Welcome to Part 4 of the Demand forecasting in IIoT blog series. In this blog we will talk about what happens when the forecast over-estimates or under-estimates the actual demand. Be sure to check out parts 1, 2, and 3 in the series as well.
Real-world supply chains are extremely complex and hence difficult to model without comprehensive data sources. It is typical for demand forecasting models to under- or over-estimate the demand, when trying to capture the patterns of the supply chain. When trying to find the best model for demand forecasting, it is imperative to understand the cost of missing the target demand. Consider the example of predicting food sales demand for a restaurant. If you predict less than the actual demand, then the restaurant customers must wait too long to get food, resulting in customer dissatisfaction and eventual loss in revenue. If you predict much higher demand than actual, then it might lead to overstocking, which will increase food wastage and eventually overall operation costs.
It is important to understand which cost is more critical to the business and adjust the model evaluation metric accordingly. For example, if the restaurant considers an over estimation by 5 units appropriate, then the model can assume the prediction to be accurate if the prediction is within 5 units of the actual demand. If the cost of under- or over- estimation is dependent on the time of the day, then the model can have multiple evaluation metrics that penalize based on time. For example, a restaurant might want to overstock during lunch but understock during dinner time.
As more industrial customers look to digitize their supply chain using sensors, it is worth reconsidering the effect of new data and information on demand forecasting models. Customizing demand forecasting model to the operating conditions and business of the customer will enable robust demand estimations, thereby driving down significant long term costs for the customer.
Are you an industrial IoT customer looking to save money by having an efficient demand forecasting model, then talk to us. Our 20 years of experience in IoT combined with our expertise in data science, will help you navigate the complex world of demand and customer behavior. For questions about how Bsquare can help reduce your operational cost, talk to us.