Efficient inventory management balances operational efficiency and customer satisfaction. d.light, a global leader in solar-powered consumer products, implemented AI-driven forecasting to enhance its supply chain. The initiative reduces costs and maintains product availability across the global dealer network.
Primary Challenges
The organization must address the following inventory optimization challenges:
- Generate accurate daily demand forecasts to prevent stock-outs and overstocking. Address low or zero demand patterns across specific dealer networks.
- Reduce inventory carrying costs to improve cash flow.
- Adapt to frequent product retirements and new introductions. Incorporate product replacements dynamically. Update and adjust forecasts based on new product availability.
Solution
d.light partnered with SquareShift Technologies Inc. to implement a solution on Google Cloud Platform (GCP). The architecture uses AI/ML technologies.
- Build native forecasting models with automated refitting to maintain optimal performance.
- Run Vertex AI pipelines for development, forecasting, and tracking using LightGBM and Linear Regression.
- Segment dealers by sales volume to create tailored models and enable precise daily forecasts.
- Evaluate refitted models monthly to apply consistently high-performing algorithms.
Results
The project delivered the following results for d.light:
- 15% inventory cost reduction. Optimized stock levels lowered overhead expenses.
- 10% decrease in stock-outs. Improved demand forecasting increased product availability.
- 12% Weighted Mean Absolute Percentage Error (WMAPE). Accurate forecasting improved inventory and procurement decisions.
SquareShift offers AI implementation services for demand forecasting projects like this one.
For the method behind forecasts like these, read our guides to demand forecasting with BigQuery ML and to hybrid time series forecasting.