Today’s blog article describes an extensive undertaking in analyzing international trade data. The work has been done by Ahmad Fahmi at Altares Dun & Bradstreet in The Netherlands, supervised by myself and Mike van Kessel. Ahmad combined his work at the company with finalizing his MSc education in Operations Management and Logistics at the Eindhoven University of Technology.
Why Identify Risks In International Trade?
While the majority of world trade is legitimate, some import/export shipments disguise illegal activities. A report of the World Economic Forum from 2015 estimated the cost of counterfeiting and piracy alone at $1.77 trillion in 2015, which is nearly 10% of the global trade in merchandise. Customs and Law Enforcement agencies have the task to find the needle in the haystack, i.e. detect these and other illegal shipments, while minimizing the impact on legitimate trade. The road to achieving this goal starts with identification of risks in international trade. Identification of risks is important also for all commercial participants in the global supply chain. Companies want to reduce the risk of reputational damage which can be the consequence of them being involved (even unknowingly) in illegal operations of third parties. How can companies reduce the risk of exposure to such risks, and how can Customs increase their ability to detect high-risk shipments?
The Approach: Finding Patterns in Correlations Between Cargo and Company Information
The approach taken by us was to analyze correlations between product / cargo information and the companies involved in the shipments. For example: a shipment of bananas, where the shipper is an electronics company, does not make sense.
Earlier Work In This Domain
Bananas and electronics, that’s an easy example, one can say. It’s easy to define such a (negative) correlation. Also, one may think that this approach is not new, because it’s common practice in existing risk management systems. Finally, you may say that others did similar work in the past, and that’s true too. But with certain limitations which we aimed to address.
The complexity in defining what makes sense and what does not make sense stems from some complicating factors:
- There are many possible products in international trade
- There are many types of companies (combine this point with the previous one, and you get a problem of exponential complexity). This might be a reason why earlier work focused only on shipments of manufacturing companies, and thus does not suffice. Namely, service companies (e.g. wholesalers, retailers) account for a very large percentage of trade.
- Often it is hard to detect in trade data who are the involved companies
- Often little is known about companies, especially in remote countries
- Companies can have complex structures. Therefore,
- one needs to understand the broader nature of the involved companies, in order to understand which correlation makes sense.
How Our Work Addresses These Challenges
We collected very large datasets of trade data from around the world, analyzed them, enriched them with Dun & Bradstreet business information about the involved companies worldwide, and subsequently analyzed the combined datasets of trade data and company data. In doing so, we applied several unique techniques:
- We used the Referential Matching technique to improve the probability of correctly identifying companies involved in trade, complemented by some manual matching activities.
- We used the rich Dun & Bradstreet B2B database to identify the broader characteristics of trading companies, as can be derived from D&B’s knowledge of the family trees of companies. For example, we came across a shipment of vehicles by a PR company. By itself it did not seem logical, and we flagged it as an exception. However, when we learned – by enriching the dataset with additional B2B information – that this PR company was part of an automotive company, the shipment made sense.
- We included in the analysis not only manufacturing companies (as done in previous efforts), but rather any type of company. By doing so we ensured that our work covers the full scope of international trade.
Results: A Means For Risk Identification In International Trade
By taking the above approach, and applying it to very large datasets of trade from around the world (i.e. not biased to a specific trade lane), we were able to guarantee the objectivity of our analysis. We created a set of correlation rules for analyzing trade data worldwide such that we can identify risks based on a combination of:
- The type of goods; and
- The identity of the companies involved in a trade shipment; and
- Rich company information about the involved companies; and
- The correlations that we identified by applying data science techniques to large datasets.
Implications For Stakeholders In The Trade Ecosystem
The work described here is of benefit for all parties in the supply chain
- Customs and Border Protection: improved ability to detect high-risk shipments (find the needle in the haystack)
- Logistics Service Providers (carriers, freight forwarders): reduce the risk that they service third parties performing illegal activities
- Shippers and consignees / buyers and sellers: mitigation of compliance and reputational risks associated with being used for masking illegal activities
If you want to learn more about how these insights can be used by your organization, contact me.
- Report Raises Concerns On Governments’ Will To Fight Illicit Trade
- Big Data For Customs At The Borders: Start Small, Think Big!