This investigation is still ongoing in a way but I feel like we have a good idea for where to begin working on the tool.

Here's my answers to the open questions here:

• Can we work off of Wikimetrics if that can save us dev time and code duplication?
  • This seems the most viable option. Wikimetrics does indeed do a bunch of things requested from the new tool and has good documentation. Dan and Madhu are eager to help us get started with using it. It seems pretty extensible too. (code at https://github.com/wikimedia/analytics-wikimetrics)
• What all features are requested in the new tool that Wikimetrics does not fulfill currently?
  • Wikimetrics only allows cohorts by (list of) users currently. We need to create such cohorts by tags, categories, templates etc. as well.
  • Another thing that we want in the new tool is the ability to create "Programs" which in turn contain cohorts. Wikimetrics has a "tag" system, which can be used to create Programs and associate cohorts with them.
  • The UI needs changing, needs to be simpler to use.
• How easy is it to get Wikimetrics up and running locally?
  • I tried this. Installing Docker and building it was the easy part but I couldn't get the containers up because of some weird permission issues with mysql. I tried for a while before giving up. I will troubleshoot this afterwards but I don't think it's a big problem. It's not super easy but not very hard either.
• What are the historical pain points of using Wikimetrics?
  • Cohorts (group of people), limited to user who created it and can’t associate it with program(s)
  • Report generation takes time and the process is complicated (sums/individuals etc)
  • The UI is too complex (which led to the magic button interface)
  • Tags are not really used
  • Some things like validation etc. are pretty slow

Proposed next steps:

• Begin to poke around on Wikimetrics
• Decide what all we want to strip out
• Finalize new tool wireframes
• Add support for programs -- cohorts mapping
• Start by adding support for categories, followed by templates, cohorts
  • This will need a lot of thinking around how we will handle different wikis

Here are some of the use cases I came up with which we should keep in mind when designing the tool:

• Alice is a teacher holding an editathon in which a bunch of her students would work on improving some pages on a bunch of Wikipedias about women scientists. She has asked her students to tag their edits with #womeninscience.
• Bob is organizing an editathon with some Arabic participants who want to create some new pages on arwiki as part of the editathon. The pages can be about anything but Bob will be keeping track of their progress using their usernames.
• Claire has decided to hold a "global" food-a-thon where people improve articles about anything involving food in whatever languages they use. She doesn't know who all are participating but she'd like to use increased editing in "food" categories in gauge impact of her editathon.
• Dave is holding a translatathon where he's picked 5 enwiki articles about well known figures from the history of politics and is encouraging participants to translate those 5 pages into their languages. The participants belong to many different language groups.

I'm not sure how accurate/useful these are, feel free to comment. Also feel free to add things which you think I might have missed.

How easy is it to get Wikimetrics up and running locally?

• I tried this. Installing Docker and building it was the easy part but I couldn't get the containers up because of some weird permission issues with mysql. I tried for a while before giving up. I will troubleshoot this afterwards but I don't think it's a big problem. It's not super easy but not very hard either.

I managed to get it up and running today!