By now, everyone and their dog has seen the true-crime thriller Dirty John on Netflix, but just in case you’re the exception, here’s a quick summary: smart, successful woman meets a charming man with a suspiciously empty past. Red flags start popping up all over the place, but each time, the heroine gives her charming man a chance to explain them away on his own terms.
Until, at last, it becomes blindingly clear that the charming man is a criminal mastermind with a very dark past, spanning several states. “Why didn’t you do a bit more due diligence?!” you hear yourself cry, as you watch this unfold. “How on earth isn’t it obvious that this man cannot be trusted?” But that’s the problem, isn’t it? When you let a person who can’t be trusted decide what data you use to assess them, when you don’t look beyond this to external sources that may give a better picture, you’re very easy to deceive.
For banks and financial institutions, this need to know who you’re dealing with - i.e. to Know Your Customer (KYC) - is a constant worry. We all know, by now, that it’s up to us to get to know our customers and identify any dodgy practices before we risk (inadvertently) becoming a party to their crimes. But how well can you really know your customer if you’re only going off the information they provide? If you’re only looking at a narrow pool of data? If you aren’t looking at the bigger picture, or beyond state lines? You need a way to get complete visibility over all the data that’s out there that relates to your customer and their past dealings. And for that, you need a risk modeling platform fed by relevant, external data.
Successive directives and regulations, including the USA PATRIOT ACT and the EU’s various Anti-Money Laundering (AML) efforts, have placed the responsibility for identifying these financial Dirty Johns firmly on the shoulders of their creditors and suppliers. For example, financial providers are expected to undertake extensive due diligence programs to identify beneficial ownership. The aim is to ensure these companies aren’t flouting government sanctions or working with criminals masked by shell companies, for example.
In recent years, governments around the world have worked on issuing comprehensive guidelines that help businesses prevent criminal elements from using them for money laundering. Using procedures derived from these guidelines, organizations find ways to better understand their customers and their financial dealings, helping them to assess and manage the risks posed by these customers effectively.
These clarifications and recommendations include advising businesses to beef up their know-your-customer (KYC) efforts using AI and machine learning.
All of this represents a big step up from the vetting processes of the past, in which institutions would rely only on the information provided by the customer to assess whether or not they posed a risk. However, to do it right you need powerful technology on your side.
That’s why, today, many companies in the financial sector use some form of risk modeling platform to assist with their KYC responsibilities. These make it faster and more accurate to identify anomalies or unusual trends in transaction data, automating tasks that would otherwise interrupt business processes, and inconvenience customers who have done nothing wrong.
At a basic level, this may involve diving into time-series data in order to identify suspicious or unusual account behavior which is consistent with money laundering activities. However, this often leads to false positives that can undermine the ease of doing business. To make these risk assessments more accurate, you need to bring in contextual data from external sources.
This external data helps you to understand more deeply the transactions you are looking at and whether they fit with your customers’ line of business. You simply can’t get all that information from your internal data alone. You need to look to other sources and datasets in order to build a complete picture of your customer’s business and sector, how they usually operate, and whether this type of transaction makes total sense or is a sign that something is amiss. Feeding this relevant, accurate external data into your models helps you get to the right answers with confidence.
The scope of information you could choose to explore is enormous and depends on the needs and characteristics of your customers’ industry. However, a few types of external data you might feed into your models include:
Are they linked to any senior political figures?
Does analysis of news media and/or social media reveal negative mentions, reviews, or other clues that this person may have earned a poor reputation? Or that they aren’t who they say they are?
These compile official lists of entities and individuals that are subject to sanctions.
Many crucial pieces of information that help you assess the trustworthiness of a potential customer already exist in the public domain. You just need to know where to look and how to use the data to draw out the relevant insights. External public datasets relevant to your KYC checks could include: regulatory registration status, mailing address and principal business address, ownership and beneficial information (both from registries/regulators and alternative sources), legal name of customer, mortgages on assets, classification and records of criminal activity, fines and lawsuits — especially when these run-ins with the law took place in a different country or jurisdiction.
Broader contextual data that reveals patterns in the behavior of your customer’s business sector, helping you identify true anomalies or suspicious activity.
The bottom line is that your KYC programs will be more effective when they include insights from external data. To avoid being hoodwinked by bad actors and Dirty Johns, you need to snoop around and do the detective work yourself. You can’t rely on a narrow pool of data or over-generalized transaction behavior patterns.
That said, you need to keep these programs and procedures as streamlined as possible. You can’t afford delays or complications. You can’t waste time harmonizing data manually or setting up tricky connections between data sources, platforms, and models. It’s vital that your risk modeling platform is set up to automate data science tasks and connections, giving you access to the accurate, pre-vetted, compatible external datasets you need.