Wikimedia Foundation Annual Plan/Quarterly check-ins/Audiences1 Notes January 2018

Present (in the office): Toby, James, Grace, Volker, Abbey, Katherine, Eileen, Neil, Danny, Joady

participating remotely: Subbu, Joe M., Niklas, Runa, Heather, Nirzar, Maggie, Pau, Dan G., Amir, Bartosz, Erica L., Sherry, Quim, Daisy, Ed

Slide edit

Slide edit

  • Work towards a single experience for mobile and desktop users, modernisation and providing the visual and wikitext editors on all devices.

Slide edit

Slide edit

  • ACTION!  label axes before uploading deck

They're labelled in the charts' title.

Could be helpful to put it on graph.

  • Graphs at the top left is time to interactive, time from starting process to being able to accept user input, times are roughly comparable
  • For roughly 50% of loads the 2010 editor is faster, this needs to be better before we deploy.
  • In contrast the time-to-loaded is better overall for the 2017 wikitext editor, which is good to note.
  • Generally the difference between the two wikitext editors is > 1 second, which sounds small but can make a big difference for how it feels.
  • Toby: Great to see this collected and displayed. Challenge when replacing something people are used to, have to be significantly better, not just slightly better. Hope this will help establish ground rules about communicating this change.

Slide edit

  • Next group goal is contributor tool maintenance. Overarching goal that touches all teams in Contribs.

Slide edit

  • James: Top level question is "can you keep editing?". During Q2, answer was yes—no downtime.
  • Deferred removal of the 2006 wikitext editor—no reason to rush into it. Only a few thousand user use it, but they're real humans too!
  • Unfortunate production data corruption issue that happened last week (related to multi-content revisions) which the team helped debug.

Slide edit

  • Subbu: Mostly work we're going to start this quarter. Improving template wikitext.

Slide edit

  • Subbu: On the top, this is the kind of wikitext users are writing now. The templates people are writing are not well-formed. On the left, this is a multi-part template. On the right, this is the problem with needing to escape within templates. This will help them write more readable wikitext.
  • Toby: Templates are such a core part of how editors create content, Subbu has some interesting further proposals for mainge templates easier to work with.
  • Subbu: We have some dev summit proposals to talk about that.

Slide edit

  • Also some of the rest of the work we did this quarter.

Slide edit

  • State today is that we have two parsers.
  • Multi-step process, can't fix right away.

Slide edit

  • Tidy is something from the '90s and we've been wanting to migrate away from for a couple of years. Last July we announced the plan to switch from this to the replacement in a year's time, i.e. by July 2018.
  • Thanks to the Community Liaisons, particularly Erica, this has been done pretty smoothly. We're on target.

Slide edit

  • The other part of use of Parsoid is supporting all the functionality. One of those is language variants.
  • We worked on the API endpoint, now working on improving testing infrastructure to make sure we're testing that properly.
  • Particularly important for the Readers team.

Slide edit

  • As a summary, going quite well.
  • Some wikis aren't making fixes though. Need to work out how to engage them to make this a success.
  • Katherine: Thank you for the work you're doing here, it's often behind the scenes. What's the timeframe on the long-term plan to unify the parsers?
  • Subbu: Definitely multi-quarter. Maybe next year. Hard to predict as we find issues as we're going. Feel quite close, but may be a year out. Long term hope is to leverage Parsoid to improve wikitext – cleaner to edit, easier to work with. Much longer term goal.
  • Toby: Also some deployment issues to work through with Technology, which we're working on.

Slide edit

  • James; This goal sits a little awkwardly, since it's now part of Audiences Design. Handing over to Nirzar and Volker.

Slide edit

  • Nirzar: Iterating on UI improvements, part of the work is improving the library (OOUI), part is accessibility improvements and language support, and consistency adoption.

Slide edit

  • Nirzar: First goal was adoption on Preferences.
  • Volker: We switched it over, on Beta we found a lot of issues with gadgets and extensions, that we need more time to get it to be a better support for users. Thanks to Bartosz in particular for heroic work on this, and we'll take it back up, but we want to make it great for users.
  • Nirzar: On background, this interface is really old and the most complex UI of the reading experience.

Slide edit

  • Nirzar: Also show something the Readers team are working on, improving the mobile preferences page, the first page using the OOUI libray, first time for the entire page.

Slide edit

  • Nirzar: Second goal was on mobile toolbar improvements, making this more responsive. Picked up as an exploration, see next slide.

Slide edit

  • Nirzar: Prateek, Volker, Pau and others considered different ways to make this happen, have a lot of prototypes now, next phase is to consider user testing. Considering editing on mobile differently so stopping at exploring ideas.

Slide edit

  • Nirzar: This goal was around optimisation.
  • Volker: Additional that we picked up mid-quarter.

