Personalising at scale: The AI platform powering real-time offers for millions of customers
Key outcomes
- Personalised offers delivered to millions of customers, drawing on real-time event data from sports markets around the world
- Event streaming architecture processing millions of events across global sports markets in real time – migrated from Apache Kafka to Amazon MSK
- Platform launched in Australia and rolling out internationally across the client’s broader group of brands
- Restive embedded across multiple squads for over two years, contributing across the full data and AI stack
The challenge
In a highly competitive market, relevance is everything. Generic promotions get ignored. The right offer, at the right moment, to the right customer – that drives engagement.
One of Australia’s leading gaming organisations set out to build a Generosity Platform: a sophisticated AI-driven system that understands each customer’s preferences, favourite sports, teams and betting behaviour, and uses that insight to serve up personalised offers that genuinely resonate.
The ambition was significant – millions of customers, events streaming in real time from sports across the globe, and a platform that needed to be fast, stable and scalable enough to operate as a strategic asset internationally across the client’s broader group of brands.
They needed a delivery partner who could work across data engineering, AI, and platform development – and keep pace with a program of this scale and complexity.
Technologies involved: Amazon MSK, Apache Kafka, AWS, real-time data pipelines, AI personalisation, machine learning
Our approach: A full-stack AI personalisation platform powered by Amazon MSK
Restive has been embedded across multiple squads on this program for over two years, contributing across the full data and AI stack.
At the core of the platform is a high-throughput event streaming architecture, originally built on Apache Kafka and subsequently migrated to Amazon MSK – the AWS-managed Kafka service – as part of ongoing optimisation work. This infrastructure processes millions of events across global sports markets in real time, feeding the data pipelines that power the platform’s AI personalisation engine.
Our work spans data pipeline engineering, AI model integration for offer targeting and personalisation, feature development for how offers surface and are redeemed across the app and website, and significant stabilisation and performance work to ensure the platform operates reliably at the scale the business demands.
The Generosity Platform is now a strategic asset – launched in Australia and being rolled out internationally across the client’s broader group of brands.
The result
The platform delivers personalised offers to millions of customers, drawing on real-time event data from sports markets around the world. What started as an Australian initiative is now part of a global rollout across the client’s broader group of brands – a testament to the platform’s scale, stability and commercial value.
If you’re currently assessing event-driven architecture for your organisation, or are curious to learn more, we’ve put together a practical guide for Australian businesses. Click here to read the full article.
Frequently asked questions
What is an AI personalisation platform?
An AI personalisation platform is a system that uses machine learning and real-time data to deliver tailored experiences to individual users at scale. Rather than sending the same offer or content to everyone, it analyses each customer’s behaviour, preferences, and context – in this case, favourite sports, teams, and betting patterns – to serve the most relevant offer at the right moment. The platform Restive built for this client processes millions of live sports events to power those decisions in real time.
What is Amazon MSK and why use it for event streaming?
Amazon MSK (Managed Streaming for Apache Kafka) is AWS’s fully managed Kafka service. It handles the heavy lifting of running a Kafka cluster – provisioning, patching, scaling, monitoring – so engineering teams can focus on building the data pipelines that matter. For a platform processing millions of real-time sports events across global markets, migrating from self-managed Apache Kafka to Amazon MSK delivered improved reliability, reduced operational overhead, and better scalability as the platform expanded internationally.
How does real-time event streaming power personalised offers?
Event streaming architecture allows a platform to ingest and process data the moment it’s generated – a goal scored, odds shifting, a market opening. By connecting that live event data to an AI model that understands each customer’s preferences, the platform can make personalised offer decisions in milliseconds. This is the difference between a generic promotion and an offer that feels timely and relevant. The Restive-built platform does exactly this across millions of customers and sports markets worldwide.
Can AI personalisation work at scale for millions of customers?
Yes – but it requires the right architecture. The key constraints are data pipeline throughput, model latency, and infrastructure reliability. Restive’s work on this platform demonstrates that AI personalisation can operate reliably at millions-of-customers scale when the event streaming architecture and AI stack are designed to work together. The platform has since expanded beyond Australia to other Flutter Group brands internationally.