Wikimedia Foundation Annual Plan/2018-2019/Mid-year review

From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search

The 2018-19 Annual Plan Mid-Year report[edit]

We are pleased to present our progress against the Wikimedia Foundation FY18-19 Annual Plan. Our FY18-19 Annual Plan is the beginning of a transition as we begin to focus on the 2030 strategy and its two key areas: knowledge equity and knowledge as a service. You can learn more about this new direction in our 2018-19 annual plan, published at the beginning of this fiscal year.

Below you will find a financial overview and status updates on each of the the planned actions and goals described in our FY2018-19 Annual Plan. Like Annual Plan, the report is organized around our priorities "Evolve our systems and structures", "Grow new contributors and content", and "Increase our reach". All financial information and program status reports refer to the period of July 1, 2018 to Dec 31, 2018.

Evolve our systems and structures[edit]

Wikimedia’s very existence is remarkable. It’s not just the idea that has proven compelling and useful to hundreds of millions of people around the globe. It is also remarkable that the Wikimedia community, and Wikimedia Foundation, has done so much with so little. Our traffic, community, use, and influence belie the lean nature of the organization. In the year to come, we will invest in the underlying systems and structures that quantify and sustain our projects, with the purpose of preparing ourselves for future work in support of our mission. This could include processes and norms internal to the Foundation, metrics that measure progress towards our strategic objectives, systems and structures that govern and support our movement, as well as the platforms and architecture that underlie our projects.

To achieve this goal, we will:

Planned action Please describe the status of these actions in 1-3 sentences.
Modernize our technology infrastructure, including our applications stack, services, and production platform; advance privacy and security-related systems; and invest in development of Wikibase and Wikidata. • Several workshops around Wikibase and improvements to the installation process to make it easier for 3rd party users to run a Wikibase installation

• Released support for Senses (meanings of words) as part of lexicographical data on Wikidata

• Released a beta feature for improved suggestions when entering data on Wikidata

• Improved constraint checks to verify data in Wikidata and made constraint violations available in the Wikidata Query Service to make it easier for editors to find issues in the data

• Refactor presentation layer away from business logic code in MediaWiki is on pace

• Parser unification work is on pace (it's a multiyear project) and porting parsoid to PHP is underway.

Increase support for the global Wikimedia community by rethinking and improving our community grantmaking, leadership development, and tools to make contributors’ work easier and more productive • Movement Organizers study: onboarded contract research lead Ana Chang, convened staff steering committee in November to develop research framework

• Movement Organizers research plan is underway with two field site visits planned for March and April, remote interviews begin late February

• A Grantmaking strategy that reflects our strategic direction is underway. Significant research and assessment work has been done and in development, and we are doing this in coordination with the working group. Explained below in more detail.

Grow the Wikimedia Endowment to ensure the long-term health of the movement On track for our goals to grow the Endowment.
Live our values by supporting diversity and inclusion, staff development, and operational excellence across our organization, and including environmental sustainability in our decision-making processes • Kicking off environmental footprint assessment with vendor Strategic Sustainability Consulting mid-January with the goal of developing an actionable sustainability roadmap by beginning of May

• We are retooling our software engineer environments to streamline development and deployment, reducing time to release and increasing developer satisfaction.

We will know we’re successful when:

Goal Results Explanation of results in 1-3 sentences
Our technological infrastructure is robust, flexible, sustainable, and secure, and can meet our needs for the future • We've maintained 99.97% uptime for our sites and services.

• Zend PHP7 has been deployed to be production ready for full migration of MediaWiki traffic in the second half of the year.

• We are ahead of schedule on deploying Apache Traffic Server as replacement for our Varnish clusters, enhancing availability, scalability, privacy, and multi-DC efforts in numerous ways.

• OpEx for US West Coast region reduced by 60%.

• Switch to new DCIM system has been fully completed, and work towards a "single source of truth" continues.

• Multiple automation-related tools have been developed, released and are being used for common workflows.

• The implementation of our improved backups infrastructure is behind schedule due to staffing challenges.

• We've made good progress on all our objectives, and they are generally on track or slightly ahead of schedule.

• Work on the backup infrastructure is now behind schedule, but this is expected to mostly recover in the second part of the year.

• Investing in automation to improve and accelerate our workflows works and yields results even in the short-term

We are able to serve 95% of our services out of more than one data center • Data center switchover (wiki read-only) time has been further reduced from 13 minutes to less than 5 minutes through increased automation through our custom written software Spicerack.

