A Guide to Combating Financial Crimes With Graph Databases and Analytics
By Heather Adams, Managing Director, Accenture originally featured in our Spring 2021 Graph + AI Summit. Watch here.
Detecting financial crime is not just a legal or compliance exercise; it’s about playing a meaningful role in society.
Why should organizations take a more data and analytics-driven approach to combat financial crime?
Well, there are several reasons:
To improve compliance and risk management. Using more data and analytical approaches enables organizations to add robust, auditable controls, and to do so without having to take on additional headcount. Many are also looking to leverage new sources of data to do this, such as more adverse media information from news articles around the world, data available from registries or paid data sources.
To reduce manual effort. For those who carry a significant headcount, due to the nature of their existing fraud and financial crime controls, predominantly banks, taking a more data-driven and analytics-driven approach can also increase efficiency. Data-driven approaches typically reduce manual effort, and analytics reduces the volume of alerts requiring investigation.
To help organizations realize benefits from digital transformation. Recently, the pace of digital transformation has been growing faster than ever seen before, especially with the global pandemic. This shift to more digital from home and less in-person interactions is fueling a lot of organizations to really invest in digital change and transforming their organizations to have a more digital customer experience. This increasing digitalization of customer experiences is a good opportunity to take data captured electronically and to enrich your business case. By then using that data to allow for much more automated checks for signs of fraud, money laundering, terrorist financing, or sanctions breaches. It can also flow through into data stores that align customer behaviors with their transactions, and information that can be used to look for unusual or suspicious activity.
What are the key areas of opportunity if we can bring in more data and analytics in combating financial crime?
Better data sourcing. Acquiring a broader range of data can be achieved via more streamlined customer interactions. We can use this to drive more automated and robust decision-making and risk identification. That data can also be fed into tools like graph technology which connects data effectively. With internal and external data sources you are able to identify relationships between parties, this enables you to look at the risk associated across these hierarchies and relationships, rather than having the narrow view of one counterparty.
We can also start to think about how we can use unstructured data. Typically, it has been a concept that we can use data from the world’s media. However, to be able to pick out all that information from large chunks of unstructured data, we need to be able to use natural language processing (NLP). Machine learning needs to be developed to sort through all of the information that’s gathered to be able to produce meaningful results that are risk scored and worthy of manual investigation. As such, we need to have ways to not only pick up and machine read the data, but also ways to qualify and to score that data using the analytical capabilities. We can also use NLP to pick out data from documents provided by our customers and reduce a lot of the manual rekeying that has happened to date.
Better detection of possible fraudulent activity of money laundering, terrorist financing, or sanctions breaches. With analytics, we can have a clearer view of factors within the network of relationships and transactions between people and companies.
We can use graph technology to help us better understand connections within our data. This enables us to examine risk in a different way, which may reveal relationships we may not have been able to see before.
We can also explore patterns. For example, how does the behavior of a particular group of customers vary against their peers? We can start to compare behaviors over time with what was normal for that customer in the past versus what they’re doing now.
We can also use patterns to establish patterns of normal and be able to use that to remove noise from the system. Often in the past, organizations like banks, who’ve had to look into trying to flag suspicious activities have suffered from incredibly high false positives. And as a result it’s been like trying to spot a needle in the haystack. But now we can build up patterns of good customers and good behaviors. We can use these to drag out some of the false positives as well as better honing in on the things that are truly unusual or truly suspicious.
Where should organizations be focusing their attention?
Have a holistic view of what you’re seeking to achieve, and have some ideas of the kinds of tools and technologies that will help you get there. But it doesn’t mean that you need to do everything at once.
It’s really important to test and learn on your data, making incremental gains with each evolution of the analytical models and the data sets that you’re using and applying what you’ve learned to the next. Why? Well, a lot of clients that I talked to, you can initially feel very overwhelmed by the vast array of data that they already have. And one of the best things to do in the first instance is not to add more, but actually focusing on what is important – different hypotheses that you want to test, patterns of behavior that you want to look for, what typical fraud cases you’d like to search for you – and test and learn from the data. Once you have a working model that’s really honed in and refined down and contains the right data that you need, then that is a very powerful asset. This asset can then be deployed and rolled out in further aspects of your business or different geographies.
What else is it important to get right along the way?
To get those hypotheses right, you have to team up your content specialists with your data scientists and data engineers.
Data engineers will help you hone in on important attributes, they’ll help you build ways of leveraging your data. Data scientists can build models that can leverage all different aspects of advanced analytics and machine learning, but they are not usually people who are experts in your business and the behaviors of your customers.
So you need people who understand your industry, where fraud may take place, where people may be exploiting you for purposes of money laundering, or which aspects of your business have the potential to be breached.
Bringing these key risk scenarios into your thinking and using them to generate hypotheses, test these hypotheses, and learn from the results, is really important to be able to explain your logic to others.
Explaining your logic to others brings me to my second point: designing for explainability. It’s really fundamental that your models align with your ethical principles and your internal policies. What you are trying to identify needs to be things you would be comfortable explaining to your customers.
And then finally, do not go out alone, leverage the experience of others. This is a very fast-evolving area. And being able to work with people who know how to go about doing this can really help accelerate, and can really help you focus on the value that you get out of this exercise. It’ll stop you from making mistakes that others have made before. And it will make your journey go so much more smoothly in terms of being able to add incrementally the value that you get from trying to bring us data more effectively and to build out your advanced analytics capability to better combat financial crime.
In conclusion, any step forward in better identifying where fraudsters are taking advantage of customers, or where your organization is being abused by criminals, is a great step forward. Data and analytics can absolutely be part of that journey and can be really important tools in making that happen. They are, however, the kind of tools where you may not get everything right the first time: you need to build things up in a way that you learn as you go, and you keep evolving it. But in return, what you get is your team learns and grows, or you work with others who bring the relevant skills and experience and has a way of really boosting up your capability to better combat fraud and financial crime in a very repeatable, scalable, and dynamic way.
This won’t be a solution that in another few years, you have to throw away and replace with the next best thing. Taking these kinds of approaches allows you to continuously evolve and learn and make your solution better and better.
Watch Heater Adam’s Graph + AI Session here.