Synthetic Data, Fueling Smarter Retail Decisions

A major e-commerce company, serving tens of millions of users each month, faced a significant hurdle in optimising its recommendation engines and fraud prevention with demand forecasting.

Customer data being a valuable asset and at the centre their business, making the right use of this data was crucial. The main challenge was the inability to fully leverage it due to privacy concerns and technical limitations. This resulted in poor personalisation, reduced resilience and missed sales opportunities.

Challenges

The company faced three primary challenges in utilising its customer data:

Data privacy regulations: Sensitive customer data like browsing histories and payment records were subject to strict regulations like GDPR and CCPA. These laws severely limited how the data could be used and creating a major obstacle to innovation.

Long-tail & imbalanced data: Most popular items had enough purchase data however, niche products lacked sufficient information to train the recommendation models. Similarly, high-cost but rare fraudulent activities, like fake returns and payment fraud, lacked enough examples for models to learn from.

Slow experimentation: Data anonymisation and compliance reviews were time-consuming leading to delay in model development and experimentation by weeks. It also affected secure collaboration with external AI vendors who could not safely access the sensitive customer data.

Solutions

To overcome these challenges, the Cloudaeon implemented a Synthetic Data Generation Platform powered by generative AI. This platform created realistic and privacy-safe datasets for model training and experimentation.

Synthetic user journeys: AI generated synthetic browsing-to-purchase journeys, reflecting real-world behaviours like cart abandonment before checkout, impulse buying during flash sales and seasonal bulk shopping (e.g., Black Friday). 

Synthetic transactions & reviews: The platform simulated purchase histories to balance data for under-represented products and generated product reviews with varying sentiment to improve sentiment analysis models.

Fraud scenario simulation: The synthetic datasets were designed to include examples of rare fraudulent behaviours, such as multi-account abuse discounts and unusual cross-border payment attempts. Returning empty boxes or counterfeit items scenarios were also added to provide models with the examples they needed to learn.

Privacy-by-design framework: By leveraging differential privacy, synthetic data could not be traced back to real customers, allowing compliance teams to validate it as safe for both internal and external use.

On-demand data sandbox: Data scientists gained the ability to generate customised datasets on demand for specific experiments and external vendors could safely test their AI models without ever accessing sensitive customer information.

Impact

The synthetic data platform solution implemented by Cloudaeon, turned a data privacy liability into a strategic advantage and significantly improved business outcomes:

Personalisation & engagement: Recommendation engines trained on the more balanced synthetic data, boosted conversion rates by 18% and increased the average order value (AOV) by 12% by showing customers more relevant items.

Fraud prevention: 35% Improvement in recall in fraud detection models, saving millions of dollars by catching more fraudulent returns and coupon abuse without increasing false positives.

Faster experimentation: The experimentation cycles from 6-8 weeks to just 2 weeks that dramatically accelerating development and the launch of new initiatives.

Operational efficiency: 20% improved seasonal demand forecasting leading to more effective warehouse stocking.

Strategic advantage: The company could now run "what-if" scenarios, such as predicting customer behaviour for new product launches, without the need for real-world, costly trial and error.

Conclusion

By deploying Cloudaeon’s Synthetic Data Generation Platform, the retailer overcame data privacy and data scarcity barriers. Once a compliance challenge became a driver of innovation, accelerating AI experimentation, improving fraud detection and AI-powered personalised customer experiences.

This transformation not only delivered measurable business impact but also positioned the retailer for scalable and AI-driven success.

Wondering how synthetic data can transform your business? Submit your use case today and discover the possibilities with Cloudaeon.