• Deployed a clustering solution for Memcached, implementing global caching of objects across our data centers, towards concurrent multi-data center deployment of MediaWiki. However, new barriers for this deployment have also been discovered because of the methods MediaWiki uses to store some data, requiring additional development (e.g. new session storage).

• We are in progress migrating all micro-services onto the new Deployment Pipeline

• All services being migrated onto the Deployment Pipeline (Kubernetes) are deployed in and able to be served from both data centers.

• We're on track for completion of migration of all micro-services by end of this year.

• While good progress has been made on the multi-DC MediaWiki deployment according to the annual plan, new and unknown challenges keep being discovered at all levels of the stack, requiring additional development work and collaboration between multiple teams, putting this project at risk for completion by end of year. This very complex project would benefit from more dedicated resourcing and less contention from other (related, needed) work such as the PHP7 migration, Platform Evolution, etc...

We see a 20% increase of services adopting the modernized metrics stack • Metrics collection jobs have increased from 23 to 30, a 30% increase Work is on track; new services put in production get Prometheus metrics as first-class citizens. Integration with the new deployment pipeline (Kubernetes) is expected to drive the adoption going forward.
100% of services involved in page views are using centralized logging • Over 90% of the existing log traffic has been moved to the new logging pipeline.

• Over 10 new application/services have started pushing logs to the centralized logging pipeline; 4 remaining candidates have been identified.

The project has seen steady progress, sourced in dedicated resourcing for it, after years of not being maintained. Work is on track and expected to be completed by the end of the fiscal year.
We have made progress in the journey to unify our core software assets by converging MediaWiki with our Node.js services • We have performed interviews, collected feedback, identified projects and discussed platform concerns with stakeholders before and during TechConf. We are currently distilling this information into Measurable goals for the PE Program. Following this we will use this to roadmap the work to achieve each goal. This will be used for planning remaining work for this FY, next FY annual planning and 3-5 strategy

• We have begun integration of Parsoid into MW Core with expected completion end of FY.

We are able to quickly iterate high-quality end-user experiences and measure success by running product experiments twice monthly At 92% (1.83/month average). On pace. Across audiences, we have completed 11 experiments (including user testing) since July, for an average per-month rate of 1.83, 92% of our goal. However, if you factor in the additional 8 still in progress, you can see we are well on pace to hit our goal by years end.

For a complete list, see:

Use of Wikidata increases by 15%, reaching 386 million data uses by the end of the 2018 calendar year As of Dec. 31st of 2018 the entity usage rose to 711 million across all projects. We surpassed the goal. The large increase comes among other things from a rollout of automated infoboxes on category pages on Wikimedia Commons.
We invest community resources according to a targeted set of funding priorities aligned with the strategic direction Alongside the the resource allocation working group, we aim to develop a funding strategy to advance the movement's strategic direction. To get there, we developed several projects that include mapping our existing systems and programs, identifying regulatory barriers in international funding, and conducting external research on other structures of grantmaking. In the mean time, we have made 253 grants, and continued to increase funds to emerging communities; 71% of grants went to emerging communities in the first half of the year.
We have grown the Wikimedia Endowment by USD 5 million and deepened our relationships with major donors On track to hit our goals. Engagement with major donors improving through events large and small, and based on preliminary data from Q2.

Grow new contributors and content[edit]

Wikimedia’s core strength is its global community of volunteer contributors. Contributors curate and create the knowledge (content) that is so valuable to hundreds of millions of people around the world. The people who make up this movement, and their knowledge, diversity, and ability to participate on our projects, make it possible for Wikimedia projects to exist and thrive. We will focus on increasing not just the net number of contributors and amount of content, but the linguistic, geographic, and demographic diversity of contributors and the breadth and depth of Wikimedia content. We will ensure that people everywhere can share their knowledge, including the 5 billion people who use smartphones to access the internet and the billions of people who only use phones, especially in underrepresented regions.

