Recovering $2.4M in Annual Revenue Through AI-Powered Personalisation
Industry
E-commerce & Retail
Services Used
AI & Automation Solutions, Data, Analytics & Business Intelligence, Website Development & Digital Experience
Key Outcome
$2.4M
Annual Revenue Recovery
The Challenge
Our client — a regional e-commerce platform with 400,000 active customers — was experiencing a sharp decline in average order value and repeat purchase rate. Product recommendations were powered by a rules-based system built five years earlier. With a catalogue exceeding 180,000 SKUs, the system was recommending the same top-selling items to everyone regardless of browsing history, purchase patterns, or stated preferences.
Revenue from returning customers had fallen 18% year-on-year. The business attributed this to increased competition, but the data told a different story: customers were browsing extensively and leaving without purchasing. Personalisation was the gap.
Our Approach
We began with a two-week data audit, mapping customer behaviour across all touchpoints: homepage, search, product pages, cart, and post-purchase. We identified three high-value intervention points where personalised recommendations would have the greatest impact on revenue and return-visit rates.
We designed and deployed a collaborative filtering model trained on 24 months of transaction data, augmented by real-time session signals. The model served recommendations through a new API layer we built, integrated with the client's existing e-commerce platform without requiring a full replatform.
A/B testing was run for 6 weeks across all recommendation surfaces before full rollout. We worked with the client's internal team throughout to ensure they could maintain and improve the model without ongoing dependence on us.
Results
“The results were beyond what we expected. The team understood our data, our customers, and our constraints — and built something we could actually own and improve. Six months later the model is still performing and our team has extended it to email personalisation independently.”
VP of Product
Confidential Client
Ready to achieve similar results?
Let's map a solution built around your goals. Start the conversation today.