What is logistics optimisation and why is it important?
Logistics optimisation is all about trying to reduce operational costs, whilst improving service delivery and ultimately, customer satisfaction. This is not only important in last-mile delivery for e-commerce products, but goes even to the establishment of distribution centers, and targeted marketing strategies. It is also an under-appreciated weapon in response to changes in competitors’ network strategies.
Most common focus areas for logistics optimisation.
- Where do we have clusters of customers and/or suppliers?
- Do we need to open a new distribution center, and where?
- Based on driver characteristics, who should deliver what, where, and when?
- What are the optimal routes for one driver, or for a team of drivers?
Data Types commonly used in Optimisation systems
- Points: representing locations of customers, retail stores, competitors, etc.
- Lines: representing networks
- Distance: this will dictate the allocation of work and resources
- Time: is mostly important in last-mile deliveries, and is closely related to the distance. It can however be augmented with real-time road network data to have a better view of conditions that may affect the schedules.
How to go about with optimisation.
This involves 2 main steps.
The first step is to outline the problem as is, and the foreseen goal. That means taking note of what you are doing currently, identifying the drawbacks, and plotting your end goal. The goal might be to reduce the total amount of time spent on the road or to efficiently distribute drivers over a said location, or anything that makes the logistics component run smoothly. It at this stage, where you have to identify the constrictions in the network or operations as well as alternatives like building new distribution centers.
The second step involves applying mathematical expressions to the conceptual framework of the problem identified in the first step. This is done to provide an easy way for us to create the optimal solution (described below), to reach our goal
The Optimal Solution
The optimal solution can be found using be either Clustering or Efficient Routing, or a combination of both, (taking into consideration practicality on the ground). The solution can be found through the application of a series and combinations of algorithms. Over the next few weeks, we will dig deep into the specific algorithms to find the optimal solution for the business’s goals. We will go through the very basics, and then dig a bit deeper into the mathematics so that you can understand further, and know-how to tweak some on these concepts for your benefit.
At its simplest, clustering refers to the grouping of data points or objects based on their attributes or specifically, the values of those attributes. This means that data points or objects in the same group should have similar properties or features. The clusters can be developed based on features such as distances between each object, or the characteristics of data points. Clustering has been historically used mainly in epidemiology, criminology, precision farming, and customer segmentation.
Categories of Clustering
Spatial Clustering can be broadly categorised into 3 types:
- Partition Clustering: Groups objects into clusters such that any one object in a cluster more similar to the objects inside that specific cluster than to objects in different clusters. Each cluster will have at least one object, and each object will belong to only one cluster. It more commonly used in the retail industry by creating clusters based on sales activity, and in delineating delivery zones for a team of delivery personnel.
- Hierarchical Clustering: Is an intuitive and iterative process. It first assumes that each data point is in a cluster of its own, and then merges the most similar clusters which are close to each other. This is then repeated until we have the desired k clusters. This type of clustering is commonly used in epidemiology but has also been used extensively in customer segmentation for targeted advertising.
- Density-based Clustering: Creates clusters based on the concentration or density of the data points or objects. It groups regions of high concentration and separates them from sparse or empty areas. Data points or objects outside of any cluster are labeled as noise. This form of clustering is used in e-commerce recommender systems, to recommend products to customers.
There other types of clustering, such as fuzzy, and model-based, but we will mainly look at the above mentioned three. But I will write about the other types in later posts.
There are a number of methods that can be used to identify the most cost-efficient route for transporting products/ or services from point A to point B. This requires knowledge of the locations of your customers, depots or distribution centres and delivery personnel team, in near-real-time, where applicable. We will go over the algorithms in greater detail over the next few weeks, but some of the more effective and commonly used algorithms are the Dijkastra’s algorithm, the C-W Savings algorithm, and the Christofides algorithm.
Logistic Optimisation requires a certain knowledge of the problem at hand, and fully understanding the goal or target. This can then be translated into mathematical expressions which can be used to find an optimal solution that saves the business money, whilst improving customer satisfaction and growing the business.