“Companies tend to overstock inventory so that when customers’ products break down, they have replacement parts readily available. But purchasing and storing all that extra ‘safety stock’ is very costly. With the proliferation of connected products, we saw an opportunity to analyze each product’s machine signals to predict when components may fail and, thus, develop a more sophisticated forecasting model,” said Mike Wooden, CEO of OnProcess Technology. “It’s clear that the more accurately you can predict failures, the lower average inventory you need. This has the potential to save companies millions of dollars every year.”
Traditionally, supply chain experts have used mathematical models to calculate the right inventory based on factors such as past demand, variations in demand, the amount of stock in the market and lead time from suppliers. MIT students and research staff developed a new inventory model that incorporates machine failure predictability. They found that even seemingly poor machine failure predictability tests can lead to a significant reduction in inventory levels. It can also enable a superior ability to predicate part demands, which leads to improved service levels.
According to Dr. Chris Caplice, Executive Director, MIT Center for Transportation & Logistics, “Improving the demand forecast for repair parts can lead to significant inventory reductions, but it is notoriously difficult. This project has shown that utilizing machine data proactively can lead to better forecast accuracy and, in turn, potentially result in higher service levels with less inventory.”
