Customs agencies around the world are under immense and growing pressure at a time of rapidly increasing cross-border e-commerce, shifting regulatory environments, and supply chain disruptions, including repercussions from COVID-19 and the war in Ukraine. The good news is that agencies also have an expanding number of digital tools that can improve performance by identifying and tracking issues before and at the border—and even once goods have left.
Deployment of advanced analytics could make a significant difference in a number of use cases. Among other applications, analytics can radically improve fraud detection, minimize revenue leakage, and bring new transparency to audit coverage. In short, if the work of customs agencies is to find needles in haystacks, machine learning and other data analytics tools could both magnify the needles and shrink the haystacks, making detection faster and more reliable.
The rethinking of risk management strategy
The customs operating model is already stretched, and that tension is likely to grow. Cross-border e-commerce alone is projected to grow from $300 billion in 2020 to $1 trillion by 2030.1 This will increase declaration volumes and could expose customs agencies to unprecedented levels of security and revenue risk.
Moreover, the COVID-19 pandemic has prompted many companies to rethink their supply chains and the locations of their physical operations. All this has coincided with heightened friction in global commerce as trade disputes have escalated. As a result, customs agencies have made developing new risk management strategies a key consideration.
Some customs agencies are not up to date in their risk management. One leading global customs player used a risk engine driven solely by human input to identify potential risks in imports and, in doing so, would deploy risk profiles that were many years old. When a random inspection exercise was conducted, the agency found the number of violations to be 20 times larger than predicted by the existing risk engine.
Similar limitations in terms of revenue management also became apparent: historical performance, rather than an understanding of up-to-date insights, was used to set post-clearance collection targets. This all led to realizing revenues that were below their full potential. It also prompted excessive auditing that ultimately burdened operations during a time of already heavy demand on customs agencies.
The benefits of deploying advanced analytics
A range of advanced-analytics techniques are being deployed with increasing success. Machine learning in particular can be used to train machines to sift through enormous volumes of data to spot patterns and anomalies, including potential fraud, which is particularly pertinent for customs agencies.
The potential applications of such technologies extend across the trade journey—that is, they can be deployed before the border, at the border, and after the border (exhibit).
Before the border. Advanced analytics can help agencies obtain information about traders early in the value chain. Using existing digital tools such as Microsoft Power BI and Tableau can provide customs agencies with a dashboard for accessing declaration data for all shipments. Analytics can help detect potentially fraudulent importers using historical data or even information from within traders’ commercial supply chain systems, including their transport management systems and manufacturing execution systems. Techniques such as natural-language processing can comb through large amounts of text data from declarations and detect anomalies that could help identify illicit trade as early as possible. Such programs can instantly flag suspicious activity—for example, if the trader is a car company but the good being imported is a bed. Natural-language processing can also support traders by giving them tools to ensure they are less likely to make mistakes on their declarations, such as by allowing them to identify the correct commodity code based on a few questions or free text.
At the border. Use cases at this step include auditing at the border, which can identify manipulation of commodity codes or goods valuation. Analytics can also improve operational performance in areas such as workforce planning, health and safety, and performance assessment of auditors. For example, advanced analytics can direct customs officials to open the right consignments using historical data and inputs from early-warning systems. And analytics can be a potent tool for strategic workforce planning, including in matching workforce schedules with demand.
After the border. Use cases at this step are about identifying and addressing revenue leakage. One case study from a G-20 country highlights the potential upside of analytics. The customs agency in that country was seeking to improve its risk and revenue management by ramping up its audit function and implementing a new targeting team. It recruited about 200 auditors and started conducting about 2,000 post-clearance audits annually, most of which were cases that had been incorrectly identified as compliant.
The agency, quickly realizing that the rate of detected violations was very low, moved to strengthen its risk-targeting engine by building two machine learning models using advanced analytics. The first was a “supervised” model that learned from past audits by selecting similar noncompliant cases and excluding any compliant cases. The second was a more sophisticated “unsupervised” model. This identified noncompliant cases that differed significantly from what was expected, essentially flagging anomalies that had previously gone undetected.
The upshot. After implementation of the post-clearance audit models, the detected violation rate doubled from 30 percent to 60 percent, and the agency’s workforce productivity jumped by 75 percent. In all, the customs agency was able to achieve a 15-fold increase in revenue per auditor per year.
Customs agencies can consider deploying advanced analytics in the heart of their operations now. The World Customs Organization, for one, is committed to it. Its priority for 2022 is “scaling up customs digital transformation by embracing a data culture and building a data ecosystem.”2
In the European Union, initiatives are being tested to use advanced-analytics tools to improve customs risk management practices. A project called PROFILE aims to facilitate and accelerate customs agencies’ advanced-analytics capabilities, including the incorporation of external data sources to enhance risk profiling of imports.3 Customs agencies under the program can access data owned by big data providers as well as e-commerce websites.
According to an analysis by the World Customs Organization, the proof of concept for this project is being rolled out in Belgium, the Netherlands, and Norway, among other countries. In the Belgian “living lab,” where the testing is taking place, analytics tools are being used to establish risk indicators for traders. In the Dutch living lab, price information is being collected from peer-to-peer online marketplaces and web stores and compared to average prices in e-commerce declarations. And in the Norwegian living lab, import and export risk is being assessed at the border through analysis of trade data.4
Adopting advanced analytics can be challenging, and many consider it an aspiration for the future, rather than something that can be achieved right away.
One important myth can be challenged: customs agencies do not need perfect data to start their advanced-analytics journeys. They can start by leveraging the data they already collect. Our analysis suggests that many customs agencies could experience the benefits of use-case pilots in as little as 12 weeks, with a tremendous potential impact. In terms of magnifying needles and shrinking haystacks, that is a very short time indeed.
This article was originally published in Border Management Today.