Slide edit

  • Volker: Covering lots of files across lots of software. Forged common guideline to clean these up.
  • 5.8% out of just 8 icons in the critical path on mobile reading experience. Even more icons on the wikitext and visual editors.

Slide edit

  • Nirzar: Last one, part of styleguide work. Icons had diverged a lot, inconsistent for users, not comprehensive so missing for some contexts.

Slide edit

  • Pink ones are the old forms, new ones shown in black (these are always shown monochromatically to users).
  • Amir: Shout out to the Design team for RTL support work, reaching out to RTL speakers early on, thanks.

Slide edit

Slide edit

  • Neil: First objective was taking reseach and coming to consensus to what the most impactful things we could do.
  • Now going to work with Community Engagement to see what we can do together, lots of opportunity.
  • ACTION! Neil & Abbey to keep Katherine informed to enable her to represent this work

Slide edit

  • Neil: Tactics in constrast to current and historic methods, e.g. expect users to have read a 10k word policy before editing.
  • Neil: Also the three tactics we identified are not specific to technology, could also be programmatic context too (e.g. the Community Programs team's 1lib1ref campaign is an example of a program focused on microcontributions), the CE teams participating.

Slide edit

  • Neil: Over to Joe.
  • Joe: As Neil outlined, those different tactics are significant projects. We're really commited to them, and we'll spend the quarter researching product and design questions, putting flesh on the bones of the ideas, working out how they might succeed, proposing things to go forward to develop, especially a suite of metrics to gauge sucess and where we're going.
  • KM: Would love to go into more depth with you when you have more meat on the bones.
  • Toby: Thanks to CE, great to have your participation, and to Neil and Abbey.

Slide edit

Slide edit

  • Joe: Talked about this before so we'll go quickly.

Slide edit

  • Joe: On track to be completed this year. Opt-out rate for Recent Changes is down to ~2% despite the very prominent opt-out call-to-action.
  • Only thing left to do here is graduate the Watchlist filters out of beta, which we will do when the time is right.
  • Toby: Highlight that this is a slight change in tempo, we'll give these tools a little more time to bake with user feedback and still having engineering teams around.

Slide edit

  • James: About 800 distinct users having SD conversations a week. Did a user survey; based on that, we prioritized a set of improvements which we're going to take on this FY. The most important is modernizing and speeding up the front-end; this will make future fixes easier (e.g. a 1-2 week project rather than a 8-9 week project.)
  • Toby: One of the products we're going to take a look at, there's a need for understanding what the longer-term plan is.

Slide edit

  • James: This slide hasn't changed since the last time we discussed. Objective is still to reduce maintenance burden and get it out of beta. Did not go well this quarter. Had conflicts about design and product ownership. The plan is to reboot that this quarter and to carry forward as we previously planned. Substantial, growing numbers of people use it.

Slide edit

  • James: One of the highlights of CX is that it's the only part of the Contributors pipeline where we prompt users with task suggestion.
  • ACTION label axes on graph (DONE)
  • Joe: The task recommendation system we're looking at for New Editors is inspired by the success of CX recommendations.

Slide edit

  • James: Other important chunk of work was Compact Language Links. German rollout took a great deal of work from Erica and Amir; problematic conversation with users mistrusting our metrics. Impact was a 16% growth in language link clicks on German Wikipedia.

Slide edit

  • Neil: I will go expeditiously.

Slide edit

  • Neil: Key metric is overall numbers of active (5+) editors. This breaks it down by component.

Slide edit

  • Existing editors component is highly stable.

Slide edit

  • However, we are seeing a decline in new active editors (down 2500 per month over three years).

Slide edit

  • Not just a seasonal effect.. Taking out the pretty substantial (but odd) seasonal cycle, we do stillsee a meaningful decline.

Slide edit

  • Lots of ways to look at this; here are the wikis which were the biggest drivers of the 2013-17 change.
  • Wikidata and Commons were major gainers; Tamil and Bengali Wikipedias had significant increases given their small size.
  • The list of decliners is topped by some of the biggest Wikipedias: English, Spanish, Portuguese, German. It's probably not a coincidence that these are ones where there are concerns about community acceptance of new editors.

Slide edit

  • This is our overall suite of movement-level metrics.
  • In December, new editor retention hit its highest level in 3-4 years: 7.4% of first-month editors return in the second-month. Not clear whether this is a sustained increase or just volatility.
  • Mobile edits over a million for the second month ever.
  • Toby: Does retention going up and new editors going down tie to each other?
  • Neil: The two measure separate phenomena and aren't necessarily correlated: for example, if you remove initial barriers to editing (e.g. with the visual editor), you might increase your number of new editors, but if you don't increase the attractiveness of the underlying community, retention of those new editors might stay flat. Will dig more into this as part of our work on new editor metrics.