Chelsea Builds Success off the Field with AWS

Football is a passion for billions around the globe and competition between the top teams is fierce. To stay ahead, modern professional football clubs have developed sophisticated academies that offer elite coaching to the world’s most talented young players. England’s Chelsea FC is such a club and enjoys success on and off the field. But the club wanted to find a way to enable all young footballers to benefit from its knowledge and expertise in elite football coaching, and not just those lucky enough to join the Chelsea academy in Cobham, UK. 

To do this, the club partnered with Chelsea Digital Ventures—a digital-first, consumer products business spun out from Chelsea FC in 2018—to create the Perfect Play football training application. The Perfect Play team needed the infrastructure and solutions that would allow it to rapidly develop, test, and iterate the app. That meant a serverless architecture that would let it spin up new environments in a few minutes. And as a data-driven business, it needed analytics and machine-learning (ML) capabilities to deliver accurate and personalized experiences to its users. Duncan Ross, Chelsea’s CTO, is a games industry veteran and has spent his career building advanced applications. When the time came to choose a cloud provider for Perfect Play, the choice was based on his past experiences of creating apps and products. “We considered all our options, but chose AWS based on my previous experience. In the end I stuck with what I had faith in,” says Ross.

Perfect Play uses both augmented reality (AR) and ML to guide the training of aspiring footballers. The app uses computer vision to track a user’s movements via their smartphone’s camera. It then makes recommendations on next steps and suggestions for improvement, learning as it receives more data. The app allows Chelsea to tap into the wealth of knowledge in the academy and share that with aspiring footballers and athletes around the world. 

Recommendations for Perfect Play are data-driven and powered by AWS ML technology. Successful AI relies on data so it can improve. To give its users the most accurate recommendations and the best advice, Chelsea developed CDV-ID, a fully anonymized and GDPR-compliant system that allows the development team and the app’s ML technology to learn from usage and offer continually better advice on training drills. 

Chelsea uses Amazon Cognito for user registration, login, and authentication and to connect user data to CDV-ID. Chelsea then uses Amazon Personalize and Amazon SageMaker to analyze the data and improve the accuracy of the training and next steps.

Chelsea Digital Venture’s plans are ambitious with further apps and direct-to-consumer products in the pipeline. And AWS allows it to plan with scale. “There was never a thought that something could not be done. And should we hit the scale we are planning to, we can move forward creatively without restraint,” says Ross. Perfect Play quickly rose to the top of the App Store’s sports apps chart after launch, which brought increased compute demand. That meant Chelsea needed the kind of architecture that could respond to spikes in demand for data processing power. 

Chelsea used serverless technology wherever possible, using AWS Lambda and Lambda@Edge in addition to Amazon Elastic Compute Cloud (Amazon EC2) to provide an environment that responds seamlessly to demand and usage. Duncan Ross describes serverless technology as “the core of the entire backend infrastructure”. While Perfect Play uses some device resources, additional compute power is sent to the AWS backend. Chelsea relies on AWS making calls to authentication services, the Amazon DynamoDB database, and its data lake. Amazon Kinesis Data Firehose lets Perfect Play load the streaming data to be used in analytics tools and to be modelled by the ML services. 

Amazon CloudWatch was used for governance and oversight during the launch. “It was essential that we could see the traffic streaming in. That visibility allowed us to tweak things as and when needed,” says Ross. He adds: “With identity and access management services for launch, we could change the limits instantly if we needed. Or if we wanted to adjust DynamoDB settings, we could do that straight away. It’s effectively like a launch control system.”

With a start-up’s energy and agility, Chelsea Digital Ventures went all-in on AWS. This lets it join the dots across multiple disciplines, from development, to marketing, to data science. Ross says he has always been a proponent of Agile methodologies from the early days, and notes that Chelsea needed a backend provider that was able to work in the same way and pivot as and when needed. “In fairness, we pivoted a lot as we developed the initial architecture.” 

One of the things he likes about working this way is that if you hit a development dead end, it’s easy to quickly move in a different direction, thanks to the modular structure of AWS. This flexibility, combined with good communication, is a key benefit, Ross says. “With AWS, there’s no you and us. It’s like being on a unified team. We have Slack set up together so we’re communicating on a daily basis. It’s peace of mind. If we have a problem, I know we can pick up the phone and work on it together.”