The client is an international sports betting and gambling company.
Headquartered in the UK, our client serves 14 million monthly players worldwide. Operating in diverse local markets with unique cultural norms, customer preferences, and regulatory requirements on online gambling, the client ran several programs to help players identify early signs of addiction and take corrective action to ensure and maintain fun, sustainable experiences, as required by law and compliance authorities in some countries. The company also operated an ML-based monitoring system to identify potential fraud in players’ paybacks or bonuses by identifying deceptive patterns among bonus recipients and flagging repeat winners.
The company had recently expanded to US markets and anticipated a considerable boost in its player base. To keep true to its mandate of putting enjoyable player experiences first, our client needed to dramatically scale up its on-premises data warehouses to power machine learning (ML) engines for real-time insights into player behaviors to identify addiction and fraud proactively. The company at the time had a fragmented and significantly siloed data landscape that hindered the creation of required data pipelines to process huge volumes and varieties of data, including contextualized, cultural, regulatory, and others. Scaling to ensure compliance and detect fraud was critical to stem revenue loss and provide all players a fair gaming environment.
The company decided to migrate its on-premises data warehouses and ML engines to Amazon Web Services (AWS) to bring in the required scale, scrutiny, and high performance. It also turned to Persistent to help augment existing ML engines to a cloud-native environment and create an AWS-hosted data platform to power future ones.
A tech-led transformation to migrate, scale and streamline experiences
Our client needed a data platform to power its ML engines, with pre-built data pipelines, cleansing, and validation mechanisms to help its data scientists glean insights from raw data on the go. We brought in a product engineering mindset, along with domain understanding and our expertise as an AWS Premier Tier Partner to operationalize an AWS-native MLOps architecture that brings in added visibility, improved business intelligence, and real-time insights for point interventions.
We realized the need to streamline the data fabric with a domain-aligned data architecture that could power the ML engines. We worked with the client’s data scientists to understand the business logic and the requisite data sources or features to be pooled for operationalizing an ML engine. For instance, features for the safe gambling ML model would include time spent, money invested, behavioral pattern etc., bringing more context to the ML engines that delivered meaningful insights. We worked to create data pipelines that would automatically cleanse, verify, and validate data for data scientists to query it from a data platform. We also rewrote ML algorithms using Python for its visualization capabilities, low barrier to entry, flexibility, readability, and platform independence. We embedded new rules to identify anomalies in player behavior, helping the operator proactively drive responsible gaming practices and reinforce corrective measures.
Once data pipelines and augmented ML models were set, we developed a data platform powered by AWS with a data ingestion and processing framework, backed by an AI factory approach to automatically manage data collection, storage, usage, security, and disposal. This minimized the administrative burden on the client’s data analyst team, freeing up bandwidth to pursue gaming experience innovations based on player behaviors.
Hosted on the AWS cloud, the data platform gave the operator end-to-end visibility into player behavior, and what actions were recommended to players with an affinity for gambling to prevent addictive habits. This visibility allowed our client to automate the regulatory reporting process and meet compliance with anti-gambling regulations.
Sustainable gaming, backed by behavioral intelligence
With a tech-lead transformational approach, and 10+ years as an AWS partner, Persistent assisted the client with operationalizing an AWS-native MLOps framework that can:
- Scale up to analyze the behavior of a growing, culturally diverse customer base
- Improve regulatory compliance with a single-window view into corrective actions taken to stem gambling addiction
- Reduce revenue pilferage by proactively identifying bad actors that gamed the bonus system
