Wikimedia monthly activities meetings/Quarterly reviews/Mobile contributions/October 2013

The following are notes from the Quarterly Review meeting with the Wikimedia Foundation's Mobile Contributions team on October 24, 2013, 1:30pm-4pm.

Present: Juliusz Gonera, Rob Moen, Kenan Wang, Tomasz Finc, Erik Moeller, Sue Gardner, Dario Taraborelli, Tilman Bayer (taking minutes), Howie Fung, Ryan Kaldari, Jon Robson, Jared Zimmerman (2:00pm-2:56pm), Kaity Hammerstein (from 2:08pm), Steven Walling (from 2:13pm), Vibha Bamba (from 3:06pm)
Participating remotely: Arthur Richards, Brion Vibber, Max Semenik, Michelle Grover, Yuvi Pandian

Please keep in mind that these minutes are mostly a rough transcript of what was said at the meeting, rather than a source of authoritative information. Consider referring to the presentation slides, blog posts, press releases and other official material

Agenda and introduction

1. Goals
2. Q1 Features
3. Mobile Editing
4. Q2 Roadmap

Tomasz: Welcome
two new faces since the last review meeting: Ryan Kaldari and Rob Moen joining from other teams, Dario helping out with analytics

Mobile Contributions team = Mobile apps team + mobile web team

Goals and Roadmap

Q4: spent time on contributions, focused on photo contributions, goal of 1k unique uploaders
Q1: 6000 unique uploaders+editors goal met
Q2: shift to retention, VE on mobile, Wikipedia app

Review of stats: 2013-14 goal for editors (1 mainspace edit/month) already surpassed by a wide margin
Vast majority web, vs. Android and iOS
Sue: Why no spike for WLM?
Erik: the number does not include the custom WLM app, which saw 5500 uploads (last year ~3000)
Howie: actual comparison should be with slide 67 (# of uploads/month)
Sue: feels like WLM causes huge spike in desktop uploads, but less so for mobile
Sue: so probably not justified to invest huge resources in WLM app
(Tomasz:)
strong readership growth, trending with the markets

Q1 themes

Arthur: reminder: 3 ways of using mobile site: alpha / beta / stable
Alpha: in Q1 worked on categories, nearby, image lightboxes
Beta: worked on user profiles, Echo, tutorials
Stable: Echo, editing improvements, navigation, thanks from watchlist
Navigation improvements: grouped sidebar into logical components
Most notable Q1 feature for stable: section-level editing (via pencil icon)
also: fixed captchas
Edit CTA for non-logged users, great driver of (especially) account creation
Sue: meaning you can't edit anonymously? correct
(Arthur:) "Thanks" from watchlist diff (may be reused elsewhere)
User profiles (beta):
intended to surface editing process, may help to drive contributions
First time editor tutorial (beta), being A/B tested
Sue: full tutorial with instructions, or just an encouragement? just an encouragement
(Arthur:) Issues overcome, e.g. HTMLFormatter (core of the mobile experience) moved into MediaWiki core, already reused by other teams, e.g. search team
Team is increasingly data driven, helped by Ori's EventLogging system

Apps

Tomasz:
Continue to work on photo uploads:
Getting people into the system works fine, but issues with retention
Use experience from campaigns (e.g. WLM)
Screenshots: campaigns app (Android only for now)
For discovery component (like map in WLM app), waiting on Wikidata
Howie: conceptually: "scavenger lists", right now created in wikitext
Sue: could it be also geography specific instead of just campaign specific (i.e. show all campaigns in vicinity)
Kenan, Erik: yes
(Tomasz:) also improved campaigns on desktop
administrative interface for campaigns: better history tracking, enable access by more admins
also looked at Commons app
modernized for iOS 7...
enable usage reports, improve detail pages, categories
stats: continuation of Q4 trends - mobile web still dominates upload #s
but apps used more often by power users, lower deletion rate
Sue: makes sense that power users are more likely to install the apps

Mobile editing

Kenan:
from last review: mobile editors tend to be newer, but mobile is also used by experienced users
two ways to look at mobile platform:
1. acquisition vehicle
2. general contributory platform
Know it is a strong account creation vehicle - spike in new registrations on July 23rd (from then on, higher than in 2012)
Howie: Hard to pinpoint a number (for the caused increase, as we didn't run any controlled test on mobile acquisitions), but when looking at it from various perspectives (with Dario), we get orders of magnitude
since about 2007, decrease in account creations, leveling lately (2012-13)
about 20-24% increase compared to 2012, probably attributable to mobile
(Kenan:)
now (enWP, Sept 2013) 18% of account creations mobile, cf. 16% for pageviews
Registrations / 1+ / 5+: for 5+, 10% mobile
Sue: so mobile signups have lower success rate in becoming active editors? yes
Sue: maybe retention (bad experiences) are not the problem, rather that those signups never had intention to engage deeply
(Kenan:) Levers:

