Cost Saving in a Supply Network with Product Affinities
Authors: Erik Berggren and Rawlings Nechavava, Supervisor: Peter Berling
This master thesis was conducted as a single case study of a supply chain that produces and distributes spare parts for heavy machinery within Europe. Throughout the study it was noted that, whilst managing inventory is a complex task, this can be made even more complex by con-sidering the demand dependencies of multiple items in the supply chain network. By running warehouses companies can better match the supply with their demand and this can be further enhanced by stacking products where possible to increase the stock per location. When prod-ucts are stacked on top of each other for the inbound flow, the expected result of this is that the distribution cost per unit goes down.
This master thesis attempts to show, through this case study, how such complex supply chains can be made more efficient. In the results, we show that if the current (S, s) policy is maintained and the products are stacked, there is a potential for savings in overall logistics cost. This is il-lustrated in table 1, where the decrease in the number of containers shipped and received, the decrease in the number of pallets handled and pallets in the system, when comparing non stack-ing and stackstack-ing, are reflective of improved work flow and reduced logistics costs.
Table 1: Simulation results for a (S, s) policy
Assuming a (S, s) policy No Stacking Allowed With Stacking Pallets in Stock 826.03 666.83 Average Number of Backorders 8.93 8.58 Average holding cost per day 837.32 880.84 Average containers received 1.127 0.99 Average Pallets Recieved 21.78 13.76 Averaged Pallets Shipped 21.81 20.66 Average Unstacking 0.0 5.70 Total Cost Per Day 7611€ 6983€
To consider product affinity, a model has been developed that considers a different ordering sys-tem that is dependant on the cost benefit of ordering a truckload at a time (see figure 1). The heuristic determines which product is associated with the largest saving, when ordering. The most beneficial pallet is added to a shipment, until the shipment is full. After which additional shipments are added until it is no longer beneficial to add additional shipments.
Figure 1: Ordering Heuristic
In the results in table 2 and table 3, we show that the proposed model outperforms the (S, s) policy, mostly through higher savings in inventory costs and fewer pallets in the warehouse. How-ever, the behaviour of the model is hard to predict, it requires a large amount of data and is also difficult to alter.
Table 2: Comparison of inventory costs, when stacking
Comparison No Stacking Model (S, s) Average Pallets in Stock 534.50 551.02 Average Number of Backorders 14.53 14.13 Average holding cost per day 616.06 730.10 Average containers received 1.01 0.98
Table 3: Comparison of inventory costs, when not stacking
Comparison No Stacking Model (S, s) Average Pallets in Stock 534.50 551.02 Average Number of Backorders 13.96 14.67 Average holding cost per day 656.66 730.10 Average containers received 0.72 0.98
Further more we show, through table 4 that, stacking the products in the new model resulted in increased potential savings in labour and other costs and a reduction in the amounts of pallets in the flow. However, while trying to keep the same service level, the tied up capital increased significantly.
Table 4: Simulation results Model, stacking or not stacking
Assuming model No Stacking Allowed With Stacking Pallets in Stock 684.01 534.50 Average Number of Backorders 14.53 13.96 Average holding cost per day 616.06 656.66 Average containers received 1.01 0.72 Average Pallets Recieved 21.76 13.77 Averaged Pallets Shipped 21.80 20.67 Average Unstacking 0.0 5.72 Total Cost Per Day 7592€ 6628€
The new model performs very well when implemented in supply chains which have a good flow of information and coordination amongst supply chain partners. Therefore, it is the authors’ recommendation that, companies that do not have seamless information flows and high level of cooperation between supply chain players do not adopt the model in its current state, since the benefits relative to the additional complexity is small. It is also not recommended for companies that have short product life cycles since they will not have the long term data needed to run the model well.