Reducing Inventory and Optimizing the Lead time in a Custom order, High model mix Environment

In this contemporary world, demand forecasting has become an effective tool for the success of any product organization. This is especially important when their components have long lead times and when the companies don’t build on order. The goal of this thesis is to reduce inventory by improving the forecast accuracy while maintaining customer lead time in a custom order, high mix model environment. In this master thesis investigation, the research questions that were formulated are answered individually. To be able to answer the research questions, a thorough literature review was done to understand the various findings within this research area. In this thesis, only the high cost commodities were considered such as engines and frames as it costs an arm and a leg to have these high cost commodities in stock. Additionally, under estimating the forecast values of these components can be detrimental to business as the lead time of these high cost commodities are too long. Firstly, the forecasting accuracy of previous years’ data is calculated and measured through forecast error measuring parameters such as cumulative forecast error, mean squared error, standard deviation of error, mean absolute deviation and mean absolute percentage error. In the empirical findings part of the thesis, the problems faced with the existing forecast method is briefed which highlights the root causes for overall forecast inaccuracy. Aforementioned forecasting problems inevitably increase the inventory level which is a serious threat to an organization due to working capital tie up. Secondly, a hypothesis model was developed as an alternate forecasting model by considering the demand patterns from the past three years (historical) data. By analysing the demand pattern it was clear that the nature of the demand has been lumpy. A hypothesis model known as Croston’s model was developed and by applying the historical demand values, the forecast values were calculated. A key performance indicator known as mean absolute scaled error was calculated for both the existing and Croston’s forecast method for the purpose of comparison. The results proved that the Croston’s method gives better forecast accuracy when compared with the existing forecast method. And finally, to improve the forecasting process as a whole, a benchmarking study has been successfully carried out. The benchmarking study is done with three internal companies within the Atlas Copco Group. The companies have been chosen by looking at the similarity in their product portfolio and business challenges faced. (Keywords: Forecasting, inventory management, forecasting methods, forecast accuracy lumpy demand