Focus Area:
Sales Forecasting for B2B Automotive Parts Wholesaler
Case and Problem:
A leading B2B Automotive Parts wholesaler was in dire need of refining their sales forecasting methods. With a vast inventory comprising hundreds of SKUs, and a complex compatibility matrix with specific car makes and models, the client found it challenging to maintain optimal stock levels in their warehouse. The unpredictable nature of their sales forecast led to either stock-outs or overstock situations, causing financial and operational inefficiencies. The main objective was to forecast sales with higher accuracy to ensure the right products were ordered from the manufacturing plant in a timely manner.
Analysis and Actions:
Upon collaboration, our team dug deep into a decade worth of the client’s sales data to uncover underlying patterns and correlations. Our findings led us to believe that a tailored Machine Learning (ML) model would be the most suitable approach to solve this complex forecasting problem.
- Data Analysis: Before embarking on the ML journey, our team meticulously analyzed the 10 years of sales data. This helped us understand sales seasonality, trend patterns, and SKU performances.
- Machine Learning Model: We developed a robust ML model in Python that could adapt to changing sales patterns over time. This model was built with a user-friendly front end to empower the client to run forecasts independently and adjust parameters according to their evolving needs.
- ERP Integration: Understanding the importance of seamless operations, we integrated the ML model with the client’s ERP system, ensuring data flow was unhindered and that forecasts could be generated in real-time.
- PowerBi Reporting: To facilitate a comprehensive view of forecast performance, we crafted a PowerBi report that allowed the client to juxtapose actual sales against forecasted values. This not only provided insights into the model’s accuracy but also helped in refining future forecasting strategies.
Outcome:
The implementation of the tailored ML model proved to be a game-changer for the B2B Automotive Parts wholesaler.
- Accuracy on Top SKUs: Our team scaled the time-series models focusing on the top 300 SKUs (representing 80% of the volume). The results were impressive – we achieved an average statistical error of only ~9% for annual demand on 75% of these tested SKUs.
- Directional Forecasting: When it came to quarterly forecasts across both major categories, the model showcased its prowess by predicting directional movements with a 71% accuracy rate.
The new forecasting approach not only optimized stock levels in the warehouse but also fortified the wholesaler’s confidence in making data-driven decisions for their ordering process.
By leveraging the synergy between historical data, machine learning, and modern reporting tools, we were able to provide a solution that significantly improved the wholesaler’s operations and bottom line.