7. Concluding remarks and future research
Over the last decade, academic research has begun to provide evidence that retail supply chain management requires a comprehensive view of all subsystems, from warehousing to the shelves. Cumulatively, these findings suggest that careful integrative planning considering the effects of all subsystems will result in cost improvement. The planning of delivery patterns connects these subsystems on a medium-term level. As order intervals are a result of the delivery pattern applied, volume effects along the supply chain occur that essentially determine overall logistics costs (Cachon, 1999).
In this study, we presented a modeling framework to explore the delivery pattern decision on retail operations and logistics. We derived a decision support model that integrates warehouse operations, transportation and instore replenishment. It comprehensively models the influences of the delivery pattern decision in all relevant subsystems, considers bundling effects across stores and reflects practical needs. The heuristic solution approach developed generates fixed clusters for all days of the delivery cycle, i.e., each store is assigned to the same basetour on each day deliveries are received. This is in line with the wishes of truck drivers and tour operators. They prefer stable and regular basetours as they need to be familiar with the characteristics of the tours and the unloading areas. Stable clusters and tours may therefore reduce short-term operational complexity and increase service quality.
We show that effective operations cost improvements are on average 2.5 percent and up to 7 percent in our numerical study using simulated data and 1.5 percent in the case example of an European grocery retailer compared to the most recent approach in literature.
Grocery chains operating their own transportation fleet can profit from the advantages of the solution approach suggested, while retailers relying on logistics service providers in transportation can strengthen their bargaining power, offering service providers delivery schemes with enhanced transportation routes supplying nearby stores on the same delivery tour.
There are some limiting factors of the model (A) and the solution procedure (B) from which further research opportunities can be derived.
(A) The model for example can be extended by considering a multi-level distribution structure. We assume one central DC supplying all stores of the retail chain. Some retailers, however, operate central as well as regional DCs storing multiple product segments that are delivered to the stores using different delivery schemes. The delivery of different product segments to stores requires a considerable coordination effort since the transportation loads have to be consolidated at transshipment points and the limited receiving capacity at the stores have to be regarded simultaneously.
Our deterministic dynamic model does not account for stochastic influences. This is justified as the model is applied on a medium-term planning level. However, as the delivery pattern also determines the length of the review period, the pattern selected influences the amount of safety stock, which in turn influences shelf restocking activities. On-shelf availability therefore may also be affected by the decision considered. The connection between determining delivery patterns and stochastic inventory management is a challenging field for prospective research.
Finally as discussed above, shelf refilling activities from the backroom are strongly influenced by delivery frequencies. This leads to the future challenge of integrating shelf space allocation with its dependencies on consumer demand into the model. Further research is also required to investigate the effects with different case pack sizes and order requirements as well as seasonal demand. Generally speaking, a better connection of consumer interaction at the point of sales with instore logistics and upstream operational activities seems to offer numerous opportunities for retail research.
(B) On the other hand, the solution procedure can possibly be enhanced by a subsequently performed metaheuristic (e.g., tabu search, simulated annealing, etc.). Our analysis showed that varying delivery clusters can potentially improve the solution further. Evidence thus exists that an improvement heuristic will upgrade the solutions, but only at the expense of additional computational time.
In addition, the approach suggested applies a strategy of “cluster first, assign delivery patterns second.” Other strategies can also be used, such as “assign delivery patterns first, cluster second.” In this case, however, the question arises as to how transportation bundling effects can adequately be anticipated when assigning delivery patterns to stores.