Introduction
MTE-THOMSON, founded in 1957 as Metalúrgica Termo-Elétrica, is a top Brazilian manufacturer of automotive temperature control and engine management systems. With over 3,000 products, including sensors and thermostats, they serve customers in more than 100 countries. Their focus on quality and innovation has made them the only producer of PTC and NTC thermistors in Latin America.
To stay competitive in the global market, MTE-THOMSON efficiently procures essential materials using data and AI/ML. This approach ensures the quality and reliability of their specialized automotive components, driving their success in the industry.
Challenges Faced
MTE-THOMSON was grappling with two interconnected issues that were impacting its operational efficiency and profitability.
- The first was accurately predicting the demand for its diverse range of products. Inaccurate demand forecasts could lead to overproduction or stockouts, both of which have significant cost implications.
- The second challenge was managing inventory at optimal levels. Without accurate demand forecasts, maintaining the right balance of inventory becomes a daunting task. Overstocking results in increased holding costs and risk of obsolescence, while understocking could lead to missed sales opportunities and customer dissatisfaction.
Solution Implemented
The team from RapidCanvas was able to finish the project in a span of just three months. The implementation was carried out systematically, making progress at every step:
Data collection and cleaning
The team began with gathering relevant data from various sources. This data was then cleaned to ensure its accuracy and reliability, which is crucial for the subsequent stages.
Automated modeling for demand forecasting
With the cleaned data, the team then built a machine learning model specifically designed to forecast the demand for MTE- THOMSON's products. This model uses historical data and patterns to predict future demand, providing a more accurate and dynamic forecast than traditional methods.
Developing the inventory optimization system
Once the demand forecasting model was in place, the team developed an inventory optimization system. This system uses the demand forecasts to determine the optimal stock levels for each product, taking into account factors such as lead times, order cycles, and safety stock levels.
Creating dashboards with insights
The final step was the creation of a user-friendly dashboard. This tool provides an easy way for MTE-THOMSON to monitor key metrics, view demand forecasts, and manage inventory, enabling more informed and effective decision-making.
Results and benefits
Increase in operational efficiency
Reducing the time spent on manual adjustments and allowing the supply chain team to focus on strategic tasks results in savings of over $200K.
Reduction in order suggestion errors
With fewer errors, the risk of stockouts and lost sales has decreased, potentially recovering $200K in missed sales annually, and also easing up nearly $500K with optimal stocking.
Optimized stock levels
In 67% of cases, the average stock levels were reduced without an increase in stockouts. This means the company was able to maintain a leaner inventory without compromising on product availability.
Improved forecast accuracy
The Mean Absolute Percentage Error (MAPE), a measure of forecast accuracy, improved by 9%. This indicates a more accurate prediction of demand, reducing the likelihood of overproduction or underproduction.
Increased visibility and control
The new system provided greater visibility into the inventory management process. This allowed for more effective monitoring and decision-making, leading to better control over inventory levels.
Boosted Productivity
The supply chain team's productivity was enhanced as they could shift their focus from managing spreadsheets to more strategic tasks. This not only improved efficiency but also allowed for more strategic and proactive inventory management.
