The development of a decision support model for the picking, staging, combining and loading activities in the grocery distribution warehouse of a large fast moving consumer goods retailer

Abstract

The Pick 'n Pay grocery warehouse situated at the Longmeadow distribution centre (DC) has 10 790 stock keeping units that are received from 123 vendors and are distributed to 421 stores located in several provinces as well as three neighbouring countries. The stock keeping units which are referred to as delivery units (DUs) are picked into handling units (HUs) (rolltainers used to transport DUs) and staged in staging lanes. Loading sta load the staged HUs onto transport units (TUs). The TUs then deliver the stock to various Pick 'n Pay stores. This process is referred to as the outbound process. The two concerns currently experienced by the outbound process is an excessive num- ber of dropped HUs (HUs that are not loaded onto their planned TUs) and excessive TU loading times. It was determined that a decision support model was required to provide insight on how to reduce dropped HUs and TU loading times. A Literature review was performed to investigate agent based simulation (ABS), system dynamics (SD) and discrete event simulation (DES) as possible simulation techniques. Tako and Robinson (2012) deter- mined that DES was most frequently used within the supply chain environment to assist in making tactical and operational decisions. It was determined that a DES model would be used in this project. AnyLogic was selected as the simulation software as this software provides the functionality required for a DES model. The simulation model was developed by simulating the outbound process's picking/staging, combining and loading events. Data captured between the 1st of May 2017 and the 30th of June 2017 was used to develop the simulation model. Key measures were identi ed against which the simulation model was validated. These measures included the number of DUs per HU after loading, loading time per TU, number of HUs staged, combined, dropped and loaded per shift and the number of TUs loaded per shift. After compar- ing the data generated for each measure from 100 simulation runs to the observed data with distribution plots and 99% con dence intervals it was concluded that the developed simulation model was a valid representation of the outbound process. It was determined that high levels of congestion in the staging lanes contribute to the excessive number of dropped HUs and TU loading times. Two scenarios were identi ed which could reduce staging lane congestion namely the increase in the number of stores with night-time receiving and the distribution of weekly volumes. The two identi ed scenarios were evaluated with the developed simulation model using data captured between the 1st of August 2017 and the 31st of August 2017. The e ectiveness of the models were determined by evaluating the number of HUs dropped per shift, loading time per TU,the number of HUs per TU, the number of HUs in the staging lanes and the ow of HUs into and out of the staging lanes. It was concluded that both scenarios could reduce the number of HUs dropped per shift. The scenarios did not indicate a signi cant reduction in TU loading times but did produce an increase in the HUs per TU and loading rate per HU. The scenario proposing an increase in the number of stores with night-time receiving was selected as the most suitable solution as this solution was the most e ective in reducing the number of dropped HUs and increasing the loading rate per HU.