Research:Mobile trends
 |  This page is currently a draft. More information pertaining to this may be available on the talk page. Translation admins: Normally, drafts should not be marked for translation. |
Mobile device usage is growing rapidly -- especially in the developing world -- and that is changing how we ought to view the desktop/mobile divide for Wikipedia. In order to make these trends clear and inform our strategic decisions, this report provides a high-level snapshot of Wikimedia Foundation's mobile projects as of July 2014.
Mobile device usage
[edit]In 2014, there are 3.5 times more mobile/cellular subscriptions in developing world than the developed world. Users in the developing world are 14 times more likely to have a mobile/cellular subscription than a broadband subscription (c.f. developed world at 4 times).[1]
Data
[edit]
The data for this report includes edits across all projects as well as sampled readership data across all projects. The start date of the data collection is 2014-04-20 and the end date is 2014-07-17.
The unsampled dataset for editors was obtained via all cu_changes tables combined. To exclude bots, the ug_group='bot' flag was used, along with a list of bots that do not identify as such in the db. The dataset for readers was obtained from the 1:1000 sampled logs and further sampled in the following way, to allow country-specific analysis: for each country with >1,000 observations in a 24-hour period, 1 in 100 views were considered. For countries with <1,000 observations in that 24-hour period, none were taken. In both cases, spiders, as detected with ua-parser, were excluded before the subsetting.
Contribution and readership
[edit]Device class vs access method
[edit]
The readers and editors (for the rest of this report, labeled as users) of Wikimedia projects, be it Wikipedia, Commons, etc., can access the website using one of the following device classes:
- phones
- tablets
- dektops/others
At the time of this report, the three main methods (or targets) for users accessing Wikimedia projects are:
- desktop site (e.g. es.wikipedia.org)
- mobile site (e.g. es.m.wikipedia.org)
- apps
All these methods are available to users from each device class, except for apps as they do not support users on desktop devices.
Readership and editorship by device and method
[edit]The following barplot shows the global distribution of contributions and readership as a function of the device class vs target site type.
From the above figure we see that:
- 97% of edits across all projects are done using a desktop device and accessing the desktop site, while only 65% of readership comes from the same category of users.
- 20% of the readership comes from tablet and phones, while only 3% of the edits come from the users using these devices. In other words, there is an increasing number of readers using devices that do not provide yet an interface that is as edit-friendly as the desktop site/desktop device combination.
To summarize, despite the uptake of mobile editing, readership is shifting faster to mobile than contributions.
Geographic distribution of contributions
[edit]The following plots show the distribution of the number of edits from each device class to each target site, for every country in the world, during the observation period. One immediately notices that contributions from desktop device accessing desktop are heavily dominated by the U.S., while the introduction of other devices and access types diversifies contributions from other countries, though clearly the volume of contributions is not always comparable to those made on a desktop device to the desktop site.
Desktop/other devices
-
Accessing desktop site via desktop/other device -
Accessing mobile site via desktop/other device -
No access to app via desktop/other
Phone
-
Accessing desktop site via mobile phone -
Accessing mobile site via mobile phone -
Accessing app via mobile phone
Tablet
-
Accessing desktop site via tablet -
Accessing mobile site via tablet -
Accessing app via tablet
Geographic distribution of readership
[edit]The following set of plots shows the distribution of sampled readership count from all countries, across all projects. The first observation, when comparing readership to contribution plots is that the world is flatter when it comes to readership, i.e., the volume of page views is less skewed towards the U.S. for every device-site combination than contributions.
Desktop/other devices
-
Reading desktop site via desktop -
Reading mobile site via desktop -
Reading app via desktop
Phone
-
Reading desktop via mobile phone -
Reading mobile site via mobile phone -
Reading app via mobile phone
Tablet
-
Reading desktop site via tablet -
Reading mobile site via tablet -
Reading app via tablet
Global South
[edit]The following plots represent the geographic distribution of pageviews, limited to countries in the Global South (using definitions based on this list.
Desktop/other devices
-
Reading desktop site via desktop (Global South) -
Reading desktop site via mobile phone (Global South) -
Reading app via desktop (Global South)
Phone
-
Reading desktop via mobile phone (Global South) -
Reading mobile site via phone (Global South) -
Reading app via mobile phone (Global South)
Tablet
-
Reading desktop via tablet (Global South) -
Reading mobile site via tablet (Global South) -
Reading app via tablet (Global South)
New users: acquisition and activation
[edit]- Read more: mobile activation trends, mobile acquisition trends
Registrations via the mobile site are steadily growing across all projects, and they do not follow the same seasonal fluctuations that we see on the desktop site.
The following plot shows the growth of weekly mobile registration in the first half of 2014 as a function of the first week of January 2014. The second quarter of 2014 shows strong organic growth in a direction that is opposite to seasonality: on the desktop site, we typically expect the volume of desktop registrations to drop as we approach the summer, in the majority of Wikimedia projects. The growth in the second quarter is organic as there was no known change in the user experience design of the signup workflow that may explain this growth rate.
The following plot shows that although we did not have any major feature release for mobile in the second quarter of 2014, the activation rate of users registered on the English Wikipedia mobile site significantly surpassed that of desktop: users registered on mobile convert at a higher rate into new editors than users registered via the desktop site. The patterns is also consistent across languages.
A similar trend emerged for new active editor rates, i.e., percentage of new editors who complete 5+ edits in the first 24 hours since registration, as shown below.
The rate at which newly registered users become active on desktop has remained constant in 2014 while the the corresponding rate on mobile has started to increase in the second quarter of 2014, and surpassed desktop towards the end of the second quarter. Note that the last two data points in each plot are after the switchover of tablet users to the mobile site, which contributed to a further increase of the activation rate of newly registered users.
Tablet users
[edit]New user activation
[edit]The increase in new user activation at the very end of the second quarter of 2014 was driven, among other factors, by the tablet switchover that the Wikimedia Foundation deployed in June. Tablet users were previously directed by default to the desktop site. As of the end of June 2014, users on tablet devices are directed by default to the mobile site. Registered users on tablets (albeit a relatively small population compared to desktop users) benefited from the switchover with a much higher rate of registered users editing and a significantly higher conversion to new editors and new active editors. As a result of the switchover and as of July 2014, the {tablet device / mobile site} combination has the highest activation rate across all device-site combinations.
Edit volume
[edit]Switching all tablet users to the mobile site had a noticeable impact on overall contributions. While producing a higher level of initial engagement for newly registered users (caused by the prominent signup calls to action), the switchover produced a significant and substantial drop in the number of overall edits and unique editors, due to the restriction on anonymous editing. The trend is pretty robust as it has remained several weeks after the switchover.
