While much has been said about IoT and how that it may potentially change how business is done, very little has been done to actually utilize IoT or machine signals to create measurable impacts in service supply chain. Through joint research with OnProcess Technology, Massachusetts Institute of Technology (MIT) recently developed a new model for spare parts forecasting and inventory planning that incorporates machine failure predictability into the equation as part of a student thesis project titled “The Impact of Installed Base and Machine Failure Prediction on Spare Parts Forecasting and Inventory Planning”. The two MIT students who led this research are Mike Brocks and Renzo Trujillo and they were guided by MIT research advisor Daniel Steeneck. An OnProcess team supported MIT with data and service supply chain ideas.

The research compared two different methods for generating a spare parts forecast. The first method utilized a traditional time-series forecasting approach, generating a forecast for future spare parts demand based on historical demand data (time series model). The second method generates a forecast taking into account the machine signals from the equipment or part in question (predictive forecast model).

The forecasts from each model were then plugged into a standard periodic review (‘R, S’) inventory planning system and their performance was compared. The time-series forecasting model was run 15 times each in order to obtain a wide sample of performance data. The predictive forecast models were run 15 times for various levels of failure predictability in 10% increments of predictability. We measured the performance of each model using four different output metrics; average inventory, cycle service level (CSL), item fill rate (IFR) and sum of the forecast error. We found that as the ability to predict failure increases, less inventory can be held on average while providing a similar level of service. Even at relatively low levels of predictive power, significant decreases in inventory are possible.

We believe the method developed provides an avenue for companies to take their predictive analytics and Internet of Things efforts and turn them into actionable business value. Beyond the direct benefits of being able to reduce the spare parts inventory needing to be held and/or increasing the service level, we also believe the results found could possibly help companies redesign their entire service supply chain distribution networks by aggregating spare parts inventories into more centralized hubs, reducing the number of local and regional distribution centers in their network.

To find out how machine signals can be leveraged to optimize your post-sales supply chain, contact us at 508-623-0810 or email us at sales@onprocess.com.