We have inconsistent and often outdated statistics all over our comms materials, job openings etc. (number of wikis, number of articles, pageviews, uniques, rank etc). The deliverable for this task is (1) updated numbers for Wikimedia's official press kit that we all agree upon and (2) a proposed process to revisit on a regular basis.
**2016-02-09**
**2016-06-20**
GOAL:
The following metrics or some derivatives of them are identified as potentially useful for Communications.
---
[] pageviews
For large projects, counting pageviews is solved to a good extent, i.e., we can tell bot pageviews from human pageviews, of course, the identification of bot versus human pageviews can and should be improved (T138207). At some point, we identified around 5% of our traffic coming from botnets. This number can be 10%, more or less. For small projects, this metric may be problematic. An example is reported in T136084.
Pageview broken down by:
1. wiki (where wiki is defined by the combination of project and and pre-fix, for example (Wikipedia, en) is a wiki, and so is (Wikidata, xx).
2. site (mobile versus desktop)
3. country
---
[] industry rank
Which service should we use to report the ranking of our projects? What metric do we want to use to rank (users, unique devices, pageviews)? Here are some options:
Alexa: We used to use it. We swtiched to comScore.
comScore: Provided high level service, we used it for free.
QuantCast: ...
TODO: Erik and Leila prepare for the meeting on 2016-07-07 with Heather. This will require more research on industry standards, preparing some background about the possibilities and what the Foundation has used in the past, etc.
---
[] number of articles
+ It may not make sense to get a number for all Wikimedia projects together. For example, Wiktionary entries are different than Wikipedia articles, and different than Commons uploads.
+ Numbers of articles alone, without the sensitivity to details are problematic. Bot created in certain projects lead the project to "pass" another project.
+ We recommend reporting the number of articles taking into account the usage of those articles. More specifically, we recommend computing the number of articles in a wiki only when those articles have a view of at least n. We currently recommend n=5, though we can dive in and find the best n if this recommendation is adopted.
+ Note that the above recommendation represents our vision more clearly. It reports the number of articles created that are read by humans, and that is to capture the "share in" component of our vision.
---
[] Number of active editors per million speakers
We believe that this can be used as a score across many of our projects. At the end of the day, what is vital to the success of each wiki is a vibrant editor community, and the proportion of that editor community to the size of the speakers of a language, as a proxy of the speed at which content can be created to serve the needs of that population.
---
[] active projects
[] unique devices
used by the Foundation and come to an initial agreement on how they should be defined and computed.
- Discuss the use of qualifiers, instead of metrics.