Scaling a small data team with the power of machine learning
In an effort to continue to grow their business in existing and new markets, DAZN – a live and on-demand sports streaming service – wanted a fast, low-maintenance way to enable their small data team to run predictive analytics and machine learning projects at scale.
The company wanted to find a way to allow data analysts who were not necessarily technical or experienced in machine learning to be able to contribute in meaningful ways to impactful data projects. Ultimately, they wanted to support an underlying data culture with advanced analytics and machine learning at the heart of the business.
The situation
Until recently, the sports entertainment industry was dominated by cable or satellite TV systems and companies; if a customer wanted to watch a particular sporting event, he had little or no choice in how to do so. Now that consumers are breaking free from traditional TV, they are increasingly turning to specialised services streaming exactly the content they’re looking for, whether live or on-demand. And while they are willing to pay for these services, it means that entertainment companies – in the absence of the a fore mentioned virtual monopoly of TV broadcasts – are held to increasingly higher standards when it comes to quality and offerings.
In other words, because customers can turn elsewhere, entertainment companies have had to up their game, so to speak. Today, that means bringing innovation by way of predictive analytics and machine learning to optimise every aspect of the business, from marketing to customer service to product offerings. To do this efficiently, they must also bring this innovation at scale, hiring fewer people to do more such that insights grow exponentially along with the amount of data being collected.
The need for Big Data with a small staff
DAZN knew that in order to accomplish their goals quickly, they would need technologies that were simple and in the cloud. They turned to Amazon Web Services (AWS) and Dataiku in combination for their simplicity in setup, connection, integration, and usability, and they got up and running in under one hour.
With AWS and Dataiku, the small data team built and now manages more than 30 models in parallel, all without needing to do any coding so that the processes are completely accessible to non-technical team members.
They use these models as the basis for a variety of critical processes throughout all areas of the business, specifically:
- Content attribution to determine what fixtures are driving sales, enabling contextual information on key fixtures in each market.
- Advanced customer segmentation to identify user behaviours, particularly regarding content and devices on which customers use the product.
- Propensity modeling to identify customers that are likely to churn, enabling improved customer targeting for retention activities.
- Survival analysis to understand customer stickiness, enabling calculation of expected revenues to understand customer return on investment.
- Natural language processing on social networks for market research
Results of more effective team members = More data science
AWS and Dataiku have noticeably shifted the data culture at DAZN and have brought innovations in advanced analytics and machine learning into the spotlight throughout the company. Thanks to Dataiku’s ease, simplicity, and huge efficiency gains, DAZN has hired two data analysts who have already gotten up to speed and are doing as much work as five analysts in the pre-Dataiku team. In addition, the company has found that each data team member 2.5x more efficient in putting models in production.
Overall, the biggest impact has been empowering a non-technical team to create more models than ever before and get them into the production environment quickly to bring real ROI to the business. DAZN plans to continue to grow the team to three data scientists and 6-10 analysts to exponentially increase the number of machine learning models in production.
Comment on this article below or via Twitter @IoTGN