Wikistats: Pageview complete dumps

Pageview complete is our best effort to provide a comprehensive timeseries of per-article pageview data for Wikimedia projects. Data spans from December 2007 to the present with a uniform format and compression.

Features of the dataset

• Unified, human-readable project names (e.g. es.wikisource instead of es.s)
• Page IDs (from 2015 onwards) that allow this dataset to be joined with Wikidata and MediaWiki History
• Corrected one-hour skewing problem
• Where available, agent type and access method dimensions exposed.
• From 2020 on, it includes numbers on agents detected through our bot detection filters
• KNOWN ISSUE: rows without Page IDs have only 5 columns, while rows with Page IDs have 6. We are applying a fix which will take some time.

Details on data segments

Sets of daily files are derived from the best data available at the time:

• From 2007 to 2011: from pagecounts-raw (to be loaded during the second half of October 2020)
• From 2011 to 2015: from pagecounts-ez
• From 2015 to the present: from the latest pageview definition

Data format

Compression of dataset is similar to that of pagecounts-ez: bzip files with hourly data embedded on each row, following this format:

• wiki code (subproject.project)
• article title
• page id
• daily total
• hourly counts

Hourly counts can be deciphered as follows:

Hour:

from 0 to 23, written as 0 = A, 1 = B ... 22 = W, 23 = X

All Analytics datasets are available under the Creative Commons CC0 dedication.