Reducing Minimum Stock Cover Levels inFast-Moving Consumer Goods Industryusing Classification Schemes

Abstract: This thesis was developed at the Demand and Supply Planning department (DSP) of NestléPortugal whose mission is to develop planning scenarios encompassing the whole supply,production and distribution cycle to support the most appropriate decisions at each operationallevel. Stock policies are among the most important parameters that DSP defines periodically.Such parameter includes minimum and maximum stock cover levels. The minimum stock coverlevels tell how many days the stock will last if demand goes as predicted. From that value themaximum stock cover levels is then calculated and stock policies are set. Currently stock coverpolicies are defined by Supply Planners with a home built tool called “Optimizer Tool” that showsoverestimation. This situation implies extra cost and inefficiencies that the company wants toaddress by the present thesis work.After study of the context and specificities of the situation the goals agreed were: 1) Complement“Optimizer Tool” operation with an innovative process to reduce the suggested minimum stockcover levels. 2) Develop a case study based on “Optimizer Tool” routine operation fordemonstration purposes.For reasons associated namely with confidentiality issues the approach used was mostlyempirical, in the sense that no analysis of fundamentals of the “Optimizer Tool” was undertaken.On that line of work, after considering that stock policies are indeed the result of the interactionbetween “Optimizer Tool” operation with human judgement on several inputs that can beadjusted, the research question to meet the objectives was: How to optimize the integration of“Optimizer Tool” operation with the inherent human judgement? This question was basedupon two hypothesis that were formulated, tested and validated.The literature review showed that classification schemes for the individual items (Stock KeepingUnits or SKU’s for Nestlé) could be used with the Simple Additive Weighting (SAW)methodology in the search of a solution to the problem under study. Furthermore, it was clear thataddressing uncertainty factors related to inventory could be based on what was called the rollinghorizon framework (basically, learn as you go). These findings lead to the development of a toolor add-on putting together classification schemes and a learn as you go process.The validation of the hypothesis mentioned above was then performed. That included a sensitivityanalysis that made clear that the options made by Supply Planners when using the “OptimizerTool” in respect to two inputs, the so called Adjusted Demand Plan Accuracy (DPA) and AdjustedMaster Schedule Attainment (MSA), were critical to the quality of results in terms of stockpolicies. A specific set of classification schemes was then developed and combined with SAWmethodology in three different arrangements.The combination schemes were prepared to be applied to the final results of an “Optimizer Tool”run. That option was dictated by the existence of company targets for Adjusted DPA and AdjustedMSA (that in principle should be adopted). Additionally, such option keeps present operation ofthe system totally unchanged, just introducing a reference that allows a deeper analysis in respectto stock policies (as illustrated in the case study and subsequent discussion).The case study was successful and the possibility of taking sound decisions on keeping orreducing minimum stock cover levels was demonstrated. It should be noted that the tool or addonis not a substitute of human experience and knowledge. It is a support to a more informeddecision. Furthermore it opens new possibilities in respect to formalization, sharing, continuouslearning and adaptation to new conditions, in line with the rolling horizon framework ofaddressing uncertainty factors in respect to inventory.