Our last article addressed the temporal aspect of scaling logistics networks. The initial question was: How can you find out when you need to create logistics resources? This is the first challenge that e-commerce providers should address, as the right timing is essential for successful and sustainable growth.
The second important challenge concerns the spatial location of logistics resources. As a rule, the question of location follows immediately after the answer to the temporal aspect: "So we need another operational site in two and a half years from now. Where would be the best place to locate it?" In the second article of our four-part series, we take a look at possible solutions to this question.
Logistics deals with the planning, control, optimisation and execution of flows of goods, information and people. In e-commerce, this usually means flows of goods. Although the spatial aspect is also important for flows of people, it no longer comes into play as often in our field. What is easy to forget, however, is that spatial distances once also played a role in information flows. Of course, that is no longer the case today.
So, in e-commerce logistics it is mostly about getting goods flowing through the network. The network consists of nodes and the connections between these nodes. From an abstract perspective, goods are mostly stored and processed at the nodes and merely moved along the connections. Since the connections – routing decisions aside – result from the nodes, spatial decisions are limited to setting nodes on a map.
Not all nodes can be placed variably. End customers cannot be persuaded to move even with discount codes. Very few e-commerce companies today already have their own delivery network and therefore outsource part of their supply chain operations to partners. The same is sometimes true at the supplier end (which is usually considered secondary) where production sites are owned by third parties, so that their node coordinates are therefore predetermined.
So, if you have fixed points at both the inbound and outbound end, you will limit yourself to optimising your own network. There are two types of variables: those that result from the connections created and those that result from the node itself.
Obvious connection variables are, first of all, costs and time. A linehaul transport from Erfurt to Barcelona, for instance, costs both time and money. The two variables are, of course, also relevant to the node itself, although the cost aspect is certainly easier to understand. When it comes down to it, it will be cheaper to operate in the countryside than at an exclusive location in a conurbation. It goes without saying that wage levels, rents, taxes, investments and related interest rates as well as other aspects also have a role to play here. Other node variables include the catchment areas of workers and customers as well as other circumstances related to infrastructure or building law. In addition, the availability of suitable buildings or land zoned for construction in sought-after regions is steadily declining.
This means that the issue of where a new site is to be located appears complex even for a single site. Comprehensive due diligence and a sound valuation matrix are needed to make an informed decision.
For a large part of the node variables mentioned above, it is sufficient to conduct research or purchase information. The aspects of costs and time, on the other hand, are extremely specific to each case and usually have to be modelled explicitly. These models come in a wide range of complexity levels. We distinguish between models for visualisation and models for optimisation.
Visualisation models are limited to calculating the results from a limited range of possible solutions. The results are then included in the evaluation matrix mentioned above and support the decision-making process. They are particularly suitable for initial orientation, decisions when time is of the essence or when there are limited alternatives. Visualisation models do not require all too much work and effort.
Optimisation models, on the other hand, search for the best options from a very large range of possible solutions. They are mostly based on mathematical concepts and programming languages or special software solutions. This makes sense, especially for large networks with significant growth, when setting up multiple locations or for long-term strategic considerations.
Both types of models work through weighted connections and cost rates and search for the lowest sum (of all costs or all times) under the given conditions. This concept is also referred to as a centre-of-gravity analysis.
It should be noted that even a correct site analysis for the next site will not provide an optimal result in the long run, because networks evolve. Five sites determined iteratively will not be able to deliver the same result as five sites determined simultaneously. Of course, this problem cannot be completely solved, but it can be contained by extending the time horizon.
It goes without saying that this brief overview only scratches the surface of the problem. These approaches hold many interesting questions, such as how to determine a good proxy model (‘an estimator’) for the delivery time perceived by the customer or how to offset costs and times against each other. We would be happy to discuss this with you. You can contact us via e-mail, or on LinkedIn and Xing.