A featured contribution from Leadership Perspectives: a curated forum reserved for leaders nominated by our subscribers and vetted by the Construction Tech Review Advisory Board.
HSBC
Too Much Data? Apply Intelligence Techniques to Financial Crime Risk Data Analytics.
With the recent advances in computer power and storage capabilities, the Banking Industry has become even better at something that has traditionally been mastered since it started to use computers: gathering data.
Data is the blood that flows and feeds any company, major or small. We use it to measure our efficiency, losses, and gaps. In the narrower context of Financial Crime Risk Management, data is used to prevent, detect, monitor, and mitigate the risk of criminals using the a financial institution’s products and services, either by detecting anomalies hidden within thousands of transactions or screening across large datasets looking for matching names, keywords or even news.
Data collection: let’s get everything
It does not take long for a big financial institution to collect vast amounts of data related to their customers and how they behave (payments, purchases, transactions), the numbers are simply staggering: 90% of all data in existence today was created in the past two years, 80% of data will be unstructured by 2025.
While storing these volumes of data might not be a problem for the analyst community, it is retrieving what constitutes the problem. Typical of any catch-up dynamics, financial crime analysts are always trying to identify patterns where illegal transactions hide. To make things worse, requirements can also be very vague or broad i.e., “What can we find on human trafficking in these markets?” This sometimes ends in huge data collection efforts, where time and resources being wasted, trying to develop data insights that will likely not answer the question. Technology can help processing data, but it’s humans who need to decide first what to look for.
Financial Crime Analytics can learn from Intelligence practices in order to help to solve this problem.
Intelligence Requirements
A series of primary intelligence requirements (PIRs) should be developed, agreed, and shared across the Risk function and Business Units in order to help to define those requirements and narrow both response and delivery times.
Narrowing the initial question to a series of known and tested typologies is the first step to optimize time-dependent processes, resources and reasonably maximize chances for success in delivery and answering the question “so what?”
PIRs should be aligned to an existing Financial Crimes Typology Library and Data sources inventory so it’s easier to decompose a potentially ambiguous question into elements that can be easily mapped to existing areas of knowledge.
A series of primary intelligence requirements (PIRs) should be developed, agreed, and shared across the Risk function and Business Units in order to help to define those requirements and narrow both response and delivery times
- Primary Intelligence Requirement: Provide analytics support to prevent, detect, and manage the finance of slavery and trafficking.
- Typologies:
• Human Trafficking
• Money Mules
- Related Typologies:
• Terrorism Financing
• Drug Trafficking
- Data
& Systems:
• Customer
• Transactional
• Transaction Monitoring
• Negative News Screening
The above steps can be further expanded to include specific data fields or filters. In this way, a question as broad as “what do we have on human trafficking in this market for the last 6 months?” could be quickly put against a framework where analysts know which systems to query and data to download and analyse.
While it seems a simple, common-sense exercise, the reality is that in a fast-evolving environment where some resources along the chain might not necessarily be financial crime experts. This simple framework that defines initial steps in the right direction can save time and help getting close to the intended mark. In a recent exercise, the author of this article requested a list of transactions, only to be presented at the end of a long extraction process with debit only data. While it may seem logical to think that the term “transaction” would also include credit ones, it clearly was not for the operator, who had recently transferred to the team and wasn’t familiar yet with financial crime analytics.
Dissemination& feedback: how was it?
While “Know Your Customer” (KYC) is an integral part of financial crime risk and control processes, knowing the internal customer who will receive the analytical product sometimes seems not to be the case. Too often, reporting templates are developed to display the same metrics that might have been relevant once but could have easily seen its scope diluted over time.
Analytics teams need to be able to understand their audience and tailor the product not only to the required specifications but also manage contents so the recipient can quickly understand and apply the insights to the decision-making process or incorporate them into the relevant knowledge base. For example, an analysis of Transaction Monitoring scenario effectiveness will likely include a broad series of conclusions and recommendations backed up by extensive technical data on customer segments and escalation rates. On the other hand, an analysis to identify money laundering exposure on key markets intended for senior managers should include short and concise messages highlighting areas where risk might be outside appetite and where thresholds have been breached.
Feedback is essential to prevent repeating mistakes and improve efficiency. The best learning occurs through mistakes, but success lies in identifying what went wrong and be flexible enough to change direction.
Never too much data
In a fast-moving data-driven world where every footprint produces a piece of information that can be quickly stored, it’s easy to get lost and miss targets when expectations are high and requirements unclear. Technology is an amazing enabler in the fight against Financial Crime, but it is still up to humans to avoid being overwhelmed by huge volumes and build a knowledge framework where data can be harnessed and maximized to support decision making.
