عملکرد موجودی تحت محدودیت ها اندازه بسته و تقاضای همبسته مکانی

Motivated by a problem facing a large retailer, we consider the impact of pack size on the performance of a periodic review inventory system in the presence of spatial (i.e., between retailers) correlation of demand—which we model using an equicorrelated multivariate Poisson distribution. Employing simulation, we utilise a full factorial experiment to provide support for decisions on product and supplier selection, and whether or not packs should be split during distribution. We consider variables such as pack size, correlation, and the number of branches, and discuss how they and their interactions impact performance metrics such as inventory and shortage levels and the bullwhip effect.

Improvements in the management of inventories have played a key role in realising gains in supply chain performance. However, theory does not address some of the peculiarities of distribution practice, such as the phenomenon of splitting packs prior to shipment to the retailer. Our research was motivated by a problem facing a major retailer with almost 100 stores. The distribution centre (DC) receives weekly replenishment requests from each store. The high variance in demand rates between stores had prompted an issue as to whether to split packs prior to the retail level (i.e., at the DC). While this would incur higher labour costs, as well as increased risk of breakage and pilferage, it would reduce inventory costs. There was a concern that the effect of splitting packs on system performance could be complicated by correlation of demand between stores. Little research was found on splitting packs, or spatial correlation, and, to our knowledge, no research has jointly considered these two phenomena. Seeking to model problems including that described above we evaluate the impact of pack size on the performance of a periodic review inventory system in the presence of spatial correlation (i.e., between retailers) of demand. We consider a two-echelon supply chain consisting of one distributor (or equivalently a DC) and multiple retailers, with retail demands assumed to be from an equicorrelated multivariate Poisson distribution. We employ simulation to assess various performance measures in a wide range of environments using a full factorial experiment. The paper is structured as follows: the first section summarises literature relevant to the influence of pack size and spatial correlation on inventory performance. The second section outlines the model development, including a description of the transaction logic and demand generation. This third section on experimental analysis discusses performance measures, parameter selection, and the simulation. A results section presents an analysis of variance, highlighting the main and interaction effects of the various parameters. We conclude highlighting contributions and potential extensions.

Through simulation and ANOVA we have confirmed that positive spatial correlation between retailer demands increases demand variation at the distributor, but results in a lower variance amplification ratio. We have also demonstrated the size of the positive impact that larger delivery pack size has on demand variance at the distributor, the variance amplification ratio, and both average SOH and average BOs at the retail level. The experimental results can be used by management in cost models estimating the impact of pack size on supply chain performance. Companies such as that motivating this research may seek a smaller delivery pack size through vendor negotiations, or by allowing for packs to be split at DCs. Such decisions can be evaluated by analysing the economic consequences of changing P, along with costs associated with average SOH and BOs at both the branch and the DC level. The impact on soft costs such as breakages, pilferage, and labour would also need to be included. The focus of our simulation is on slow-moving items in a two-echelon supply chain, under an (R, S) inventory policy and synchronized ordering, within a retailer context. The results may not hold for other environments such as fast-moving items, or for a balanced ordering mechanism. Similarly, it may not be applicable in an extended multi-echelon supply chain or under a different inventory policy. Moreover, the outcomes are only valid when the factor levels are within the range covered by our simulation. The work may be extended in several ways. Firstly, financial considerations may be incorporated, e.g., by including a shortage cost penalty as another factor in the experimentation. Secondly, additional analytical modelling may, at least under restricted conditions, derive how inventory policy and control parameters (e.g., R, S, and k) should be modified to incorporate both pack size and spatial correlation.