Topic on Talk:Wikistats 2.0 Design Project/RequestforFeedback/Round2

After a quick look around the prototype (very pretty!), I'm left with the question: what use will this be for sub-communities within a project?

That is, as an editor on English Wikipedia, I don't really care about its 5 million total articles and the gazillion active editors (most of whom appear to be engaged in philosophical debates on Jimbo's user talk page), but I do care about what's going on in my little corner of it (w:WP:BARD WikiProject Shakespeare). How will Wikistats 2.0 let me answer how many active editors there are working on my WikiProject's articles; the most viewed and edited articles within its scope; the number of edits over time on those articles; etc.? On English Wikisource this might be scoped either by a WikiProject, or by a particular set of authors. In other settings the scope might be defined by a category, or a set operation (sum, intersection, exclusion, etc.) on categories.

Overview stats at the project level like those on the current dashboard prototype, or cross-project (wiki) comparison, is far too coarse a level to useful for actual editors. Stats at that level are mostly interesting to Wikimedia management and press; all actual editors will be interested in lower levels of granularity. For instance, I would be interested in stats scoped to a WikiProject on enwiki combined with stats scoped to a category on Commons. All-project stats (totals for a wiki) for Commons would be at best a curio, and of very little actual use to me.

Or if I occasionally lift my head a little, and wonder about the health of the movement, I might be wondering how many people are editing actual content pages (an enwiki article) vs. navelgazing in Wikipedia:-space (judging by my Watchlist, enwiki editors spend most of their time on WikiDrama in Wikipedia:-namespace pages, but it'd be nice to have real metrics for this). Or what's the distribution of edits between Talk: and mainspace for a given article?

Anyways, my point is that the current direction appears to be aiming for too coarse a level of granularity to be useful to actual editors, and if this is so you might want to reassess it.

Cheers,

Directly, we'll track this work here: https://phabricator.wikimedia.org/T163113

Xover, thanks for the perspective. Before I answer, it's important to note that the current prototype is coarser than we envision our first draft. When we fully flesh out all of our metrics, we will have some of the namespace breakdowns you're talking about.

That said, we are first aiming for feature parity with the existing wikistats: stats.wikimedia.org, and that doesn't yet go as deep as Wiki Projects and categories. We do want to go that deep, just not at first. And I should point out that while it's true the larger communities have many sub-communities, this first level of granularity is hopefully relevant to the smaller wikis that are just taking off.

For the larger communities, our next annual plan includes "the community backlog". In my mind this is a way to look at any and all work being put into the wikis by our different communities. I would love to get more detail from you about the examples you give. Scoping by WikiProject, cohort of authors, or category sets sounds like a great start. Any in-depth detail you can provide on this will be super-useful in design that's going on right now for the back-end API that will send Wikistats data. We can bake in these ideas now, even if just in concept, and implement them in a few months.

We're dedicated to the idea of being relevant to each and every member of the community and giving them relevant information that helps with their work. I'm not super excited about proving or disproving how much drama is going on, on different talk pages :) I hope we keep this positive and aimed at what we all want: sharing the sum of all knowledge.

I find myself, too, often being most interested in very specific questions/cross correlations to answer design questions or check assumptions. However, I am not familiar with the old wikistats well nor do I know if to check hypothesis like "The more experienced a community member is, the more they use talk pages" or the like is the purpose of the tool.