FOX Sports Elevates the Fan Experience With Databricks

FOX Sports has a long history of driving the evolution of broadcast technology, from its high-definition coverage to experiments with virtual reality. Eventually, as the internet has grown more complex and overloaded with information, we wanted to invest heavily in our web and mobile search experience to ensure that information was easy to find.

As AI-powered search has evolved over the past few years, brands and businesses have adopted new approaches to improve user engagement and reduce search bounce rates. Increasingly, these efforts leverage data and AI to create smarter, more intuitive on-site search experiences.

At the heart of this shift is a focus on semantic search, which seeks to use AI to understand the intent behind a query. This allows search experiences to deliver intuitive results that align with the user’s expectations, moving beyond simple keyword matching and accounting for synonyms, misspellings, and complex relationships between concepts. Incorporating near real-time data on trending topics further enhances these systems, ensuring results stay relevant by dynamically adapting to rapidly evolving search patterns and user interests.

For FOX Sports, implementing these technologies was critical to delivering fast, context-aware search results that keep pace with our reader’s expectations.

Elevating the Sports Fan Experience

Our mission at FOX Sports has always been to deliver fan-centric experiences that keep sports enthusiasts engaged and informed. Just as we lead in sports entertainment, we aim to redefine how fans interact with digital content. As AI evolves, we see a tremendous opportunity to transform search into an intuitive, dynamic tool for our audience. Three key goals guide this vision:

• Enable natural, contextual searches for fans - Sports fans deserve a search experience that adapts to them — not the other way around. Typing exact keywords or sorting through irrelevant results has been the norm for too long but now, search can understand context and connect users to answers quicker and more efficiently.
• Integrate semantic search for richer discovery across all content - Fans shouldn’t have to dig through multiple platforms to find their favorite sports highlights. By embedding semantic search into our search bar, we enable seamless content discovery across the FOX Sports ecosystem, whether it’s first-party articles, syndicated content, or videos.
• Deliver a real-time, low-latency search experience - Timing is everything in sports. Fans turn to us for an informational experience that keeps pace with the fast-moving world of sports. Our goal is to provide a real-time search experience that delivers relevant results as users type.

Weaving these priorities into our approach, we are setting a new standard for how search can leverage AI to elevate the fan experience. For us, search is more than functionality — it’s a gateway to delivering the thrill and connection that define sports.

Fixing a Broken and Disjointed Search Experience

Before partnering with Databricks, we knew our search function needed improvements to empower our fans with modern search powered by data and AI.

Imagine searching for “Lionel” and seeing dozens of other athletes named Lionel before the soccer superstar Lionel Messi. Or typing an incomplete query like “christian pulis” would previously return unrelated people with a first name “Christian” or last name “Pulis” instead of the American soccer player Christian Pulisic. Fans, who were looking for quick and accurate results, often had to carefully craft their searches to get the right results.

Lastly, we couldn’t unify search across entities and our rich library of content. Let’s say a fan wanted to explore a topic such as the Dallas Cowboys offseason moves. The prior system wasn’t set up to accommodate surfacing relevant videos, articles, and entities all in one place. Instead, users had to jump between sections of the site navigation to piece together the information they were looking for, and this led to the user experience feeling disjointed and time-consuming.

All of these issues boiled down to one thing: the search experience wasn’t keeping pace with what modern fans needed. People come to FOX Sports expecting fast, relevant and easy-to-navigate results, and when that doesn't happen, it becomes a missed opportunity to keep them engaged and coming back to FOX as their go-to source of sports news.

That’s when our leadership teams knew it was time to make a change. We needed a smarter and more intuitive search solution that could understand the context and give searchers the results they wanted — and that’s where Databricks came in.

Helping Fans Discover Content Faster with AI

With Databricks, FOX Sports made key improvements to its search experience, aiming to deliver context-aware results to its readers.

The first step of this transformation was the implementation of real-time data ingestion pipelines in Databricks leveraging Spark Structured Streaming and Databricks Workflows. These pipelines continuously ingest and process entity data (e.g. athletes, teams, and leagues) as well as FOX Sports articles, videos, and third party content as it is published. Sports change quickly — players get traded, rosters change, new stories break and fan interest shifts with every game. Our data ingestion pipelines ensure these updates are reflected almost instantly.

To create dynamic relevance scores for entities, we built additional ingestion and processing pipelines for user interaction data, such as search queries and clicks. We use these insights to provide real-time data intelligence about what is popular or trending. For instance, a search for “Washington” might prioritize the Washington Nationals during baseball season and the Washington Commanders during football season. However, if the Washington Commanders sign a star player during baseball season, the scoring engine will surface this higher. By continuously ingesting and processing engagement data, Databricks enables us to ensure search results remain relevant, no matter the time of year or shifting user interests.

Mosaic AI Model Serving and Vector Search form the backbone of our search system. All data is synchronized continuously to Delta Sync Indexes and vectorized automatically using embedding models served with Model Serving. Entities and content are stored separately to support different retrieval patterns.

An additional Model Serving endpoint serves all search requests coming from the client by orchestrating various calls to Vector Search. Retrieval of entities prioritizes exact matches while content retrieval performs a time-weighted semantic search based on the publication date of the content. The final search result is constructed by joining the content and entity results and pulling in additional key entities tagged by FOX editors in the retrieved content. This endpoint achieves low latency under high load, ensuring that results are responsive.

With this enhanced search experience, users can explore abstract concepts like “Cowboys offseason updates” and receive a rich set of relevant results that includes Dak Prescott, their star quarterback battling an injury, and clips from NFL on FOX with analysis of the team’s decision to hire a new head coach.

For more granular searches, we save users a click by sending them directly to the requested subpage. For example, a query for “MLB schedule” will direct you to the MLB schedule page instead of the default home page.

Finally, our popular searches feature dynamically highlights the top entities per the scores computed from user interaction data before a user types anything into the search bar.  This assists fans in easily discovering what’s capturing attention across the sports world.

Databricks allows FOX Sports to unify data ingestion, processing, and model serving. This combination of real-time updates, dynamic search relevance, and AI creates an experience tailored to modern sports enthusiasts.

Creating a New, Agile Era of Sports News

The results of our search transformation at FOX Sports have been impressive. The new endpoint from Model Serving now handles hundreds of thousands of requests per day with significant spikes in traffic during weekends and live events.. 

Thanks to a caching layer that optimizes performance, the total number of user requests fulfilled through FOX Sports' web and mobile platforms is significantly higher. The popular searches feature alone accounts for over 25% of all search requests, highlighting the importance of incorporating real-time data intelligence into the experience. With this enhanced capacity, we can ensure our fans can quickly find the content they need, even during the busiest times.

To support this rollout, we also implemented a Databricks Lakehouse App that demonstrated the impact of our changes by providing side-by-side comparisons between the old and new search systems. Visually showcasing the improved relevance, speed, and accuracy of the new implementation, the app continues to secure stakeholder confidence and buy-in for this transformative search upgrade.

Overall, this overhaul of our search has not only improved the experience for fans but has reinforced FOX Sports’ commitment to keeping pace with the ever-changing demands of sports enthusiasts. Through Databricks, we’ve built a solid data and AI foundation that combines real-time performance, advanced technology, and dynamic search adaptability, ensuring our search remains a step ahead of our competitors.