The Client

Our partner is one of the leading corporations in logistics, serving as the supply chain backbone for hundreds of major domestic and international businesses. Operating on a massive scale, their monthly import of a strategic material—at a volume of thousands of tons—is not just a routine operation. It is a complex financial problem.

The Challenge

The client's previous procurement strategy relied heavily on fixed schedules and historical experience, leaving them reactive to sudden market shifts. The challenge was clear: how to transition from a passive to a proactive purchasing model, capable of forecasting the optimal price point in each cycle to buy at the lowest possible cost?

The Solution

Recognizing that no single model could capture all market dynamics, we built a proprietary predictive engine, custom-tuned to the client's purchasing contracts, that combines multiple methodologies to deliver the most reliable results.

• Hybrid Modeling Approach: The system integrates classic time-series models (Prophet, SARIMA, LSTM) to capture long-term trends, alongside powerful multivariate machine learning algorithms (XGBoost, Random Forest) to analyze immediate, influencing factors.
• Technical Indicator Integration: To enhance sensitivity to short-term volatility, the solution incorporates financial indicators like RSI and algorithms designed to detect local peaks and troughs (Local Maxima/Minima).
• Seamless Deployment & Usability: The model was deployed via an API data stream and integrated directly into Google Sheets for maximum accessibility and ease of use by the procurement team. This was accompanied by a custom-built dashboard for real-time performance monitoring.

The Result

Our solution delivered an exceptional outcome, creating a direct and significant impact on the client's bottom line.

• Achieved 30% in cost savings on procurement compared to the previous schedule-based method.
• Transformed the procurement function from a reactive operation into a strategic, proactive one.
• Empowered the client with a powerful, data-driven tool to make confident purchasing decisions in volatile markets.

Technology & Tools Used

• Analysis & Modeling: Python, Prophet, Scikit-learn (Random Forest), XGBoost, GARCH
• Deep Learning: TensorFlow (for LSTM)
• Platform: Google Cloud Platform (GCP)