User blog comment:DaNASCAT/Technical Update: May 16, 2016/@comment-24473195-20160517145124/@comment-24473195-20160518112640

The best insights are those that help contributors focus their efforts on creating or improving pages. "Popular pages" are clearly useful, but lack context. For example, are these pages people reached through a search engine or other links, and are bots included in those views?

Some insights which would be particularly helpful for building or improving content would be:

Typo / spelling detection really needs a separate language library / extension to augment search. For instance, wiki search doesn't account for "advertize" (UK) vs "advertise"(US),  yet google will find both in a wiki. But yes, this would be an interesting insight, especially if it also accounted for synonyms.
 * Search count -  pages (keywords) often searched (using the wiki's internal search)
 * Mobile vs Desktop vs other device (e.g. TV, fridge) views - a clear separation between these views would be particularly interesting even without user agents. For example, a game wiki may want to focus on providing more information about its mobile version if it is popular.
 * Unique device count - a recent blog by wikimedia highlights the usefulness (and code) of those counts (http://blog.wikimedia.org/2016/03/30/unique-devices-dataset/).
 * Redirects views / count - there's actually academic research that explains this rather well (https://mako.cc/academic/hill_shaw-consider_the_redirect.pdf)