Francisco Mainez: How HSBC is Using Data to Detect Money Laundering and Fraud

HSBC is modernizing its fraud and money laundering detection
capabilities by rolling out algorithms designed to identify ‘bad
apples’ in the most efficient way possible Digital transformation
is not unique to businesses. Since the onset of the COVID-19
pandemic, criminals have also moved their activities online. Now,
HSBC is using data to fight back. In this week’s episode of the
Business of Data Podcast, HSBC’s Head of Data and Analytics and
Business Financial Crime Risk Francisco Mainez explains the
motivations behind this global initiative. “We need to find
something that will divert our attention to the customers that we
really wanted to analyze. The ones that have the potential [to be]
the bad apples in your basket,” says Mainez. “And also, that’s
going to have a knock-on effect on cost efficiency.” Using Data to
Identify High-Risk Customers Historically, customer-focused
assessments looking for fraud and money laundering might involve
the manual review of thousands, if not tens of thousands of people
and accounts. This process was costly, inefficient and
time-consuming. By employing an algorithm to give individual
customers a personal score based on predetermined risk factors,
HSBC can quickly identify high-risk accounts. “What the algorithm
does is embed different key risk indicators,” says Mainez. “Are you
moving countries? Are you transacting with virtual currencies? Are
you over a certain age?” He continues: “In the old world, we will
be looking for things that we know for a fact from previous
experience that could be suspicious. With this [algorithm], you’re
scoring customers because you’re actually measuring customer
behavior.” Improving the Efficiency of Fraud Investigations By
using data to identify the high-risk accounts, HSBC is making sure
that their investigative resources are being used as efficiently as
possible. “You don’t want to spend 80% of your time, energy, budget
and resources, especially on the operation segment, checking a
false positive. Then 20% of the time rushing to find out if those
customers are the ones that you’re looking for,” Mainez says. “We
wanted to reverse that.” He continues: We’re going to spend minimal
time because the machine is going to help us make a decision on
which customers we need to review. And then we’re going to spend
the rest of the time properly analyzing the customers.” Mainez
points out that this initiative is designed to assist human
decision-making, not replace it. The human element of fraud
detection is still essential, especially when it comes to adapting
a global initiative to local realities. Taking Stock of Cultural
Factors Big financial institutions typically work in a very
decentralized way. To make this global initiative successful,
Mainez knew that the algorithm would have to take account of local
and cultural factors. “You also need to take into consideration
cultural factors, he says. “Every country is going to have to worry
about their own typologies, not the ones from [any other] country
because that’s going to produce false positives that they’ll have
to review”. By asking individual regions to specify the cultural or
regional typographies that best indicate risk, Mainez can tailor
the algorithm to that region. “You’re going to tell me which are
the typologies that are keeping you awake at night,” he says. We
want to help you by configuring the system in a way that can detect
those types of behaviors.” The new initiative has already been
rolled out in several regions, but the future has plenty in store
for Mainez and his team. “Over the next few months, we’ll be
deploying in more markets and continually tweaking those
typologies,” Mainez says. “Because of all possible times, we
started to roll this out in the middle of COVID.” He concludes:
“[Criminals] are adapting to a more digital and remote environment.
That’s reflected in the data, and we need to be able to figure out
how those typologies, and how they are evolving.”