Hello Analytics people!
I have a specific analytics question about how the category tree on
commons is used. To get started I have drafted a schema at
https://meta.wikimedia.org/wiki/Schema:CommonsCategoryTreeUse
The description on the talk page of that schema is copied below.
Thanks for considering it!
Best,
Daniel - [[User:Dschwen]]
Question
How are anonymous users using the commons category tree to find
images, compared to logged in users. Is the category tree being used
to discover images?The proposed schema should emit events on page view
and on category link clicks. The event data should contain the log in
status (logged in/not logged in), and the current namespace number.
Analysis
The following analysis on the dataset would be performed:
Category page visitation frequency compared to image page visitation
frequency for logged in and logged out users.How much relative "time"
is each group spending the category namespace? This could indicate if
categories are a significant path for the discovery of images (as
opposed to direct jumps to image pages from internal/external
search).Category link click rates in category and image
namespaces.These metrics (again for each logged in and logged out
users) would indicate if the category tree is actively browsed (rather
than stumble upon).
* Category link clicks in the image namespace are an indicator for the
effectiveness of categories to find similar content.
* Category link clicks in category namespace are an indicator for
browsing the category tree to find specific content
Rationale
Motivator for this study is finding out the significance of the
category tree in content discovery on wikimedia commons. This directly
impacts decisions for gadget default deployment, such as the FastCCI
Gadget which would benefit anonymous users (if the category tree is a
significant funnel for content discovery. The schema is designed to
collect a minimum amount of data in a maximally anonymized way.
The data to be logged should be considered inexpensive (standard
identifiers isAnaon and pageNS in the schema). I have no clue how the
link click action will be logged, but determining the namespace from
the link target should be rather trivial (using mw.title for example).