CTA (for newly registered ones - e.g. not just take user back to article after signup, but directly to edit)
CTA workflows (for 1+s), e.g. GettingStarted

edit CTA drives conversion: 35.9% edit within 24h
Sue: people register for reading customization features, eg. watchlist - see that on desktop too
Erik: how much do users use mobile watchlist?
Jon: looked at usage of list (bookmark) vs. RC view...
Erik: how about email notifications?
Jon: email is optional on signup
Tilman, Kaldari: RC watchlist view is currently length limited, can be as short as only 1-2h back
(Kenan:)
types of first-time edits (hand-coded)
higher vandalism rates
but productivity rates close to desktop
Howie: looking at better ways to measure reverts (so far distorted by 1-edit users)
Kaldari: updated AbuseFilter last week
Dario: also need to look into how it really affects productivity of entire community (easy reverts)
(Kenan:)
most editors that register on mobile stay on mobile (in first 30 days)
Steven: makes sense, considering what we know about first edit sessions. [Note: a portion of users that make it to 5 edits do so within the first session-ish. So this metric may be picking up on this behavior]
(Kenan:)
..but many of the heavier mobile signups do migrate to desktop, too
Erik: what happens when a (new) mobile user gets a user talk page warning?
Kenan: they get an Echo notication, but there isn't a good way to respond
Erik: today people may be more aware of the hostile environment, speculation: perhaps stay on mobile on purpose to avoid this?
Steven: mobile may be a good way to test delivery of revert notification (Sue: yes, they are ephemeral)
Erik: question: is it actually desirable to get mobile users in closer touch with community ;)
Sue: people should have the opportunity to warn new users, but warnings could be more ephemeral
(Kenan:)
most mobile edits are from new accounts, but not all
Dario: massively driven by new acquisition. same U-shaped distribution on desktop (2006/07 cohort overly active)
(Kenan:)
Wikipedians of all (account) ages are editing on mobile
plot #mobile vs # desktop edits
Sue: so does this show that heavy editing will still be happening on desktop? (because the few awesome superactive users will always move to desktop)
Kenan: hard to say, e.g. because popular half-automated tools (Huggle etc) not yet available on mobile
Howie: jury still out on whether mobile signups become part of "heart and soul" of WP
Sue: support mobile-friendly routine tasks (Huggle-like tools could become part of this), but also treat it as acquisition tool, assuming that the writing of long articles will always happen on desktop
Howie: should work on that mobile --> desktop bridge
Sue: e.g. maybe after 100 edits, send a notification about desktop editing? or maybe it's unnecessary as they are sure to know about this possibility - but there may be other ways to encourage migration to desktop
Jared: should focus more on other types of contributions that are equally valuable (as writing articles) but more suitable for mobile
Sue: agrees, but worried about the future of pure article writing (need multiple tabs, access Google Books), could become a niche pursuit, desktop becoming a white collar / first world thing
Steven: keep an open mind about this, especially with tablet usage becoming part of mobile
Sue: yes, my concern is about now, would be great if things turn out better
Jared: should not ignore tasks like triage, "Liking"/"Unliking"
Sue: nothing wrong with "Liking"/"Unliking", but not contributing to encyclopedia
Jared: I disagree
Howie: ... team will look at *how* (not whether) to make tablets work
(Kenan:)
look at what the top 10 mobile editors did
interview with experienced editor: likes opportunity for mobile editing, but lack of features - not for me, but dumbed down for new users
summary: lot of acquisition, probably much of it incremental,... (see slide 53)
Device strategy:
on smartphones: web: wikitext editor for now. apps for experiened users
on tablets: redirect to mobile front-end, include VE. not yet clear on apps
Back to goals: 1000 active editors by Q2 - kind of reached already (952)

Q2 roadmap

(see slide)

Asks

Tomasz:

More A/V support in the office - some improvements were made a few months ago, but more needed in e.g. 6th floor

little confidence in pageviews data regarding e.g. user agents, maintenance of stats.wikimedia reports

