Financial Inclusion Does Not Happen By Accident
By Tushar Agarwal, Product Manager and Joshua Linn, VP of Machine Learning Product Management
Regulations tied to The Customer Identification Program (CIP) have been in force since 2003 with the intent that banks and other financial institutions form a reasonable belief that they know the true identity of their customers. The spirit of CIP is not in the aggregation of data, but in evaluation of a holistic identity that goes beyond just the information contained in an online application. A bank must obtain a residential or business street address for an individual customer because the Treasury and the Agencies have determined that law enforcement agencies should be able to contact an individual customer at a physical location, rather than solely through a mailing address. This makes address classification a crucial part of determining an individual’s holistic identity.
The US Census Bureau estimates that over 31 million people changed addresses within the United States in 2019 — over 13 million households [1]. COVID-19 accelerated this trend, with Pew Research Center estimating that one in five Americans relocated in 2020 [2]. This unprecedented mobility has stressed a financial system that has for decades relied on traditional and outdated methods of identity verification, which includes validating addresses. Simply put, banks are not prepared.
As business-critical as preventing fraudulent and incomplete data from clearing KYC (Know Your Customer) is to our customers, it is equally critical to ensure that no good consumer is disenfranchised of the country’s financial system due to local well-established whimsicalities and acronyms, minor errors or algorithmic inadequacies. To this end, our goal is to minimize friction for consumers while maintaining a high degree of accuracy.
But how reliable would we be if we rejected a consumer who mistyped ‘Main Street’ as ‘Msin Street’? Or if we couldn’t correct ‘100–52102nd Street’ to ‘100–52 102nd Street’? States often have different terminology for the same entity — US Highway, State Route, Interstate; and not recognizing such well-established entities means rejecting good customers and robbing them of their opportunity to be part of the American Dream. Even more so in minority US Territories like Puerto Rico where identical street names and address number ranges can be found within the same zip code. Here, the municipality urbanization is the only address element that correctly identifies the location of a particular address but oftentimes is also the very thing that causes address classification issues.
Financial Inclusion doesn’t just happen by accident. It takes all actors in the financial services industry to bring something to the table, incrementally working to harness the collective intellect and contributions of marginalized rural communities, Native Americans, and others — a helping hand that positively impacts the economy by ensuring that people have equitable and seamless access to the financial products they love most.
To this end, Socure commits significant resources to ensuring that underserved areas receive their fair share of consideration and representation within our algorithms and conversations, striving to let no bias creep into the models that source a wide-range of data sources to validate addresses. It is only an added bonus that by helping the communities most in need, we are also able to serve our customers as a whole. Less than 1% of the addresses Socure processes are not served by leading delivery services in the country and these addresses tend to disproportionately disadvantage ultra-rural communities in the hinterlands and Native American reservations. If our investments lead to more people being validated correctly and quickly every year than the last, Socure will have successfully played its part in the industry. We will have spurred other firms to follow our lead or risk being left by the wayside as we make headway toward our mission: identifying 100% of good identities in real time.
As part of building a holistic identity, our engines are trained to understand that simply resolving an address isn’t sufficient if our customers cannot deliver physical mail to that address. Auto-accepting customers only to have their mail returned is a colossal waste of resources, time and the trust our customers have placed in us. Socure’s pre-processing engines continue to be trained to recognize ever-increasing permutations in addresses — from handling New York City’s boroughs and neighborhoods referred to as cities themselves to knowing that in Texas, “Farm to Market’’ is a spoken term in addresses, recognized by leading last-mile delivery and address validation services as only ”FM.” These insights and the improvements that follow are only possible because Socure’s people and processes focus not just on eliminating fraud but doing it in a way that fosters frictionless consumer experiences.
A secondary practice that creates and exacerbates credibility issues throughout the industry are false positives, created by data aggregators when they resolve addresses that “look” correct. Very often, when data aggregators fail to resolve a certain address, they simply look for common mistakes in addresses and if no such mistakes exist, the address is classified to be “correct”. Such addresses may be undeliverable and can artificially prop up resolution numbers. Socure does not auto-classify an address as correct unless proven wrong. We are proud to play no part in plaguing the industry with false positives. Every address must be incorrect unless proven otherwise.
Over the last quarter, tremendous improvements have been made by Socure to serve users residing in the Virgin Islands, Puerto Rico, Guam, Hawaii and Alaska, where addresses continue to not be classified and resolved at the same rate as the rest of the mainland. Inability of last-mile delivery services to make successful portage here combined with a unique Urbanization system of municipalities, makes for a problem that plagues the entire industry.
Our commitment to these areas remains absolute. Socure’s current industry leading resolution rate for these addresses will only improve with smarter algorithms, improvements in data acquisition and expanded access into these territories year-over-year. The tremendous progress witnessed in recent years with the advent of a customer that demands no unnecessary friction and cherishes being part of an equitable financial ecosystem, heralds Socure’s greatest strength and opportunity.
Sources
- US Census Bureau: Calculating Migration Expectancy Using ACS Data
- Pew Research Center: 1/5th of US Adults moved due to COVID-19