To achieve this goal, we will:

Planned action Please describe the status of these actions in 1-3 sentences.
Invest in the overall health of the Wikimedia community with tools and programs that help contributors respond to harassment and increase positive collaboration The Anti-Harassment Tools Team has improved the ability of Wikimedia administrators to apply blocks through the new Partial Blocks feature. Evaluating harassment reports is now easier with the Interaction Timeline tool; this tool still requires some features to meet its potential. The program has been preparing for the major User Reporting System project in FY18-19 by conducting workshops and design research
Support new and existing contributors through a richer suite of onboarding tools and intuitive mobile contribution experiences Growth team is working with Czech, Korean and Vietnamese WP to model first-day experience for registered users, and design interventions to help retain them. The "Help Panel" is currently being tested on Czech & Korean, giving users in the edit window to browse and search help pages, and ask questions on the local Help Desk. In addition, the Editing team is well underway on VE section editing for mobile; next, the team will redesign the VE mobile toolbar to make the mobile editing experience easier to understand and navigate.
Create techniques to measure contributor demographics at a granularity suitable for understanding improvements in this area Projects such us "Understanding first day" driven by audiences department are using our Analytics infrastructure and newer data sources like the Analytics Data Lake to gather insights about he most common workflows that new account holders go through during their first 24 hours. Special focus is given to what workflows lead to new edits.
Expand contributors’ capacity to address content gaps, bias, and more diverse content via machine learning and artificial intelligence (AI) and improved translation tools Our content quality measurement models have been extended to 5 new projects. Our initial analysis of biases originating from automated contribution strategies in Wikidata is complete and will be reported by the end of the FY.
Increase the participation of local communities in underrepresented regions through funding, mentorship, capacity building, and strategic partnerships support Through mutually agreed terms, we have rolled out 2 new machine translation systems for the Content Translation tool. These are LingoCloud and Google Translate. As a result, we have extended automatic translation support for 15 new languages, and improved automatic translation quality for a number of languages already being served through other machine translation systems. We expect to see the impact of these changes by end Q3, particularly from Google Translate which was rolled out in early Q3.
Advance the Wikimedia policy agenda and build a robust case for our values around the world through education, advocacy, government relations, and community mobilization We participated in education by publishing blog posts, statements, and other material in support of our public policy positions. These activities included endorsing the proposed WIPO Treaty on Education and Research Activities, hosting and moderating several sessions at the Internet Governance Forum 2018, supporting events in Brussels, and participating in global policy multi-stakeholder meetings. We also met with lawmakers and filed comments directly with governments, including to the Indian Joint Secretary Ministry of Electronics and Information Technology on a data collection bill, to the Minister of Electronics and Information Technology on intermediary liability, to US National Telecommunications and Information Administration on consumer privacy, and joined coalition letters on new proposed regulations in Europe. We support the Wikimedia community mobilization to oppose problematic sections in the proposed EU Copyright Directive.

We will know we’re successful when:

Goal Results Explanation of results in 1-3 sentences
The number of English Wikipedia administrators who report having the skills and tools to intervene in cases of harassment increases by 25% Initial metric is being expanded to measure impact in a broader area of program goals; queries to guage usage and effectiveness of new tools have been established, but need month-over-month metrics to give a full picture. Contractor is being engaged for qualitative surveying on general community health indicators. This work will not be finished until Q4
New contributor retention in target languages increases by 10% year-over-year No measurable impact to date, but the majority of our product interventions launching in the latter half of the year. Overall, the retention rates for these smaller wikis is fairly volatile, but tends to average lower than our overall average. We have yet to launch and measure interventions that we would expect to have a significant impact, but will have done so by the end of the fiscal year.
Number of edits on mobile increases by 20% year-over-year in target languages 40.4% increase, likely due to ongoing global trend The feature that has already launched has led to a >20% increase across wikis on iOS. However, due to iOS's relatively low proportion of overall edits, this did not have a significant impact on the 40% lift we are seeing. That is part of a global trend. Since we kicked off this brand new effort, it's going to be the second half of the year when the high-traffic interventions land and we expect to see more of our footprint.
We have the capacity and baseline data to measure diversity in contributor demographics We have started conversations with Legal, Reading Web, and Research to prepare the infrastructure and the exact research questions that we want to be able to address in Q3 (January-March 2019). We expect to run the first surveys to collect baseline statistics on demographics in Q4 (April-June 2019), pending on the current conversations and the results out of them.
The rate of content growth in target languages increases by 10% year-over-year The increases in articles published compared to the past fiscal year have been +18% for Q1, and +12% for Q2. More precise numbers to meet the defined metric would require to discount deleted articles and focus only on the subset of target wikis, which we are unable to do currently due to delay in data analysts support.
Our research leads to better understanding of knowledge gaps across languages We have developed an improvement on Wikipedia section alignment models first created last year. We now are able to align Wikipedia article sections across many languages. The models developed do better in 80% of the language pairs when compared with the baseline (Google Translation). The ability to align sections will allow us to surface content gaps across languages more effectively, and this is especially important for onboarding newcomers to Wikipedia.

