Interoperability in international trade is an old problem both in its implementation and in the mathematical theory on which it is based.
To simplify, we can express the problem as several countries wanting to exchange data with each other, ignoring for now an additional layer of complexity which is determining who has permissions/agreements to exchange data and what data is agreed to be exchanged. For now let's simply say that N number of countries want to exchange data, all with all others.
This is a well-known problem, connecting N nodes of a network, either directly or indirectly between them. This gives rise to different types of networks, on one hand we could have a centralized network, where there is a supernode that bridges between all the nodes of the network and on the other end a fully distributed network where all nodes connect to all other nodes. If we consider the complexity of the network directly proportional to the number of ties between nodes, then we have that the simplest network would be the centralized network, where there are N ties to link N nodes (N-1 to be precise). On the other end, the fully distributed network has N x N ties, each node has N ties to the remaining N nodes (N-1 to be precise).
What happens when the network grows?
Well, centralized networks grow very easily. Adding another node to our network is simply connecting that new node, which costs us 1 more link. That is, if we had a network with N nodes and we add another node… now we have N + 1 nodes or links. A fully distributed network, on the other hand, grows very differently. Adding a new node to a distributed network costs us N links. That new node must connect to the existing N nodes, thus adding N new links. This, seen specifically in the practical case of VUCE interoperability, is represented as the cost of negotiating and agreeing on the connectivity of a new VUCE to an existing network of VUCEs.
However, the benefit that centralized networks bring us brings with it another problem, the dependence on a centralized supernode that ultimately acts as a judge in the communication between the VUCEs. If this node were to fall, the entire network would fall. This node will ultimately be in practice in one country, so the network will ultimately be dependent on the regulations and laws of this particular country. Thus, centralized VUCE networks will be an exception.
But this only represents the cost of connecting two VUCEs to each other. We are not considering in this theoretical model the cost of having different types of documents, such as the Aladi COD, the ePhyto, eBL, various UPU documents, private sector documents such as invoices, insurance, purchase orders, among many other documents handled in the sector. In addition, there are customs declarations for import, transit and export. Each different type of document adds another layer of complexity, since the cost of connecting two VUCEs is proportional to the number of different types of documents that they intend to exchange.
This leads to the degree of complexity of the model being not only quadratic (NxN) but cubic (NxNxM). We then manage two variables, N the number of VUCEs and M the number of types of documents to be exchanged. This being the base cost For each node, this cost is proportional to the number of VUCEs I want to connect to and also proportional to the number of types of documents I intend to exchange.
The base cost problem
A link has a base cost, adding a new node to a cubic network quickly balances the cost/benefit. That is, the cost of adding a new VUCE to the network of N VUCEs contacted by exchanging M documents easily exceeds the benefit of the exchange.
So we see that the data exchange network between VUCEs has been taking place exceptionally, exchanging in a limited way between small groups of VUCEs that occasionally exchange only a few types of documents.
Interoperability projects based on models with cubic costs (exponential with exponent greater than one to be precise) are doomed to failure, because it is sooner rather than later that the cost-benefit balance is reached and from that point on, adding a new node simply costs more than the benefit achieved. They are victims of their own success - the more successful the project, the more quickly the cost of growth does not justify the benefit.
Condemned?
There is a natural limit to any solution proposed to this problem. The benefit is linearThat is, the gain from adding new nodes to these networks is simply proportional to the number of documents exchanged between that new node and the remaining ones. In real life, we say that the benefit is then proportional to the number of commercial operations that that country has with the others. But the number of commercial operations does not change because the VUCE is connected, it will remain more or less the same as before, maybe it will grow a little, but it is not significant for the problem.
Pillar of a solution: Moving from cubic cost to linear cost.
So for a solution to be sustainable over time, the cost of growing has to be fixed or at least proportional to the number of nodes. You can't put a 'per' in the equation. The cost of adding a new document or a new VUCE to the network, has to have a fixed cost which is not related to how many other documents or countries are already on the network. Or to put it another way, the cost of moving from 3 to 4 VUCEs has to be more or less the same as moving from 30 to 31.
The quadratic part of the problem of having N VUCEs connected to N other VUCEs has already been proposed as a solution, having the capacity to have a centralized solution. Obviously this involves the problem of the supernode and its delicate contribution to the solution. But with regard to the cardinality of the network of N VUCEs we could then say that a centralized network will serve the purpose of reducing the quadratic cost of growing a distributed network to a fixed cost.
Leaving aside the supernode problem, at least this part of the problem would be solved.
Now, we have another part of the problem. How to add new document types to the exchange with a fixed cost and not quadratic.
This is where the famous and never well-regarded enough comes in standardization of data and in particular the WCO Data Model.
El WCO Data Model is a mechanism where any type of document is represented using a finite dictionary of elements. Any type of certificate, license, permit or whatever (LPCO) can be represented as an instance of the OMA model to a fixed cost.
What is that cost? The cost of mapping that document to the OMA model and eventually growing the OMA model to encompass the new documents that need to be represented.
But once this work is done, it is done only once for each new type of document to be exchanged and it does not depend on the number of documents already exchanged. In fact, experience shows that the cost is significantly reduced the more documents have been mapped. That is then, a fixed cost.
The supernode problem
We have left one problem pending, the weakness of a centralized network based on one or more supernodes that act as judge and jury in the distribution of information. This problem can again be overcome, using the same tool that we used to reduce the problem of adding new types of documents to the network. Standardization!
In this case, not applied to the data, but to the services on which these exchanges are based. For this purpose, the UN/CEFACT It offers a standardization of services focused on enabling portability in IT services. The proposed standardization opens up the possibility of having the benefits of a centralized network without the risks that a supernode implies.
In the particular case we are trying to solve, unlike in network theory, nodes (VUCEs) do not appear or disappear frequently. It is, we could say, an exceptional milestone in the life of the network. That is why standardized computing services, commonly called APIs, allow us to assume that connecting to one or N nodes has the same fixed cost, which is to implement standard network services. UN/CEFACT he proposes a API standard so if all VUCEs implement the same services, then connecting to one or all VUCEs is ultimately a fixed cost.
We are working on this theory in the module of Customs-hub interoperability 100% WCO Data Model, 100% UN/CEFACT, 100% NO-CODE.
The author is a Systems Engineer with specialization in Foreign Trade, Customs and Single Windows. Recognized as an expert by the World Customs Organization (WCO) and accredited by the United Nations Center for Trade Facilitation and Electronic Business (UN/CEFACT).
He has developed his experience in the implementation of the World Customs Organization Data Model in Latin America, being director of the consultancy Customs-hub, a company dedicated to evangelization and support in the implementation of the WCO Data Model in the governments and private sector of Uruguay, Argentina, Brazil, Costa Rica, Honduras and Panama, among others.
He actively participates as a private sector expert in the WCO Data Model Projects Team and collaborates in the evolution of version 4.0 of this data standard. He participates in UN/CEFACT as an expert in the domain of Single Windows.
