Every year, without fail, many banks and financial institutions embark on internal projects to improve their compliance operations. And every year, also without fail, they stumble and get stuck. What is it about customer screening that makes it so hard for firms to do it themselves?

On the surface, it ought to make sense for these organizations to set up their own compliance platforms. They have large groups of people with deep knowledge of KYC/AML compliance, customer onboarding and customer due diligence, plus IT teams experienced with banking systems. In theory, they should be able to build a customer screening platform quickly and inexpensively.

The Devil is in the Data

The problem is that once you dig into the data and technology needed to create a high-performance customer screening solution, a great deal of complexity rapidly appears.

Here, for example, are just some of the questions you need to answer about data:

• What do we collect? Just sanctions and OFAC lists? What about other watchlists? How do we collect this information? Who collects and keeps it current?
• What about content from other regions and countries? How do we get it? Do we have to localize everything?
• For our purposes, how do we define politically exposed persons (PEPs)? Are they all equal or do we need to develop and implement some type of risk-based rating approach for them? What about their relatives and associates? How do we define and track this?
• How do we incorporate adverse media/negative news into our screening model? How do we collect this data? How do we define what to collect? How is risk relevance defined? How do we avoid “fake news”?
• How much data do we need before we can start screening with some level of confidence?
• What can we do to avoid mountains of “false-positives”?
• How do we do all of this without investing in huge armies of content managers, editors, translators and writers?
• How do we keep all of this data current?
• What kinds of partnerships do we need in order to have a steady supply of reliable data? How do we establish all of those relationships worldwide?

It’s clear that accumulating the right amount, type and quality of data is a complex endeavor. It can be expensive in terms of headcount, workload and data licenses. Answering these questions requires skills, experience and insight that few banks and financial institutions possess.

Beyond the data challenge lies the second stumbling block: technology. Most high-performance customer screening programs are based on two core technologies: the Cloud, and name-matching. Cloud technology is still relatively new for many financial institutions, where traditional solutions have been developed on-premise and behind firewalls. Developing Cloud-based applications is beyond the scope of many technology teams. Yet the Software-as-a-Service (SaaS) model is ideally suited to compliance because of its ‘anywhere/anytime’ access and easy upgradeability.

 Name-Matching:  Tricky Stuff

Second, at its heart, customer screening relies on name-matching. The commonality of many names around the world (e.g. “Smith”, “Li”, “Kim”, “Mohammed”, “Johnson”) makes it very difficult to determine whether the person opening the account is a good actor or a potential financial criminal. Screening engines rely on advanced technologies including Natural Language Processing (NLP) and Machine Learning (ML). Natural language processing helps computers communicate with humans in their own language, and scales other language-related tasks.

For example, NLP makes it possible for computers to read text, hear speech, interpret it, measure sentiment and determine which parts are important. This is a critical part of customer screening. Interpreting different names, name order, spelling and punctuation is essential to two tasks: 1) the ability to efficiently ingest information into the database; and 2) efficiently matching names and minimizing false-positives.

Similarly, machine learning (ML) is now a critical component of a customer screening platform. Today’s algorithms can analyze more language-based data than humans, without fatigue and in a consistent, unbiased way. Considering the staggering amount of unstructured data that’s generated every day, from medical records to social media, automation will be critical to fully analyze customer data and records more efficiently. ML algorithms find natural patterns in data that generate insight and help you make better decisions and predictions. They are used every day to make critical decisions in medical diagnosis, stock trading, customer onboarding, and more. ML can take years of customer data and screening results and identify patterns that help to improve screening outcomes and efficiency. Few banks and financial institutions have the technical resources to develop world-class capabilities in this area.

Broader Insight is Better

The third and final stumbling block is around expertise and best practices. Even a KYC/AML compliance expert at Bank ABC still only knows about Bank ABC’s data, systems, processes and risk tolerances. Independent customer screening providers deliver screening for thousands of customers. This base of historical data provides a rich foundation for creating best practices and identifying trends and techniques most firms wouldn’t see.

In a highly competitive industry besieged by disruptive newcomers, DIY customer screening can be the project from hell. It involves a tremendous investment in data, technology and process optimization at precisely the time when in-house technology resources need to be focused on customer experience.