We also finalized a comprehensive study of readers that allows us to understand the needs and motivations of readers across 14 Wikipedia languages. The learnings from this study can be found at Research:Characterizing Wikipedia Reader Behaviour.

Article section recommendation technology that helps contributors expand already existing articles is available in at least 10 languages We now have section recommendation API available in 23 language pairs. More languages can be added to the API on a pull basis. We are currently in the process of improving the quality of the recommendations through, including but not limited to, user feedback mechanisms.
Artificial intelligence technology providing edit and article quality prediction is in use on five new language versions of Wikipedia Our article quality models have been extended to 5 new language versions of Wikipedia and our edit quality models have been extended to 10 new language versions of Wikipedia.
Wikimedia values, projects and communities are present in global policy discussions on issues that directly impact our mission We are collecting baseline data to qualitatively evaluate presence, engagement, and impact in policy discussions. Data will be available at the end of Q4. We know that our efforts are increasing our presence in policymaking discussions because we are receiving an increase in invitations to participate in high-level or closed-door meetings, provide additional input, or speak about our positions. Our challenge is interpreting progress when policy outcomes require multiple years and are influenced by external factors.

Increase our reach[edit]

Much like growing new content and contributors, increasing our reach throughout the world is fundamental to our ability to achieve our vision, which calls upon every single human to share in knowledge. We currently reach a very large proportion of the planet. But there is no question that there are many more people who are not yet able to participate in free knowledge. We will focus on expanding our reach - the awareness and usage of the Wikimedia projects - across demographics. We are already a resource to many, we should become a service for all.

To achieve this goal, we will:

Planned action Please describe the status of these actions in 1-3 sentences. Explanation of results in 1-3 sentences
Bring more people onto the Wikimedia projects by investing in Wikipedia’s search engine optimization (SEO), and creating delightful, understandable experiences with improved user experiences for targeted geographies SEO intervention aimed at new readers is in the planning phase. Our SEO efforts for the first half of the year focused on all-wikis, but there is an intervention that could primarily impact smaller wikis that we are planning for the second half of the year.

The project to build product experiments to increase retention is just kicking off with the new Senior Product Manager. We are in the process of hiring an agency to build concepts, test them, and then create functional html prototypes.

Our SEO efforts for the first half of the year focused on all-wikis, but there is an intervention that could primarily impact smaller wikis that we are planning for the second half of the year.
Raise awareness of Wikimedia, particularly in underrepresented communities and groups, through targeted communications campaigns and support for local communities to conduct their own outreach The Communications team lead an effort to raise awareness in Mexico, working with Wikimedia Mexico, a local creative agency that WMMX selected, and the Global Reach team. The video earned 5.4M views during the campaign. We're currently working to understand the impact to awareness through phone surveys.

The Community Resources team is wrapping up the grants that came from the Inspire campaign in last fiscal year, and will be launching new grant templates to support awareness rapid grants, built off of those grants. Audiences and Communications partnered to support communities in raising awareness through small-scale social media experiments. Process and strategic learnings from those experiments will be rolled into the new grant type.

Build and maintain partnerships with mobile providers and invest in third-party infrastructure for offline access to Wikipedia to extend reach of Wikimedia content for users in Africa, Asia, the Middle East, and Latin America Continued conversations with mobile operators in the Wikipedia Zero network to explore other paths of collaboration. Offline projects in Nigeria started and worked under New Readers.
Improve discoverability and verifiability of knowledge through structured data integration on Wikimedia Commons, support for the WikiCite community, and technical partnerships Release multilingual file captions on Commons. These multilingual fields make it easier for search algorithms (including ours) to return relevant results in many languages. In addition, they are searchable via the API, so it is easier for 3rd party reusers to find and reuse media and the captions themselves.
Expand our network of institutional allies in libraries, education, and cultural heritage who contribute knowledge to Wikimedia projects The Education Team is currently undertaking an extensive mapping exercise to document and analyze worldwide program activities past and present, though past data tells us that education activities have been carried out in more than 90 countries in hundreds of educational institutions. Since 2016 the team has been working on a strategy to strengthen and scale local activities through strategic partnerships with like minded organizations, and has initiated formal partnerships with UNESCO and the British Council. The Education Team also supports the newly formed Education User Group, and is supporting the Wiki + Education conference organized by the Basque User Group in partnership with the Basque government.

The Wikipedia Library fully integrated into the library card research access platform, reaching 90% translation in 7 languages and 25% in 19 languages. TWL Developed a community-based #1Lib1Ref campaign strategy with a group of global ambassadors, while also rebuilding the hashtag tracking tool from scratch. The team conducted 'Wikipedia 101 for Librarians' as a Wikimania preconference, helped the libraries user group submit its first annual user group report, co-organized Wikicite 2018, and ran the #OAbot campaign adding 5000 free-to-read citations to Wikipedia.

