2015 Wikimedia Foundation Product and Technology Highlights

Highlights from the Product teams

In April of 2015 the product team organized into audience teams in the areas of Discovery, Fundraising Tech, Editing, Reading and Community Tech. The goal of this organization is to create integrated approach in understanding and serving groups of audiences. Our hope was to give our teams more autonomy in making decisions, cross-collaboration and driving initiatives; we believe we have started to see this approach bear results. We also increased focus on community feedback and engagement. The Product department has created a product hub, expanded a cross department monthly showcase, is exploring development process and is engaging users in strategy, goals and testing. The following are just a few of the highlights.

The Discovery team launched with a data dashboard showing search, api, and feature usage. Two new services were launched into limited production, Wikidata Query Service and a new map tile service which is gaining increasing usage. We launched the beta of a better search Completion Suggester geared towards improving relevancy. Efforts to gain a better understanding of usage on the portal to Wikipedia were also launched this year.

The Fundraising Tech team continued to do stellar work supporting French and English fundraisers. We provided updates to Amazon payments, PCI gap analysis and improvements. A new Banner History feature in CentralNotice was also completed and provides the team with better tooling to understand banner performance.

The Editing team launched VisualEditor to more users and continued to improve. VisualEditor is available for all new accounts on the English Wikipedia and logged-in users on the Spanish Wikipedia. Auto-citations, editor switching, math extensions, image upload enhancements, graph editing and improved performance were also improvements in VisualEditor this year. Exploration began on mobile editing as well. The Content Translation tool has seen solid growth and been well received. Over 40,000 new translated articles have been created. Article suggestions was recently added to the experience, close to doubling translation output. Flow was released to beta and continues to be tested by many users. Echo notifications experience received enhancements and the ability to do cross-wiki notifications will be a beta early next year. Parsoid has experienced improvements in html size and improvements in performance.

The Reading team completed its strategy process and focused on reaching new readers in emerging countries and updating experiences for our existing readers. The Web team worked with the performance team to reduce page load time for the mobile web, which will improve the experiences of readers around the world but particularly in Asia, Africa and South America. The team presented their vision for a new services based low latency infrastructure at the Developer Summit in January and is partnering with Design Research, Advancement and Comms on ethnographic reader research in Mexico in February. The Android app team’s focus on features was recognized in the Google Play store and as a 2015 app of the year. Android has seen some nice additions including improved tabbed browsing, language switching, link preview and a map view (using Discovery’s new tile server). We’re looking forward to more social features in 2016. iOS included some nice improvements early in the year with Share a Fact, a clean design with short descriptors, a prominently displayed image at the top of each article, improved search functionality, a read more section, and an enhanced image viewer. The next major release of iOS is in beta and will roll out a new version early next year. The update to the app includes simplified navigation, more conventional look and feel and a new feed interface which allow Wikipedia to tell you when there’s something you should know. The Reading Infrastructure team created a dashboard for measuring the impact of their efforts to modernize MediaWiki authentication system (AuthManager), measured MediaWiki API usage, and data dump loads, made api and performance improvements. Two factor authentication will be the first service on the the AuthManager architecture and should land in early 2016. The Community Tech team launched the first Community Wishlist Survey for WMF. The process generated over 100 community driven feature requests and lots of positive feedback. The entire Reading team is looking forward to delivering more reader and community requested features in 2016.

Highlights from the Technology teams

The technology teams accomplished many impactful things this year. A very large change came in the form of implementing HTTPS to encrypt all Wikimedia traffic. This many year effort involved teams from across the Wikimedia Foundation. The final steps carried out by the Operations team went well for such a major change increasing privacy or safety. Equally challenging and also a multi year collaborative effort was completion of the Single-User Login. A large change that went well, makes a users experience better and provides a platform for many more cross-wiki improvements.

The Analytics team completed a needed improvement to our pageview definition, making our pageview data easily accessible through an API. They also supported a number of teams with dashboard builds including core dashboards for pageviews and Wikimetrics, support for HTTPS, Hadoop cluster, legal cases and overall improvements including capacity and support on EventLogging. The team also made progress in reduced failed state reports by 50%.

The Architecture team has been focused on preparing for the Wikimedia Developer Summit 2016 reviewing and coordinating submissions, evolving the Architecture Committee format and reviewing the RfC's.

The Release Engineering team, on top of that keeping a regular stream of ongoing releases, continued its excellent support and service improving Continuous Integration and deployment tools, maintaining timely releases of MediaWiki, Phabricator support and transition over from other tools, tooling improvements and automated testing coverage.

The Research team had many great contributions and had many showcases and studies made available. They created the Objective Revision Evaluation Service (ORES), which brings automated edit and article quality classification to everyone via a set of APIs. The impact ORES has started to have is around improving quality control and curation across many projects. More than a dozen editing tools and services are already using it. Projects like Wikidata are using ORES to reduce the amount of time and attention that patrollers need to spend reviewing the recent changes feed by up to 99%. ORES supports more than just counter-vandalism work. Article quality prediction helps WikiProjects curate articles and WikiEd identify student drafts that are ready for mainspace. Soon, ORES will also help mentors identifying good new contributors and improve our measurements by classifying edits by the type of work done. The Research team also explored active editor usage, new editor retention, clickstream analysis, reader behaviour, supported pageview improvements and research, improved linking, article coverage and translation, content remix, examined our unique tracking challenges and conducted research on the mobile experience and usage. They also did a large amount of additional analysis to provided support work for the NSA case, HTTPS analysis, VisualEditor analysis, fundraising analysis and testing systems, analysis on micro contributions and released a number of open data sets.

The Design Research team also conducted studies on new editor discovery, supported the roll out of VisualEditor, researched workflows, mobile collections, Flow research, link preview usability testing, user persona understanding, developed a database of research participants and provided education on how everyone can do design research. They conducted product development process and how we can strengthen our survey programs with several teams.

The Performance team brought huge improvements to MediaWiki via efforts like the HipHop Virtual Machine (HHVM) which reduced load time, reduced the editing slowness by 50%, improved tooling and analysis and also reduced the costs for server infrastructure. The team has also created a portal for seeing ongoing performance and extended this to product teams to improve the overall understanding. The have setup webpagetest to evaluate browser performance. Support for HTTPS, VisualEditor improvements, mobile performance all made large impacts. The team also made another great advancement for users by decreasing first paint-times, improving page load times by over 40%.

The Security team has continually worked to achieve a high standard responding to many incidents, performing audits on our infrastructure and features and driving best practices across the organization with training sessions. They also completed work on a weekly dynamic security scanning tool and promoted efforts to improve the our attention to privacy concerns.

The Services team has continued to develop and scale the RESTBase API. This effort improved VisualEditor save times, editor switching, reduction of html size and made efforts like the pageview API easier to deploy. The team made many improvements supporting features, improving code deployments and apps with API performance and simplification.

The Technical Operations team supported many efforts like the HTTPS change and they also improved reliability of Wikimedia Labs, evaluated distributed cluster environment for Tool Labs, reduced our security attack vectors, improved compression, prepared 5 year capital plan, preparations and setup new data centers (codfw, eqord), and made improvements to Wikimedia mail systems. Deployment effort was reduced using consolidation, increased abstraction and automation.