In my previous blog in this series on the future of land ports, I looked at how to apply smart, digitally-enabled segmentation of goods passing across the border. In this latest post, I shift the focus to the inspection of those goods to catch fraudulent or illegal shipments. It’s an area where smart targeting using digital technologies is opening up opportunities to enhance speed, efficiency and “hit rates” for catching fraudsters.
It’s also an area that goes to the heart of the customs agency’s mission. Across the world, every customs agency has a responsibility to protect its nation against crime while optimising the free flow of legitimate trade. This involves applying an overall degree of control, while going to deeper levels of inspection depending on how suspicious a consignment appears to be.
However, what all agencies want to avoid is an intrusive manual inspection where a customs officer actually opens the goods. This is not only slow and expensive, but also raises the risk that the goods might degrade or be damaged, especially if they consist of perishable foodstuffs or fragile electronics devices. For these reasons, agencies try to rely more on non-intrusive inspections.
Current inspection technologies: costly and hard to scale
A wide spectrum of technologies are currently available to support non-intrusive inspections, such a X-ray machines and high-energy scanners. However these devices are expensive to buy and operate, and can have very challenging safety requirements – all of which makes them difficult to scale up as trade volumes grow. As a result, more and more agencies see “smart targeting” as the way forward.
How does this work? To target illicit goods more accurately for inspection, an agency needs to have a risk engine that ingests a wide range of data about each consignment prior to its arrival. This data encompasses everything on the customs declaration, supplemented by information from various other sources – including data on the broker, and intelligence shared by other national and international trade bodies.
Based on this diverse data, the risk engine assigns a risk score. A score of zero might mean the goods are given direct clearance through the ‘green lane’, with minimal inspection – while a score of 1 might require an X-ray, 2 could be a scan with a high-energy device, and 3 may merit a physical inspection by an officer.
By applying this process, a customs agency achieves two objectives at once. First, legitimate traders are inconvenienced as little as possible, since they will usually be directed through the green lane. Second, the agency is able to optimise its use of resources and customs officers as trade volumes grow, by directing its inspection efforts only at the small minority of shipments that are truly suspicious. This targeting also means that when goods are inspected, the hit-rate for uncovering criminal activity is higher.
Three layers of technology
So, how does the technology to enable all this work? It essentially consists of three layers. The first is data ingestion – where it’s vital to have the right technology to ingest the right data with the right degree of accuracy. This requires integration with data from sources such as IoT-connected smart sensors, and with international bodies and other countries in the economic zone. Key technologies here include integration platforms and data lakes, helping to ensure the data is clean and stored in one place.
The second layer is a new level of knowledge developed through machine learning (ML) and – as the risk engine matures – artificial intelligence (AI). The ML combs through the history of fraud incidences at the port and elsewhere, and identifies combinations of characteristics that indicate a high likelihood of criminal activity. As a basic example, if someone says they’re importing bananas from Germany then a consignment probably merits a closer look.
The third layer is to integrate the risk engine with the port operations and introduce AI. As officers act on the outputs from the risk engine to conduct inspections, a pool of experience and feedback is built up showing how accurate those outputs are. The AI learns from this experience to fine-tune the risk engine and make it more accurate over time. So every time an officer feeds back after an inspection, the engine gets smarter.
Navigating the journey
The advance to smart, data-enabled targeting of inspections is a journey that many land ports in the Middle East region have already begun. The approach we recommend is to start by running the risk engine in “silent mode” as it learns and as the data improves in both quality and range, and then transfer the engine into live operation once it reaches the right level of accuracy.
We know of several ports that are now approaching this switchover. A key factor in making the transition successfully is the quality of the data: as with other systems, the mantra with smart targeting is “garbage in, garbage out”. So high-quality data is imperative.
As I mentioned earlier, one of the benefits of getting smart targeting right is that legitimate, trustworthy brokers are directed down the ‘green lane’ for direct clearance. This links to issue of how to manage the different lanes to maximise both security and throughput. So I’ve chosen this as the topic for my next blog, which will focus on ‘dynamic lane management’.
Disclaimer: This content is provided for general information purposes and is not intended to be used in place of consultation with our professional advisors.
Copyright © 2021 Accenture. All rights reserved. Accenture and its logo are registered trademarks