How we built AI-powered personalisation and retail media for one of Australia’s largest supermarket chains
Key outcomes
- AI-powered recipe recommendations live in production, matching each customer’s purchasing patterns against a major Australian recipe database
- Gen-AI meal planning agent – chatbot recommender responding to customer flags, preferences and dietary interests
- Retail media platform serving targeted advertising to millions of shoppers, with self-service supplier tooling and campaign reporting
- Event streaming infrastructure on Kafka and Confluent handling real-time customer interaction events at scale
- $1.5 billion in sales uplift attributed to the broader Restive partnership across this multi-year engagement
The challenge
For a major Australian supermarket retailer operating at scale, the ability to personalise customer experiences in real time – across digital, in-store, and advertising channels – is a strategic priority, not a nice-to-have.
Achieving it requires more than good ideas. It demands a connected martech architecture, a robust data foundation, event streaming infrastructure that can handle millions of real-time interactions, and AI and ML capabilities that can turn raw customer data into meaningful, timely experiences.
Restive became a long-standing technical partner on this journey – working across multiple squads and capability areas over the three year project, contributing engineers, architects and data specialists to one of retail’s most ambitious personalisation programs.
Our partnership
Personalisation, CDP and loyalty
The engagement spanned the full martech and data stack. Restive contributed across personalisation, CDP implementation, loyalty platform development, consent management, and the broader marketing architecture that connects these capabilities together.
A Customer Data Platform (CDP) is the connective tissue of a modern martech stack – it unifies customer data from every touchpoint (online, in-store, app, loyalty) into a single persistent profile for each individual. For a retailer at this scale, a well-implemented CDP is what makes real-time personalisation possible: without it, you’re working from incomplete, siloed snapshots of customer behaviour rather than a live, unified picture.
Retail media and ad-tech
On the ad-tech and retail media side, we built the digital platform for serving personalised advertisements to customers based on browsing and purchase behaviour – surfacing relevant brand advertising across app, web and in-store screens, with self-service tooling for suppliers and front-end reporting for campaign performance. The platform incorporates Salesforce, observability tooling, and real-time AI and ML for targeting and relevance.
Event streaming architecture
A core part of the engagement involved the event streaming architecture – built on Kafka and Confluent using event-driven design principles. This infrastructure handles real-time processing and distribution of customer interaction events at scale, powering add-to-cart flows, engagement triggers and an extensible capability layer that other platform features build on.
Event streaming is what separates personalisation that feels real-time from personalisation that feels stale. Rather than batch-processing customer data overnight, an event-driven architecture captures every customer interaction – a search, a product view, an add-to-cart – as it happens, and makes that signal available to downstream systems within milliseconds. For a retailer serving millions of weekly customers, this infrastructure is foundational: it’s the reason a recipe recommendation can reflect what someone bought yesterday rather than last month.
AI recipe recommendations and Gen-AI meal planner
The recipe recommendation engine is built on an AI REST API that pulls purchase history data from both the ERP and ecommerce platform, consolidates it into a data lake on Microsoft Azure, and then applies inference models to match each customer’s typical purchasing patterns against a major Australian recipe database.
The AI identifies which recipes are most likely to resonate with each individual and surfaces them in the app and online shopping experience – with add-to-cart automation built on top so customers can go from inspiration to checkout in a few taps.
More recently, this has evolved into a Gen-AI meal planner – an AI agent and chatbot recommender that responds to customer flags, preferences and dietary interests to generate tailored meal plans and product suggestions.
Technologies used
This engagement drew on a broad stack of enterprise-grade platforms and tooling:
- Data and AI: Microsoft Azure (data lake), custom AI/ML inference models, Generative AI (meal planner agent)
- Event streaming: Apache Kafka, Confluent (event-driven architecture)
- Martech and CRM: Salesforce (retail media platform, observability tooling)
- Platforms: ERP integration, ecommerce platform, loyalty and consent management systems
The result
A connected personalisation and retail media ecosystem now live in production – spanning AI-powered recipe recommendations, a Gen-AI meal planning agent, a retail media platform serving targeted advertising to millions of shoppers, and the event streaming infrastructure that ties it all together.
This engagement contributed to $1.5 billion in sales uplift across the broader Restive partnership – a multi-year collaboration that continues to expand in scope and strategic importance.
Frequently asked questions
How does AI personalisation work in retail?
AI personalisation uses a customer’s purchase history, browsing behaviour, and real-time interactions to predict what they’re most likely to want next – then surfaces relevant products, offers, or content at the right moment. At scale, this requires a connected martech architecture: a CDP to unify the data, event streaming to process it in real time, and ML models to generate and rank recommendations.
What is retail media?
Retail media is advertising sold by a retailer to its suppliers – letting brands pay to surface their products to shoppers at the moment of purchase intent, on the retailer’s own digital channels (app, website, in-store screens). It’s one of the fastest-growing ad formats in Australia because the targeting is based on actual purchase data, not third-party cookies.
What is a Customer Data Platform (CDP)?
A CDP is a system that collects, unifies, and activates customer data from multiple sources into a single, persistent customer profile. Unlike a CRM (which typically stores contact and sales data), a CDP is designed for real-time activation – feeding personalisation engines, loyalty platforms, and ad-tech systems with up-to-date behavioural data.
What is event streaming architecture?
Event streaming is an infrastructure pattern where every customer action – a click, a purchase, a search – is published as a discrete event and made available to other systems in real time. Apache Kafka and Confluent are the leading platforms for this. For large-scale retail, event streaming replaces batch processing and enables millisecond-latency personalisation.
How long does a martech implementation like this take?
Engagements of this scope typically run 18 months to 3+ years. The Restive partnership with this retailer spanned three years, working across multiple squads simultaneously – from CDP and loyalty through to retail media, event streaming, and Gen-AI features. Phased delivery is standard: you go live with core capabilities early and layer in more sophisticated features over time.