[Editor’s note: This is a guest post from Eric VonDohlen, Chief Analytics Officer at Elevate. Elevate is a Bronze Sponsor at LendIt USA 2016, which will take place on April 11-12, 2016, in San Francisco. At LendIt, CEO Ken Rees will be speaking on the panel, How Do you Expand Credit to the Thin File?]
Consider two scenarios:
You are a prime credit customer looking to secure a car loan. The bank looks at some information about you including identity ensuring you are who you say you are, and your FICO score – which, if you’re a prime credit consumer, is above 700. Pretty simple.
You are a non-prime customer unable to secure credit for an unexpected car repair because of your lack of credit history, or some defaulted loans in the past few years. You rely on a payday loan shop to secure immediate cash in exchange for a high interest rate. Your credit history likely isn’t even considered.
There’s a clear gap between these two scenarios, both in what’s available to prime and non-prime consumers in the credit and lending marketplace, and how their information is considered when borrowing money.
While loan decisioning for prime borrowers is relatively straight-forward with a robust amount of background and financial information readily available, the same is not the case for the non-prime consumer. Despite its massive size and clear demand for access to credit, the non-prime market is generally not well understood by the majority of the financial services industry.
At Elevate, we look at the non-prime consumers as those who have a FICO score lower than 700, and those who lack sufficient credit history. This could be the single parent who has derogatory credit information due to some delinquent payments, or the recent grad who has no credit profile except for a student loan on record. These customers, no matter what put them in the non-prime space, present a uniquely difficult situation for credit decisioning and underwriting. Our staff of analysts make it look easier than it really is given the speed (within seconds) in which we deliver a loan or line of credit.
Imagine a funnel. At the top, we have massive amounts of data, and at the bottom, we have the results of whether we find the applicant to be a good credit risk. With a mix of our machine learning techniques, and data analytics staff, we aggregate all of this information from a variety of different sources including bank accounts, non-prime bureaus, social media accounts, geolocation, how the applicant navigates our website, employment history, and more – much more than what’s typically gathered from the prime applicant. It’s critical to go deeper for information when considering a loan or line of credit for the non-prime set, and leveraging multiple sources that can verify the same data point, improves the veracity of the information.
Given our amount of experience and time spent serving the non-prime customer, we have the ability to learn from that data and know the best way to curate that data through predictive models – and ‘predictive’ is the key.
In traditional statistical modeling, people are teaching the machines to use the data, whereas with machine learning, machines apply techniques and actually teach us about the data. This creates a tremendous advantage for us to scale and see data in a really systematic way. Our technology also allows for small adjustments to be made every day based on rapid success and failure. For example, if we learn a bad fact about an account we underwrote two months ago, we have the bandwidth and capabilities to see if it was an outlier, or we see risk and the subsequent opportunity to mitigate as necessary. We are able to re-implement learnings on a nearly continuous basis – the platform is truly a living, breathing mechanism.
While all this may sound a bit too much like the movie I, Robot where AI-enabled robots go rogue, fear not. Just like another other tool in a toolbox, we recognize it for just that – a tool. The staff here – 25 with advanced degrees and ten with PhDs – have created the technology and have their fingers on the pulse of market fluctuations. We still have the ability to dissect the information as humans presented by the machines. The data is the data, if you will, and what’s important is how we arrange and consume it in a way that derives meaningful and actionable information.
We see great opportunity for even more growth and advancement in providing non-prime customers access to credit. Through continuous efforts to perfect our technologies, and ongoing investments in talented analysts, Elevate will be even better positioned to serve more people from this severely underserved group of consumers.
The host of this blog, LendIt, is the largest conference series dedicated to connecting the global online lending community. Our conferences bring together the leading lending platforms, investors, and service providers in our industry for unparalleled educational, networking, and business development opportunities. LendIt hosts three conferences annually: our flagship conference LendIt USA as well as LendIt Europe in London and LendIt China in Shanghai. Visit our home page to register for the next event and to subscribe to our newsletter.