageview counts being published at that time.
|
2012–2013 |
December 2012 – January 2013 |
Statistics availability |
"[C]ounts for last two weeks of Dec and first week of Jan were broken (much bogus traffic). Data for these weeks have been extrapolated from unaffected days."[13]
|
2013 |
|
Observed trend (general) |
Home broadband use peaks in the US.[28]
|
2013 |
|
Observed trend (general) |
US adult internet usage at 84%, where it will stay for 2014 and 2015.[29]
|
2013 |
August 8 |
Observed trend (Wikipedia) |
Berit Schreck, Mirko Kampf, Jan W. Kantelhardt, and Holger Motzkau publish a paper (available on the ArXiV) on the use of various language Wikipedias. According to the abstract: "This report summarizes the results of a short-term student research project focused on the usage of Swedish Wikipedia. It is trying to answer the following question: To what extent (and why) do people from non-English language communities use the English Wikipedia instead of the one in their local language? Article access time series and article edit time series from major Wikipedias including Swedish Wikipedia are analyzed with various tools."[30]
|
2013 |
August 21 (actual release), August 1 (announcement) |
User experience |
Wikimedia Foundation turns on HTTPS for all logged-in users (announcement August 1).[31][32]
|
2014 |
March 1 |
Statistics availability |
A systematic log of problems starting this day with pagecounts-raw is available.[33]
|
2014 |
June 18 |
User experience |
Mobile site becomes the default for tablet devices.[34]
|
2014 |
September 5 |
Observed trend (Wikipedia) |
Wikimedia Foundation and comScore publish a report showing that uniques to Wikipedia measured by comScore have fallen, with the decline sharpest in North America (including the US), with desktop declining a lot and mobile mostly flat.[20]
|
2014 |
September 23 |
Statistics availability |
Wikimedia starts releasing a pagecounts-all-sites dump that includes desktop as well as mobile web and mobile app traffic.[11][35] This data is not immediately incorporated in any easy-to-reference format; stats.grok.se continues to use the pagecounts-raw dump, and the new Wikimedia API is several months in the future.
|
2014 |
October 1 |
Statistics availability |
Mobile-site sitewide analytics for the Wikimedia projects are available via the legacy API for data starting this day.[36]
|
2014 |
December 5 and 27 |
Statistics availability |
Wikimedia Labs Discovery dashboard displays data for mobile apps search metrics (search sessions, result sets, and clickthroughs) on Android (December 5) and iOS (December 27) mobile apps going back to these days.[37] Android mobile app search load times also go back to December 5.[38]
|
2015 |
January 1 |
Statistics availability |
Webstatscollector stops powering pagecounts-raw.[39] Instead, pagecounts-raw is powered by pagecounts-all-sites.[40] pagecounts-all-sites "uses the same pageview definition as pagecounts-raw, and so also does not filter out spider or bot traffic. But it does include access to mobile and zero sites, and is built on a more reliable logging infrastructure."[41] Note that although pagecounts-raw is derived from pagecounts-all-sites starting from this day, pagecounts-raw still does not include mobile data; since stats.grok.se is based on pagecounts-raw, it too still does not include mobile data.[explain implication]
|
2015 |
January 8 |
Statistics availability |
The first iteration on the Wikipedia clickstream dataset is available for data starting at this time. The project is managed by Ellery Wulczyn and Dario Taraborelli of the Wikimedia Foundation.[42]
|
2015 |
January 29 |
Statistics availability |
Webstatscollector is turned off.[43][explain implication]
|
2015 |
January 30 |
Statistics availability |
Wikimedia Labs Discovery dashboard displays data for iOS mobile app search load times going back to this day.[38]
|
2015 |
March 25, 26, and 27 |
Observed trend (Wikipedia) |
The post "The great decline in Wikipedia pageviews" by Vipul Naik is published, first on his personal site (March 25)[44] and then on LessWrong (March 27).[18] There is also a thread in the Wikimedia analytics mailing list about the post.[45]
|
2015 |
April 14 |
Statistics availability |
Wikimedia Labs Discovery dashboard displays data for desktop search events (search sessions, result sets, and clickthroughs) and desktop search load times going back to this day.[46]
|
2015 |
Spring (actual release), October 26 (announcement) |
Search engine use experience |
RankBrain is presumably rolled out, which could shift pageviews away from Wikipedia to more domain-specific sites?[22][47][48]
|
2015 |
May 1 |
Statistics availability |
Pageviews dataset is available in bulk form from this day.[49] This is the dataset that powers the Wikimedia Pageview API. Note that this is the earliest day of those days for which the dataset has data; the announcement (and presumably when the dataset first became available to the public) is on March 23, 2016.[41]
|
2015 |
May 19 |
Censorship |
There is a significant decline in traffic to the Chinese Wikipedia starting this date, likely because the Chinese government implemented censorship of Wikipedia starting on this date in response to the Wikimedia Foundation's announcement of a switch to HTTPS (making page-level blocking hard).[50]
|
2015 |
June |
Statistics availability |
WMF publishes comScore's unique visitor counts for the last time (data up to May)[51]
|
2015 |
June 11 |
Statistics availability |
Wikimedia Labs Discovery dashboard displays data for mobile web search metrics (search sessions, result sets, and clickthroughs) and mobile web search load times going back to this day.[52]
|
2015 |
June 12 |
User experience |
The Wikimedia Foundation publishes a blog post stating that all properties (including Wikipedia) are being switched over to HTTPS; previously, HTTPS was used only for logged-in users. It seems the switch is being made immediately.[53][54][55] One side-effect is that sites that get referral traffic from Wikipedia or other Wikimedia sites are no longer able to see those as referrers, cf. Research:Wikimedia referrer policy.
