Trade processing: going digital

Technological innovation and digitization have transformed many aspects of our personal and business lives. In the financial world, companies nowadays expect digital data at their fingertips. However, efforts to digitize trade flows – despite enormous investment and commitment – have achieved only partial success to date. Why is this and what can be done?

• Government regulations associated with trade are significant in scale and vary between countries, resulting in processing complexities. Some countries, for instance, may require multiple paper copies of a document with official stamps while others may need a manual signature. Such a range and variation of regulatory requirements makes it difficult to introduce new technologies that are already in use with other financial products, such as digital signatures. 

• Unlike other parts of the banking industry where there are a limited number of parties involved in a transaction, trade transactions encompass importers, exporters, multiple banks, customs authorities and shipping companies, to name a few.  As such, trade documents are generated in large quantities and by multiple parties. 

• Many documents associated with trade are unstructured, meaning that pieces of information required to process the trade could appear in a variety of formats or locations within a document. Because of the regulatory complexities and various parties involved in trade transactions, developing universal formatting for supporting documentation is unlikely.  This is in contrast to the documents used in many other parts of the financial services industry, such as credit card applications, which are fairly standardized. 

Efforts to digitize the trade process have tended to be wide-ranging and ambitious. The goal has been to digitize the whole flow from start to finish, and although exciting developments are taking place, including the use of new technologies such as blockchain in the market, these solutions are still some way off achieving real traction, never mind wide market acceptance. While banks continue to work towards conceptualizing the ‘ideal digital solution’ of the future, some parts of the process will clearly continue to be paper-based for the near future, and therefore banks should seek to digitize core processes themselves and not develop solutions that require all parties in the transaction to provide fully digital data.

At the processing centres of many major trade banks, documents are imaged, categorized, undergo sanctions checks and are read for processing: these manual steps are often repeated given their importance and potential for regulatory fines. To facilitate a letter-of-credit-based trade, 25 to 40 data points are typically required: these details are manually extracted from the various documents delivered to the bank and entered into the bank’s trade-processing system.

Although this process is more efficient than using paper-based documentation throughout, it remains time-consuming and expensive.

A newly developed alternative is to take the electronic images of those documents and use optical character recognition (OCR) and named-entity recognition (NER) software to determine the type of trade document and perform sanctions screening. Artificial intelligence software enables a machine to read the information located in each document and identify names of parties, ships and goods, then extract that information and enter it into trade-processing systems.

The documents are still forwarded to processing teams but their role is to review and verify and act as quality control, which speeds up processing time considerably. Any anomalies or potential risks are highlighted by the software and checked by an operator, or manually entered if there are omissions. 

This combination of technologies is already proven to deliver results in time and cost saved and risk management. The objective now is to improve processing accuracy and accelerate fraud and sanction screening times, which can be hampered by false positives. The ideal model is similar to that used by credit card companies, which do not just depend on spotting keywords but use sophisticated pattern recognition to identify unusual activity in cardholder spending. Over time, the feedback loop provided by this data will allow the learning software to increase its accuracy and therefore help to improve the speed of processing.

For many companies, trade remains a relatively slow, manual process with numerous variables. The benefits of these new technologies will be significant for all trading parties. 

It is important to remember that most trade transactions involve physical movement of goods. Any problems associated with documentation must be resolved by the time the cargo gets to its destination or the shipper will be hit with demurrage or other charges. The use of OCR, NER and machine learning can help to improve turnaround times, especially during peak processing windows, and reduce costs and days sales outstanding (DSO) for companies.

The big data produced by this approach will mean that more information about trade transactions is available to companies at a strategic level. Depending on the banking platform used, information about the progression of an individual transaction can also be accessed via a tablet or phone in 24/7 real-time. 

Many leading banks are investing time and money into potential future solutions to digitize trade flows end to end. A bank that has already digitized its core processes will be ready when the market moves towards a truly digital solution. In the meantime, the use of artificial intelligence will deliver tangible benefits to corporates; the first steps have been taken, and as this technology leaps forward there is still much more that we can do. 

^{This article is for information purposes only and does not constitute legal or other advice. The information contained in this article is believed to be accurate, but Citi makes no representation or warranty with regard to the accuracy or completeness of any information contained herein. Citi is not liable for any consequences of any entity relying on this article.}