This research deals with the inventory classification via multi-criteria optimization models. The idea behind this study is to educate readers about what factors must be considered when selecting inventory classification models. The factors that are more or less important depending on the datasets and variability in the input variables have been explored in this study. It has been learned that demand increase can make an initial inventory classification invalid, which is a critical factor considered in this study. This research shows how to first determine demand increase levels specific to an inventory item, which can change its classification. The next step is to test the feasibility of the model in classifying inventory items. Here, a discriminating power test has been introduced. Then, methods have been proposed to compare the performance of different multi-criteria models. Model selection criteria specific to datasets have been introduced and are intended to enhance customer service levels (CSLs) as well as reduce safety stock costs at the same time. Sensitivity analysis is used to show readers that varying the levels of input variables affects the performance of the selected model. Regression analysis is used to determine those levels that show when an originally selected model no longer gives the highest customer service. It can be concluded that although there are many classification models available in the literature, it is not possible to recommend any one model for obtaining the best performance in all cases. The procedure explained in this study should be followed in order to select the best model for a given dataset, thus improving service levels and reducing inventory cost.