The Nilson Report

THE CURRENT ISSUE: Issue 1149 | Mar 2019

FEATURED COMPANIES

Companies featured in this issue include:

Top U.S. Merchant Acquirers

Mastercard Makes Two Acquisitions

Installment Loans for Walmart Shoppers

Pan-European P2P Service

Fingerprint Cards’ Biometric Software for Payments

ZestFinance Machine Learning Underwriting

Cardlytics Offers Targeted Rewards

Investments & Acquisitions—February 2019

Fleetcor Expands B2B Services

NFC Specification Alternative to QR Codes

Top Acquirers in the U.S.

Investments & Acquisitions—February 2019

U.S. Acquirer Totals—2018 Ranked by Purchase Volume and Purchase Transactions

Top Merchant Acquirers in the U.S.—2018 Ranked by Visa/Mastercard Volume

Card Not Present (CNP) Acquiring in the U.S.—2018

Merchant Acquirers in the U.S. 2018—Ranked by Visa/Mastercard Volume

Based on Visa and Mastercard credit, debit, and prepaid purchase volume processed in the United States, the 5 largest merchant acquirers are listed here.

1. Chase Merchant Services, Texas
$1,041.95 bil. V/MC volume, +15.0%
2. Bank of America, Georgia
$684.57 bil. V/MC volume, +6.2%
3. Worldpay, Ohio
$655.59 bil. V/MC volume, +5.3%
4. Wells Fargo, California
$421.19 bil. V/MC volume, +23.9%
5. Global Payments, Georgia
$333.84 bil. V/MC volume, 3.4%

Full access to the Merchant Acquirers in the U.S. 2018—Ranked by Visa/Mastercard Volume results is available when you subscribe to The Nilson Report.

POSTED MAR 18, 2019 | PRINT

ZESTFINANCE MACHINE LEARNING UNDERWRITING

Traditional credit scoring to support lending businesses depends on mathematics called logistic regression to analyze 20 to 40 variables to support a decision. That math is effective when examining the credit histories of consumers in the prime or super-prime categories. It is less effective when considering consumers with little or no credit histories—thin file, no file, or subprime prospects.

Complex machine learning mathematics combined with what is effectively unlimited computer power is now available to make predictions on the creditworthiness of thin file, no file, or subprime prospects by examining hundreds or thousands of variables. While machine learning can improve risk prediction for prime and super-prime prospects, it is particularly valuable for applicants with credit files that are messy, are missing data, or include incorrect data. 

However, one problem with machine learning for credit scoring is that the mathematical computations are so complicated that humans can’t explain why the model’s algorithm reached the decision it did. This “black box” problem has kept financial institutions from deploying machine learning because they could not definitively explain to regulators why their model did not present a challenge to the safety and soundness of the financial system. Nor could they guarantee consumers would not be harmed.

Yet, unregulated fintech lenders, including Affirm, Avant, and Upstart in the U.S. as well as Aire in the U.K., do use machine learning to create credit scores. This is the main reason they approve more credit applicants than general purpose and private label credit card issuers. Fintechs partner with banks to open loans and then almost immediately buy those assets, removing the originating banks from regulatory scrutiny. 

ZestFinance believes it has solved this issue of regulatory concern for banks. Its software-as-a-service offer lets credit card issuers and other lenders build and deploy models—and explain them. It provides software “robots” that monitor and document all decisions during the modeling process so bank examiners can see and understand what is in the black box. Credit grantors receive regulatory approval for their models before they go into production. 

Discover and Synchrony Financial are ZestFinance customers. Zest offers a premium product for lenders with portfolios greater than $1 billion and a more streamlined product for lenders with smaller portfolios. The company’s technology can be used to help issuers raise approvals without the risk of higher charge-offs or help them reduce losses below current levels. It can also be used to analyze data sources that traditional linear regression models can’t support, including payment card transactions and call center data. 

Kareem Saleh is Executive VP at ZestFinance in Los Angeles, California, kns@zestfinance.com, www.zestfinance.com.

© Copyright 2019 The Nilson Report