Wikimedia Discovery/FDC Proposal

As part of the Wikimedia Foundation's Annual Plan for FY 2016/17 (July 2016 - June 2017), individual departments are submitting annual plans to the Funds Dissemination Committee. This document contains supplementary information which will be submitted to the FDC along with the proposal, and is intended to help explain Discovery's scope and rationale for the work to be done in FY 2016/17. This plan is a projection based on current understanding, and in line with principles of Agile software development, it is expected that the both the team's work and the plan itself will change and evolve over time.

If you have any questions, we'd love to answer them! Feel free to leave a message on the talk page or contact Tomasz Finc (Head of Discovery) and Dan Garry (Lead Product Manager, Discovery) directly.

Background of plan for FY 2016/17

The Wikimedia movement has content discovery issues. We have a wealth of content buried in Wikipedia, Commons, Wiktionary, and our other projects, which users simply cannot find in an easy fashion. Discovery’s goals are to improve content discovery mechanisms to connect the dots and build bridges between projects to alleviate this problem, and expose users to the wealth of relevant content that exists but they cannot find.

Discovery’s focus in FY 2015/16 was to measure and improve the fundamentals of search, and experiment with other content discovery mechanisms such as maps and complex queries to Wikidata. To that end, we launched the completion suggester in production on all wikis, piloted improvements to the search experience on wikipedia.org, launched experimental versions of a maps tile service and query service for Wikidata, and migrated all Wikivoyages to the new maps service.

In FY 2016/17, the Discovery Department strategic direction will focus on improving search and content discovery on Wikimedia projects. We will do this through improving our user satisfaction KPI across all Wikimedia projects focusing on surfacing imagery, maps, metadata, and sibling project content for on wiki and www.wikipedia.org search results. We will continue to explore new experimental mechanisms to encourage creation and discovery for community participation and innovation.

Strategic: Focus area 1

Improve Search and Relevancy of Wikimedia projects

Project description

Improve our search result relevancy by incorporating new heuristics that affect result ranking, surface cross-project and cross-language results, and explore commons media search.

Reasoning

We have a wealth of content available in other languages/projects/media that is not exposed; we should expose it!
Users should not have to rely on external services to search our site; we know our content better than anyone else
GLAM contributes a significant amount of content to Commons but has one of the worst search experiences

Alignment with strategy

Reach: Improving our search systems so that users depend less on external search services will help increase our reach. We aim to encourage users who come to our sites to stay and use our search system rather than leaving and using an external service, which aligns with reach approach two focusing on understanding of how and why our users come to and stay on our project, which was identified as the most selected strategy approach. Most of our users are not aware that there are a wealth content in Wikimedia projects beyond Wikipedia; surfacing content from those projects via search will increase the reach of those projects.

Dependencies

Continued support of a community liaison assigned at 100% time to Discovery to facilitate an increase in communication that otherwise would not be sustainable under Comms
Continued support of an agile coach assigned at 100% time to Discovery to maintain and evolve team practices in Discovery, ensuring optimal delivery of user value
Technical Operations: maintaining Wikimedia Labs which our dashboarding runs on to not incur additional load on our operations hire
Wikimedia DE: A commitment for structured data on Commons so that we don't have to develop an interim data pipeline

Milestones / Results / Impact

Significantly improve user experience of on-wiki search results by overhauling user interface to include imagery and metadata
Push changes to turn www.wikipedia.org into a portal for exploring content across all Wikimedia projects and languages
Expose relevant results from other languages and projects to users who search on-wiki
Explore incorporating structured data into relevance rankings across projects like Wikipedia, Commons, etc

Foundation impact

Readers and Editors can discover relevant content more easily
Less reliance on external services to search Wikipedia
Sibling projects content is exposed more consistently and encourages broader participation

KPIs

Strategic: Focus area 2

Evolve Content Discovery and Interactive tools on Wikimedia Projects

Project description

Explore and implement new ways for users to create and discover content

Reasoning

A picture is worth a thousand words. But if we do not simplify and encourage rich media, our mission of educating will not be as successful.
Users expect and request new ways of using rich media to find, interact, and contribute content.

Alignment with strategy

Reach: By building new, rich content experiences for our wikis, we'll attract new users with our dynamic content and disseminate information to them more effectively. Taking currently static content and adapting it into more dynamic content aligns with which aligns with reach approach one focused on adapting user experience to their user needs, which was identified as the second most selected strategy approach. This will encourage and promote further use of our sites to contribute and disseminate the world's knowledge.

Dependencies

Hardware for Maps to move from experimental to production
Consistent and extensive community engagement to evaluate and focus our efforts
Reading: Engagement of front end and design resources to improve maps and graphs

Milestones/Results/Impact

Launch maps service for Wikipedia
Empower community to surface more content based on maps, Wikidata, and other data sources - Create more ways to interlink content across projects, allowing easier content discovery.

Movement impact

Technologies allowing improvements to content quality and increase educational value, fulfilling our primary goal.
Increases goodwill - these new capabilities directly correspond to the community and readership requests

KPIs

Map usage
Graph usage
"Discoverability" increase - number of links followed due to new technology
Increased usage of WDQS

Core: Focus area 1

Maintain Search API’s

Project description

Split ElasticSearch cluster into two separate "big wikis" and "small wikis" cluster to improve performance and future scaling

Rationale

Our ElasticSearch cluster is monolithic which will have implications for scaling up to meet increased server load and demand

Alignment with strategy

Reach: Scaling our search cluster to support the demands that are placed on it is essential to ensure the success of our goals to improve reach by ensuring our users have a strong, stable search experience.

Dependencies

Technical Operations: data centre ops racking up servers
Technical Operations: coordination with Discovery Operations to maintain cluster

Milestones/Results/Impact

Maintain 99th percentile load time below 1.1s for prefix search, 4s for full text search, 2s for more like, and 2s for geodata
Maintain an n+1 configuration of ElasticSearch clusters such that the remaining data centre can serve traffic load whilst maintaining 99th percentile targets
Rack up additional server capacity for Q2 in preparation for entering full service in Q3

Movement impact

Maintain existing functionality with expected latency
Support smaller wikis better with more dedicated hardware
Better tolerance and resilience to outages

KPIs

Core: Focus area 2

Increase Server Capacity

Project description

Scale and expand Core services (WDQS and Elastic Search) to meet usage demands

Rationale

Demand on the ElasticSearch cluster has increased significantly due to extra functionality such as "more like" functionality; search is critical to content discovery and should be well supported.

Alignment with strategy

Reach: Scaling our search cluster to support the demands that are placed on it is essential to ensure the success of our goals to improve reach by ensuring our users have a strong, stable search experience.

Dependencies

Technical Operations: data centre ops racking up servers
Technical Operations: coordination with Discovery Operations to maintain cluster

Milestones/Results/Impact

Improve ability to monitor, alert and understand what is happening in the clusters.
Improve performance of APIs
Improve documentation of APIs

Foundation impact

More consumers can use "more like" functionality
Existing consumers can continue to depend on "more like" functionality
Cleaner and better documented interfaces will increase usage