|
2015 |
July 1 (data from this date), December 14 (publicly announced) |
Statistics availability |
The Wikimedia pageviews API data is available for data starting July 1, with the API itself publicly released on December 14.[56] This gives pageview counts at a daily granularity, broken down by access type (user vs. bot vs. spider) and agent type (desktop v. mobile web v. mobile app). Wikipedia Views' data from July 2015 onward is drawn from this source.[57] The "Vital Signs" dashboard for easy tracking of total pageviews by day also becomes available for data starting July 1, 2015.[58]
|
2015 |
July 28 |
Observed trend (Wikipedia) |
SimilarWeb publishes "Is Wikipedia Being Hit By a Google Penalty?", the original piece talking about a sudden decline as a result of Mobilegeddon(?).[59] Various articles are written about this,[60] and even Jimmy Wales responds.[61] There is also a report produced by Wikimedia.[62] A Stone Temple Consulting post from September 2015 concludes that "Wikipedia did slide a bit in the rankings".[63]
|
2015 |
September 17 |
Regular report availability |
The first readership metrics report is published.[64] The report would be published approximately once every three or four months from that point on.[65][66]
|
2015 |
October |
Statistics availability |
Wikimedia Labs displays data for traffic from external search engines going back to this period.[67]
|
2015 |
November 18 |
Statistics availability |
Wikimedia Labs Discovery dashboard displays data for the Wikipedia.org Portal clickthrough rate and last action going back to this day.[68]
|
2015 |
|
Observed trend (general) |
comScore data suggests that United States desktop use peaked in 2015.[69][70]
|
2016 |
January 1 (data from this date) |
Statistics availability |
The Vital Signs dashboard now includes data on unique devices and monthly unique devices.[58]
|
2016 |
January 21 |
Statistics availability |
stats.grok.se ceases to be operational for new days' data.
|
2016 |
February |
Outbound traffic measurement |
The Wikimedia Foundation rolls out an update to the HTTPS meta referrer policy, that reveals the Origin rather than the full path of the referring domain. This means that websites that receive traffic from Wikipedia can once again calculate how much traffic they are receiving from Wikipedia, an ability that was lost in the switch to HTTPS. However, unlike the pre-HTTPS situation, full referral paths are not accessible, so websites cannot know what Wikipedia pages are sending traffic to them. For more, see Research:Wikimedia referrer policy.[71]
|
2016 |
March (onward) |
Survey data |
The Wikimedia Foundation's New Readers project, part of the Global Reach program, conducts surveys in Nigeria, India, Brazil, Mexico, and Egypt. The surveys are conducted off Wikipedia, and their goal is to better understand how people in these developing or semi-developed countries interact with Wikipedia and the Internet.
|
2016 |
July 15 |
Survey data |
Issa Rice publishes the blog post "Wikipedia usage survey results" discussing the results of surveys conducted through SurveyMonkey of how people use Wikipedia. The surveys were conducted in late May and early June of 2016.[72]
|
2016 |
July 20 to August 15 (approximately) |
Observed trend (Wikipedia) |
Anomalously high traffic on desktop sites for some major language wikis (English, Russian, and Dutch). It is believed to not be bot traffic but rather a result of unusual behavior of Chrome on Windows. It is discussed extensively in the readership metrics report published on July 31, 2016, while the anomaly is still ongoing.[58][73][74] It seems that a lot of the extra traffic affected only the main page, and therefore would not affect the analysis of pageview trends for other pages.[75][76]
|
2016 |
August 5 |
Statistics availability |
pagecounts-raw and pagecounts-all-sites are deprecated, with the last dump being generated for 12:00.[77][78][79][80] The deprecation had originally been announced on March 23, 2016 and scheduled for May 2016.[41]
|
2017 |
January 25 |
Statistics availability |
The Wikimedia pageviews API now supports querying pageviews at monthly granularity. Previously, only daily granularity was supported. The change applies to all past analytics (data going back to July 1, 2015).[81]
|
2017 |
February 17 |
User motivation |
The paper Why We Read Wikipedia, a collaboration of researchers at the Wikimedia Foundation, Stanford University, and elsewhere, is uploaded to the ArXiV.[82] A talk with the same title, covering the research, had been delivered by paper co-author Leila Zia in the November 2016 research showcase.[83] The paper is based on surveys conducted on Wikipedia, with a response rate of about 0.2% on desktop and slightly lower on mobile. It discusses three dimensions: motivation, information need, and prior knowledge. Breakdowns by dimension, correlations between dimensions, weekly trends, and correlation with user session behavior (time spent per page, number of pages) are all discussed.
|
2017 |
August |
Statistics availability |
stats.grok.se appears to be down from around the beginning of the month. The exact date it went down is not recorded.[84]
|
2017 |
November |
Statistics availability |
The Wikipedia clickstream dataset is published regularly for five languages (English, Russian, German, Spanish, and Japanese) starting this month. Clickstream data had previously been available for a few months but was not being published regularly. The dataset availability is announced in mid-January 2018.[85]
|