Streamlining Wholesale Reordering with AI-Powered Recommendations

The Challenge

In wholesale, tailoring proposals to each client’s past behavior and expected needs is essential. Yet most reorders were still based on agent intuition or manual spreadsheet analysis. Agents needed help identifying not just the right products to offer, but the right quantities, at the right time. The client had more than a decade of historical purchase data, but no system in place to turn that into actionable recommendations. Integrating fragmented data sources and surfacing insights in real time was the first hurdle, and making those insights usable through automation was the next.

Our Solution

We began by aggregating and preparing over 10 years of client purchasing data, applying RFM analysis (Recency, Frequency, Monetary value) to segment clients and surface buying patterns. From there, we layered in Market Basket Analysis to understand which products were typically ordered together, helping build a logic of complementary recommendations.

“Decades of purchasing data turned into precise, personalized product suggestions.”

Using AWS Glue, we pulled and cleaned data from multiple internal sources and centralized it in Amazon S3 for fast, low-latency access. We trained multiple machine learning models within AWS SageMaker, including K-Means Clustering for segmentation, Apriori for product association rules, and XGBoost to predict reorder likelihood and quantities based on client-specific behavior.

The result was a powerful, lightweight API, served via AWS API Gateway, that integrates seamlessly into the CPQ platform used by the sales agents. It delivered personalized product recommendations and quantity suggestions tailored to the client’s profile, purchasing patterns, and sales targets. This drastically reduced the time spent creating proposals and increased the relevance and acceptance rate of each offer.

Impact & Results

• Delivered a real-time API with personalized reorder recommendations based on historical buying behavior.
• Enabled sales agents to build proposals faster, with higher precision and customer alignment.
• Improved product bundling and upselling opportunities through Market Basket insights.
• Integrated data from multiple systems using AWS Glue and centralized it in Amazon S3 for optimal performance.
• Built and trained ML models in AWS SageMaker, combining segmentation and prediction to power smarter decisions.
• Transformed a manual process into an intelligent, scalable system with measurable business impact.

What’s Next

We continue to refine the recommendation engine with ongoing model retraining, feedback loops from agent behavior, and deeper personalization capabilities. As the client explores broader AI adoption across their supply chain, this solution lays the foundation for smarter forecasting, inventory management, and automated sales support at scale.