Spinning a Web of Space and Time
8/12/23
Supply Chain Complexity
Supply chain networks contain a plethora of intricate webs of relationships, processes, and flows that enable goods and services to move from raw material origins to end consumers. When analyzing how a supply chain network functions, you must consider a myriad of complicated entities – manufacturers, suppliers, logistics providers, retailers and more – each interlinked through a sequence of transactions and dependencies.
The dynamics involved in a supply chain network, such as fluctuating market demands, environmental concerns, and geopolitical events make managing and optimizing these networks an interesting dilemma. Unique technology must be used to attack these problems efficiently. As the era of tracking assets, movements, products, and relationships in ledger-based rows and columns grow more complex, we turn our attention to confronting these challenges directly with the emergence of Space and Time networks.
What is a Graph Database?
While relational databases have long been the gold standard for structured data storage and retrieval, their tabular nature can make them less efficient when dealing with highly interconnected data. Graph databases, on the other hand, are designed for interconnectivity, making them a preferred choice for problems that deal with how entities move and not what specific items are in motion.
A Graphing Database uses a series of nodes (entities) and edges (relationships). Each, not only storing properties that describe the data they represent, but also describe how their data interacts with other data. The how is the key element that allows one to model interactions both inside and outside of their network to help figure out why elements interact the way they do.
Graphing databases gain a huge advantage over tabular, structured databases when dynamic relationships are the primary importance of interconnecting datasets. This makes graphing databases ideal for recommendation systems, network analysis, and in our case, complex supply chain networks.
Figure 1 – Connecting Dataset Company Nodes with Warehouse and Ports
Space and Time Networks in Supply Chain
Supply chains, by their very nature, traverse across spatial (space) and temporal (time) dimensions. The movement of goods and information doesn’t just follow a sequence; it unfolds across geographies and over time. Capturing these dual dimensions in analytical models is crucial to gain insights into supply chain dynamics. Consider the following:
Spatial Dynamics: Every entity in a supply chain either occupies or traverses a spatial location. These spatial locations affect and sometimes inflict challenges on the most resilient supply chains. External factors like weather, transportation costs, accessibility, and regulation need to be considered to ensure that supply chains thrive and operate efficiently.
Temporal Dynamics: Supply chains are not static entities. Orders, shipments, and production run on ever evolving schedules. Costs and demand depend on seasonality and ever-changing geopolitical events, in turn causing temporal dynamics to evolve as well.
Traditional analytical tools often struggle with this type of multi-dimensional problem but putting these attributes and constraints into a network format gives us an opportunity to more intuitively model, analyze, and optimize our supply chain problems.
In essence, the supply chain isn’t just a sequence of processes and events – it’s a dance of goods, information, and money across the globe, choreographed by the rhythms of time. Learning to dance in a network format will unearth insights that were often obscured with tabular data.
Figure 2 - Calculating Dynamic Temporal and Spatial Relationships
Analyzing Space and Time with ArcGIS Knowledge and Neo4j
At ESP our tech stack consists of many different types of databases: structured, unstructured, graphing, geospatial, and vector. For this paper, we will just focus on the graphing and geospatial database technologies and will discuss some of the others in future articles.
We have partnered with ESRI in our endeavors and make use of their spatial technologies in our Space and Time network, specifically ArcGIS Knowledge with a Neo4j managed datastore. ArcGIS Knowledge is a platform within the ArcGIS ecosystem that focuses on graph databases and graph theory. It enables users to create and manage knowledge graphs for spatial data. By integrating with other ArcGIS products such as ArcGIS Pro and ArcGIS Online, ArcGIS Knowledge provides organizations with a comprehensive view of their data. It allows users to combine data from various sources, including maps, imagery, and other geographic information to create interactive maps and perform real-time data analysis.
Figure 3 - Spatial Representation of Warehouses Connecting to Customers. The relationship is colored by the number of orders each customer has made from their respective warehouse. Yellow representing a smaller number of orders and Red representing a larger number of orders.
By using an ESRI component like ArcGIS Knowledge in conjunction with a graphing database like Neo4j, we gain the competitive advantage of a wide range of data sources, including maps, imagery, and geographic information. This integrates with other ArcGIS products, such as ArcGIS Pro and ArcGIS Online, enabling users to create interactive maps and perform complex geospatial analysis.
Figure 4 – Leveraging spatial technology gives us a unique view into how our supply chain is laid out geographically. In this figure, we perform a hotspot analysis on our customer nodes, analyzing our most and least densely populated regions where our customers are located.
A Web of Webs (Network of Networks)
One of the most exciting things about using graphing databases is that they not only connect similar nodes to each other, but they can be configured to interact with each other.
Take, for example, a cargo vessel traveling across the ocean. We can define a cargo vessel as a ship node that typically interacts with a port node. The relationship that connects the two can be defined with how they interact. At a macro level, the relationship can be defined as a shipping lane, because this defines at a macro level how ships interact with ports.
Figure 5 -Shipping Network Observed Geographically, with nodes and their relationships
Those are built using historical data, data analysis, machine learning and temporal relationships such as seasonality.
This is a very powerful network to have, but it is rendered useless after cargo is dropped off at a port. To have visibility after the port, we need to bring in other networks such as customer, drayage, and warehouse networks.
Figure 6 – Geographical representation of Customer Nodes, symbolized in Red, and the relationships that connect them, symbolized in Green. Notice that we transformed our relationships between the two into routes, so they not only represent a relationship symbolically, but also geographically.
If it seems complicated, that’s because it is! Just these datasets mentioned have numerous spatial, numerical, and temporal relationships that change swiftly over time. However, by storing them in a graphing network, these complexities are streamlined into interconnected components to shape your network.
Let's take a closer look at how this vast array of data can be distilled and represented:
Figure 7 - Network of routes, warehouses, and customers
In this foundational network, you have all your nodes, their connections, interactions, and temporal changes, offering a clear understanding of your supply chain's overall structure. And remember, this is just the beginning.
The intricacies of modern supply chain networks, underscored by their intertwined spatial and temporal dynamics, necessitate a transformation in how we conceptualize, analyze, and optimize these systems. Gone are the days when traditional tabular data representations served our every need. The advent of graph databases, as spotlighted in this paper, provides a new lens through which we can capture the interactions, dependencies, and fluctuations inherent to supply chains. The intricate choreography of goods and information on a global scale demands tools adept at capturing its nuances. Graph databases, as outlined here, are leading this transformative march.
If you would like to investigate some of the technologies that were mentioned in this paper, here are some of the sources you can look at.
AUTHOR
James Burns, Director of Data Science
James Burns is an experienced Air Force veteran, Department of Defense contractor and Data Engineer with 14 years of expertise in analyzing unstructured data using diverse collection methods to develop solutions for complex problems. At ESP, he spearheads data collection efforts aimed at enhancing problem-solving for diverse problem sets, with a specific focus on supply chain and space-time network analysis. James’ role involves overseeing the team's data gathering activities and developing innovative solutions to complex issues.
About ESP
ESP was created by combining over 20 years of supply chain and logistics expertise with technologists that have focused on the science of location and location analytics, the practice of layering geographic data along with business data to extract valuable insights. The advantage that our customers have with ESP is the ability to access and capitalize on a continuously growing infrastructure that enables more focused location-based analytics, helping the supply chain become smarter. ESP has designed a Space and Time Network which enables data (real-time, historic and near real-time) to be consumed, analyzed, viewed, and translated into action. Critical to our customers’ success, our network enables the analyzation of data sets that are beyond typical supply chain data, resulting in more accurate and actionable information.