Dario: data collection pipeline more robust now, should improve soon.
Erik: Varnish -> kafka pipeline, app type pageviews still to be done, challenge that our own apps use API, is that addressed now?
Dario: need to check.
Erik: ballpark figure right now: apps 20-25% of mobile pageviews(?). could include e.g. push notifications in apps.
Sue: we chose not to push people to apps.
Erik: for ideological and other reasons (avoid walled gardens, limited team capacity), yes.
Tomasz: need reassess, e.g. because some power users told us they lack features from mobile web on apps.
Sue: apps customized for one particular purpose would be a different issue than apps in general.
Erik: need better data for these decisions

With Ops, increase pace on OSM work

Erik: an ops engineer is assigned to that right now

continue work with VE team, both teams push changes in frontend infrastructure

Erik back to AV:
Tomasz: want this in every room. various audio issues
Max: some rooms have really strong echo
Sue: should invest in some sound baffle
Erik: what is blocked by OSM issue?
Tomasz: E.g. needed for Nearby feature, campaigns

Closing discussion

Sue: this was really useful. The quarterly reviews are getting good in general, too
Erik: the numbers are very good
also, looking at the top 10 editors was useful, avoiding stereotypes
Want to look at mobile contributions in the bigger picture of editor decline
Many saw the recent Technology Review story - which was like 1 year behind the curve
We see in your slides that we get into an area where the numbers actually start to matter
want us to become able to tell that story in the next few months
Howie: mobile alone not going to solve the problem, but the combined efforts of the Mobile and Growth teams can
Thought exercise: imagine mobile as "Wikipedia 3.0", fork of interface
Sue: do you mean mobile interface will be extended to all including desktop?
Howie: more as think of it as Petri dish
Tomasz: eventually, mobile team shouldnt exist as such, become part of all engineering instead
Sue: ideally yes, but considering natural organizational inertia probably not
anxiety about article writing, WP becoming part of Mary Meeker's "losers" in shift to mobile
Mobile first ... mobile only ... desktop as specialized tool ?
Erik: how can we replace back patterns, e.g. profiles vs. old user pages
VE or Flow teams would not take on user profiles
Tomasz: but also *we* need to catch up on Echo and Flow
Sue: if e.g. Flow can't do both mobile and desktop (lowest common denominator), perhaps it should be mobile only?
Erik: we can do mobile first in other team without diffusing the Mobile team
Sue: if mobile as green field means that implementation costs are much lower than on desktop...
Erik: but that's not quite the case for Flow
Tomasz: mobile can impact the editor decline over the next year, if not solve it. can provide patterns for desktop
Kenan: make sure user profiles actually make collaboration easier, need to be used within workflows
to assess that, need data model...
Howie: team hasn't yet had opportunity to think deeply about migration of user profiles to desktop
Sue: the migration for desktop could be a distraction for team
how and when will other teams need to switch to thinking mobile first?
Erik: next year might be another growth year for Engineering, new teams might adopt that view from the beginning
switching tablet to mobile will be huge milestone
Tomasz/Kenan: end of Q2
Sue: the ideal time for any team to showcase their achievements ;)
Erik: power users will be the test if tablet version has needed features
Connect with Communications within next 2-3 months on turning around the editor decline narrative
Julius, Jon: Flow on mobile would be rather difficult right now, should be incorporated into design from the beginning
Erik: lots of technical debt in that area (consistency)
Kaldari: QA could help with that
Erik: Flow also creates opportunities to engage users
optimize for workflows
Erik, Sue: Thanks a lot for all the useful details

Summary of discussion topics

(summarized by Howie, but feel free to edit:)

Using mobile to create friendlier patterns than we see on desktop (e.g., User profiles can help newbies understand that our editors are human. How to we use this to introduce this positive pattern on the desktop?)
- Who does this work? Question is how do we do mobile-first without diffusing the focus of the mobile team?
Migration path for new users from mobile to DT. We should keep observing this behavior.
- Perspective #1: consider building the right kind of bridges from mobile to DT (e.g., after x number of edits, let the user know there are more features on desktop)
- Perspective #2: delay migrating users to DT if mobile is a kinder, friendlier space?
Apps: we're historically not pushed apps for a number of reasons (e.g., low usage, we're a proponent of the open web), but we should continually re-evaluate (e.g., the mental model for some of our power users could be mobile Wikipedia = Wikipedia app). Next step for this coming quarter is to get a reliable baseline for the current app usage.