14 Nov Income and Employment Validation – Our Final Fraud Frontier
I wonder why I saved income and employment validation for the final article in my series? It could have to do with the war wounds associated with poor validation processes that I would prefer to not revisit. Or it could come from the frustration that income and employment misrepresentation (aka fraud) are the larg.est source of fraud we face in indirect auto. Let’s re-visit some statistics that I shared in my first article on the Taxonomy of Fraud in Indirect Auto:
- According to Decision Logic, at least 10 percent of all consumer loan appli.cations (not just auto, or non-prime) are fraud, and income inflation is also at a meaningful level (some observa.tions are in the 25 percent range)
- Point Predictive estimates fraud in auto to be a $6BN/year problem, and two-thirds of that ($4BN) number relates to misrepresented income and employment
Did the dinosaurs have a process?
Traditionally, loan stipulations (stips) are gathered indirectly via the dealer, and submitted to the lender along with the rest of the physical loan documents in a single mail package. Many lender sales represen.tatives still round up the loan documents and drop them off at the lender’s office. These packages can often be incomplete, or the stipulations provided may not be sufficient or legible enough to satisfy the lender’s requirements. Ultimately, the information provided must be validated, and not simply taken for face value.
So, lenders must take the next step of actually verifying the information pro.vided. I have observed overly manual pro.cesses where internal fraud analysts use magnifying glasses to look for miniscule text or line discrepancies on photocop.ies of photocopied documents only to declare with absolute certainty that some of these documents are clearly fraud while others pass. In this world of mysticism and insanity, my Magic 8 ball and Ouija board should qualify as “artificial intelligence”.
This process ultimately creates unneces.sary conflict between sales and operations, and ultimately proves nothing because there is no data involved.
When we engage the data available and try to impose it on the process, we quickly see that there are still many gaps and roadblocks. What if you have a thin file or zero score and the FICO is not par.ticularly insightful? What if the customer’s workplace won’t assist with a workp
A review of the data and solutions in the marketplace revealed to me that there are absolutely solutions available that lenders (and dealers) can deploy to help “stop the madness”. The good news is that there are many automated tools that can help add data to the verifications process that also increase speed and perhaps even book-to.look ratios for both lenders and dealers. Let’s take a look at what is available today and explore what the near-term future holds.
Credit bureaus and quantitative/data science organizations have built up income models that essentially validate income based on a variety of data sources. Much like a credit scoring model sorts applica.tions and rank-orders them, applications can be sorted in much the same fashion. For stated incomes and employment that makes sense for other factors such as utility and rent data and are within a comfortable level of tolerance (say +/- 15 percent), you can pursue one path of verification that requires less scrutiny. For incomes that are outside of the tolerance levels, you can require more verification steps.
Employment can be verified in an auto.mated fashion using a third-party service to absolutely validate the employment. According to this source, 85 percent of for.tune 500 companies participate, resulting in hit rates generally in the 19-36 percent range. The good news with a model like this is that you only pay for the positive hits, and once you get a hit, you have the employment verified.
For applicants that are not captured by this data provider, lenders are relegated to manually verifying employment (call, e-mail, fax). This process has issues since it is difficult to legitimize that you are really speaking to someone’s manager and that they are being truthful. The weight that I would place on this type of validations would be very small relative to the other options.
When the applicant has a bank account, and consents to allow direct access to view that information, lenders have an extremely powerful data source to use in verifying income. These solutions pro.vide instant bank verifications and return valuable information. For example, many providers of bank verifications show run.ning balance, summary analytics, and help you build an overall financial snapshot. If you are a data junkie, there are multiple variables available from the bank, as many as 30 categories that can be returned and evaluated. What is particularly exciting is that a number of these solutions are mobile enabled, which provides a strong initial handshake with your applicant that I believe can be the spark for a strength.ened connection with your borrower, and ultimately improved loan performance. The bank statement is very predictive of behavior, so much so that there are some that say that it could even supplant the need to even pull credit at all.
The downside with this solution is that not all non-prime borrowers are banked. So, for deals that make it this far and still can’t be verified, the lender would need to add some other verification or send the deal back to the dealer.
So, at this point we have a straw-man of a process that lenders can implement and verify most of their applicants (see Figure 1. Example Income/Employment Validation Process).
A better way to gather stipulations
A promising development of the recent few years has been mobile penetration that allows lenders to gather stips directly from the customer. When done right, this makes the dealer’s job more focused on getting more deals through versus admin.istratively shepherding deals through the lenders’ pipeline. At the same time, the lender has a direct connection to the bor.rower that streamlines the process and minimizes the opportunity for fraud from a third-party.
One solution that I reviewed provides the customer the ability to submit their own stip documents via their mobile device. The real benefit of this solution is that the lender creates a direct connec.tion with the applicant and removes any noise or potential fraud introduced by involving third parties. While the par.ticular solution that I reviewed doesn’t employ any natural language pro.cessing to analyze the information submitted, it does improve upon the process of manually gathering stips and submitting crumby look.ing scanned copies to the lender.
Another promising solution that I recently reviewed integrates identity verifications with bank verifications on a mobile plat.form that provides a view into cash flow and assets in a way that promises to solve for both income and employment. This solution has obvious relevance to lenders, but it is also being marketed to dealers to help them strengthen relationships with their lending partners. I really like the idea of getting the dealers to buy-in on the over.all process, because it builds a significant amount of trust.
A world with no stips?
While no solution exists today that eliminates the need for stips altogether, I believe that we are no more than three to five years away from just that. I truly believe that it comes down to two or three things: 1) mobile engagement coupled with 2) blockchain-enabled identity tokens, and 3) natural language processing (ability to take something like a PDF and make sense of the words on the page using AI). Taken in concert, these solutions not only solve the income and employment validation issue, but also verify identity. Until these solutions come online, we have a series of intermediate steps that we can deploy that will drive incremental improvements.
Joel Kennedy is chief operating officer at TruDecision Inc. and a NAF Association board member. He has a passion for grow.ing and improving auto finance ecosystem. He has over 23 years’ experience helping big banks down to start-up finance companies to build, grow, improve, and repeat. Kennedy can be reached at 240-308-2169 or firstname.lastname@example.org.