The GLAM team focused outreach efforts on the 2018 GLAM-Wiki Tel Aviv Conference and supporting the growing community of interest in Wikidata and Wikibase among GLAM professionals: attending a NYC Wikibase workshop, supporting the development of a Wikidata + Libraries White Paper by the Association of Research Libraries, and a Wikibase blog series with WMDE. Our upcoming focus is on refining the Commons upload process, by supporting the newly deployed Structured Data on Commons with community run pilot projects, and a series of documentation projects focused on making high-impact Commons projects easier to implement. With the hire of Satdeep Gill, we are beginning research and documentation into the needs of communities who want to work with WikiSource to enable digitization and content partnerships.

We will know we’re successful when:

Goal Results Explanation of results in 1-3 sentences
More people in underrepresented demographics, countries, and languages have been introduced to Wikipedia, use it, and come back[14] See our metrics below in this table for the results we are tracking.
Global traffic increases by 1 - 10% (pending baseline analysis) We are currently seeing a 3% YoY increase in pageviews, due to search, but have not attributed this to our efforts We are seeing a 3% YoY increase in our traffic. This was primarily caused by an increase in search traffic from Google. We are testing 3 search engine-related product improvements and should be able to report on our impact this quarter.
15% increase in awareness of Wikipedia in target regions In the baseline survey in Mexico (our targeted campaign for this fiscal year), 53% of internet users reported having heard of Wikipedia. We are currently collecting endline data and expect to have the results within Q3.
Wikipedia is meaningful to 1 billion women We reached our first milestone of 1 million through the Mexico awareness campaign which focused on the centering women in knowledge through a youtube influencer via video and music.
1 million users in underrepresented regions can directly access Wikipedia through mobile provider and offline partnerships Worked on getting confirmation on product viability for offline preloads on WP app, which was done via New Readers work. No actual preloads deployed. Partnership for SMS/USSD access was put on hold due to regulatory issues, and it becoming less relevant. Zero was fully deprecated.
More institutions contribute high-quality content to our projects, integrate our projects into their workflows, and see Wikimedia as a natural ally in pursuing their missions Released file captions, the first feature in getting high-quality structured metadata on to Commons. It's an iceberg feature that has a lot of infrastructure work under it (MCR, 1st instance of federated wikibase) and more features to make our projects attractive to GLAMs and other open knowledge institutions.

Identified partners for GLAM pilot projects: Structured data/GLAM/Projects

Kiwix, a core part of the offline content distribution ecosystem, improves their platform and increases the number of end users by 15% Baseline metrics can be seen here: Kiwix/Plan 2018-2019#Targets.
Ending metrics are in progress.
Citations and links to external sources cited from Wikipedia articles are immediately and comprehensively archived and remain available to readers in perpetuity Released a comprehensive report analyzing the accessibility and openness of works cited in Wikipedia across all languages.

Started the evaluation of requirements to produce a live stream of changes to external links, to allow the Internet Archive to seamlessly cache all sources Wikimedians cite in near real time. The first MVP will be released by the end of the fiscal year.

We hosted the largest conference in the WikiCite series in November, bringing together developers, librarians and volunteer contributors to help build a structured database of metadata on all sources cited across Wikimedia projects in Wikidata. Our annual report will be published in February.

Financial Summary[edit]

Consolidated report of program related spending and non-program related spending[edit]

Consolidated budget: Program and non-program expenses

(All expense figures are in US Dollars, displayed in thousands)

Consolidated report (Jul 2018 - Dec 2018)
Actual Budget Variance
Fundraising & Other income 88,863 75,813 13,050
Staffing 17,923 18,759 -836
Data Center Expenses 2,229 2,342 -113
Grants 2,804 1,489 1,315
Outside Contract Services 1,598 2,613 -1,015
Legal Fees 536 667 -130
Travel, Events & Conferences 1,514 2,008 -494
Other expenses 947 1,556 -609
Sub Total Program Expenses 27,551 29,435 -1,884
Staffing 4,713 4,895 -182
Donation Processing Fees 3,747 3,363 385
Outside Contract Services 1,060 1,276 -216
Legal Fees 221 180 41
Travel, Events & Conferences 155 239 -84
Other expenses 1,788 1,961 -174
Sub Total Non-Program Expenses 11,683 11,914 -231