Data Products at FanDuel

Why we are investing in Data Products at FanDuel.
Varun Chugh, Senior Director of Data Products discusses how investing in Data Products is helping FanDuel become highly data driven.

At FanDuel, we are on a mission to make our organization highly data driven. As the gaming industry evolves, we must stay ahead of the competition, and we believe data will give us the edge. To achieve this, we are investing in our data organization across three key areas: People, Processes, and Technology.

We are investing in our people, expanding our Data Product and Tech teams, investing in new technologies, and improving processes that are helping us unlock the value of data.

Data Products at FanDuel are broadly viewed through two lenses:

1. Trusted Source of Truth; and
2. Data Enhanced Products.

Trusted Source of Truth
As FanDuel is growing there is an increased emphasis to use data as the foundation for decision-making, business operations, measuring product & platform performance and identifying trends.

In order to achieve business objectives, we need reliable data to make decisions and below I discuss the steps we are taking to make reliable data a reality.

Centralized Data Platform
We needed to create a centralized data platform to serve the organization’s data needs as we scaled. Multiple teams were replicating the same data, leading to different versions of it, creating inefficiencies and degrading performance of the data platform, as well as increasing costs. Establishing a single source of raw and derived data is helping address these issues.

Uniform and Consistent Data Delivery
Creating a centralized data platform is vital, but if teams can’t use the data quickly enough, we won’t be able to realize its full potential. To this end, we are establishing repeatable patterns for teams to access data based on use cases. For instance:

1. Enabling internal delivery teams to use Confluent + Flink + Clickhouse (TinyBird) for real-time streaming data.
2. Enabling internal data science and data analytics teams to use dbt + Redshift/Delta Lake for batch use cases.
3. Developing self-service tools and services to unlock the value of data. For example, we are providing customer data on demand to our internal applications, so various internal processes (like fraud, audit etc), can be handled quickly and accurately.

Data Quality
There is a lot of focus on data quality within FanDuel. Due to the highly regulated nature of the business, we go the extra mile to ensure data accuracy. However, we continue to look for ways to improve and to this end, we are implementing additional quality checks in our data pipelines in dbt. These checks will catch any errors that may occur during pipeline execution and allow us to fix them on the spot.

Data Governance and Security
We are setting standards for data collection, storage, and access to ensure reliability, timeliness, and accuracy. This helps us ensure our data is trustworthy and up-to-date. We are investing in data security measures such as encryption and access control. This helps keep the data secure and prevent any compromises.

As part of our Governance initiative, we are also making data cataloging and discovery simpler. This makes it easier for data to be found outside of the Data teams. To achieve this, we have implemented technologies like dbt and Alation.

Data-Enhanced Products
Data Enhanced Products are designed to provide customers and stakeholders with an enhanced and engaging experience. We utilize data and analytics to create products that are tailored to our business needs and interests. By leveraging the power of data, we are able to create products that are more efficient and provide unique insights.

Different kind of Data Products that fall under this umbrella are:

Algorithms — We build algorithms (also referred to as models) as a Data Product. Algorithms are a powerful tool for creating further data products that offer valuable insights and solutions. Data Products that build on algorithms are described below.

Analytics Dashboards: Analytics dashboards are used to track and visualize data for various businesses inside FanDuel. They are typically used to measure and monitor key performance indicators (KPIs) like APDs (Average Player Days), GGR (Gross Gaming revenue) among others.

Machine Learning Models: Machine learning models use algorithms to analyze large amounts of data and make predictions based on the data. These models are used to automate processes, identify customer trends, and make predictions about future events. At FanDuel, we rely on ML models for various use cases inside FanDuel from Customer segmentation to building out Fraud detection.

Recommendation Engines: Recommendation engines are used to make personalized recommendations to customers based on their past behavior and preferences.

Automated decision making: The end goal is to develop the capability for automated decision-making, no matter how complex the task may be. To build a system that can accurately and reliably make decisions for a variety of tasks, with minimal to no human intervention.

Data Products are Interdependent
Data-Enhanced Products and Trusted Source of Truth are closely intertwined. Data-Enhanced Products rely largely, but not exclusively, on Trusted Source of Data. Conversely, the structure of Trusted Source of Truth is largely, but not solely, informed by Data-Enhanced Products.

Data-Enhanced Products offer numerous benefits, such as consistency in business performance measurement, increased analyst throughput, scalability, and improved confidence in decision making. Additionally, they facilitate progress in embedding data into decision making processes at FanDuel.

Finally, I would like to illustrate how all the above came together with a real-world example. In Q4 of 2022, FanDuel experienced a spike in fraudulent activity. To prevent this, data had to be made available for analysis and decision-making on a more frequent basis.

At the time, the data was only available daily. However, by utilizing our centralized platform, partnering with teams, and employing a repeatable pattern of using Confluent for streaming data, we were able to deliver the data near real-time. This enabled the fraud team to build automated decisions to stop potential fraud and saved FanDuel millions of dollars. It also continues to protect our customers’ accounts and maintain our #1 Sportsbook market share.

To find out more about Data at FanDuel check out our previous blogs here and here.

Interested in working with us?- check out our careers page here.

Data Products at FanDuel was originally published in FanDuel Life on Medium, where people are continuing the conversation by highlighting and responding to this story.