Square: Using machine learning to de-risk small business lending

Square is using transaction data to de-risk small business loans.

Square first came to market in 2009 with the Square Reader, a small device that enabled anyone to use a smartphone to accept credit card payments. Square has since developed a catalog of point-of-sale devices and services that help small businesses accept credit cards, manage inventory, and process employee payroll. As Square has expanded its services to cover more and more small business functions, the data it collects about these businesses paints a clear picture of their cash flow. Through it’s new group, Square Capital, Square has recently started using machine learning to analyze transaction data to extend lines of credit to small businesses – and the revenue from these loans has outpaced the overall growth of the company [2].

Square earns the majority of its revenue through a share of the credit card transactions that it processes. This revenue model aligns Square’s success with the success of its small business customers. Square, which has grown more than 500% since its 2015 IPO [3], has found success by leaning into this alignment and developing services that continue to help small businesses grow their business.

Small businesses have historically had trouble accessing the capital they need to purchase inventory, buy equipment, or cover short term expenses [4]. According to the 2017 Small Business Credit Survey, small businesses report financial challenges at a higher rate than larger firms [5]. Applicants tend to choose a lender based on their perceived chance of being funded (rather than on product cost), and 77% of applicants received less than the amount sought or did not obtain financing at all [5]. Square has recognized that access to capital to help cover short-term expenses and expansion plans is a key challenge for many small businesses.

So how does Square Capital work? As small businesses use Square to collect payments, Square monitors this transaction history to analyze the credit-worthiness of the business. Square’s credit algorithm appears to weigh the business’s transaction volume, most recent transaction, mix of new and returning customers, and revenue growth (among other factors) to determine whether the business is eligible for a loan [6]. Square proactively approves businesses for loans and since their bank account is already connected to Square, supplementary capital is available at the click of a button.

To date, Square Capital has offered over $1B in loans to more than 100,000 businesses and the average loan size was just $6,000 [7]. Square’s data-driven approach and small loan size is creating a new market that replaces the traditional path of asking for small loans from friends and family [8]. Not only has Square expanded the small business loan market, they have done so while achieving just a 4% loan default rate [9]. In comparison, the aggregate small business default rate through traditional lenders has hovered around 8% since 2012 [10]. This low default rate can be partially attributed to the data-driven method of determining eligibility, but Square has also made it easier for businesses to repay their loans. Repayments are automatically deducted from daily credit card sales at a rate of 9-13%, so the business pays more when sales are strong [11].

In 2016, Square partnered with Upserve, another credit card processing startup, to expand Square Capital to service small businesses that do not currently use Square [9]. While this partnership helps Square access more businesses and expand its credit service and core business, it does expose Square to additional risk. The data that Square has about these businesses may not be as comprehensive as the data it has collected through its own payment processing services. Lending to businesses with an incomplete picture of their transaction history may lead to an increase in default rate.

Square faces an exciting future as it continues to leverage customer data to empower small businesses to grow. Square’s use of machine learning may be able to help small businesses better manage their cash flow by predicting trends and cycles in revenue, identifying patterns in customer behavior to provide marketing insights, and identify opportunities to reduce expenses. As Square’s customer base continues to expand, Square may also be able to identify trends in local economic activity. How might this data be valuable to existing businesses, aspiring entrepreneurs, and local governments. Are there other opportunities to apply daily monitoring of cash flow to de-risk lending? How much of a threat does Square pose to traditional small business lenders and how should these banks respond?

(word count: 742)

[1] Image accessed from the Square Website, https://squareup.com/us/en/software/point-of-sale, accessed November 2018.
[2] Nathanial Popper, “Square, the Twitter Boss’s Other Company, Could Pass It in Value,” The New York Times, October 25, 2017, https://www.nytimes.com/2017/10/25/technology/jack-dorsey-twitter-square.html, accessed November 2018.
[3] Bloomberg, “SQ: US, Square, Inc.” https://www.bloomberg.com/quote/SQ:US, accessed November 2018.
[4] Jaime Toplin, “Why Square Capital could be a reliable source of revenue for Square”, Business Insider Intelligence, March 10, 2016, https://intelligence.businessinsider.com/post/why-square-capital-could-be-a-reliable-source-of-revenue-for-square-2016-3, accessed November 2018.
[5] Fed Small Business, “2017 Small Business Credit Survey” (PDF File), downloaded from Fed Small Business website, https://www.fedsmallbusiness.org/medialibrary/fedsmallbusiness/files/2018/sbcs-employer-firms-report.pdf, accessed November 2018.
[6] Square Website, “Tips for Small Business Financing: How to Get Noticed by Square Capital”, https://squareup.com/townsquare/tips-for-small-business-financing-how-to-get-noticed-by-square-capital, accessed November 2018.
[7] Square Website, “Square Capital: $1 Billion in Funding to Over 100,000 Sellers”, https://squareup.com/townsquare/square-capital-1-billion-in-funding-to-over-100k-sellers, accessed November 2018.
[8] Sara Vera, “A Peek into Machine Learning at Square”, Square Corner Blog, August 4, 2017, https://medium.com/square-corner-blog/a-peek-into-machine-learning-at-square-bdd75b272e23, accessed November 2018.
[9] Jaime Toplin, “Apple pushes back against big Australian banks —Square Capital begins lending to non-Square merchants — Samsung Pay’s potential security flaw”, Business Insider Intelligence, August 11, 2016, https://intelligence.businessinsider.com/post/apple-pushes-back-against-big-australian-banks-square-capital-begins-lending-to-non-square-merchants–samsung-pays-potential-security-flaw-2016-8, accessed November 2018.
[10] Wain Street Website, “Business Default Index”, http://wainstreet.com/platform/business-default-index/, accessed November 2018.
[11] Jackie Zimmerman, “Square Capital: Quick Business Loans for Square Merchants”, Nerdwallet, November 8, 2017, https://www.nerdwallet.com/blog/small-business/square-capital/#works, accessed November 2018.


Duolingo: Machine Learning Our Forgetfulness


Open Innovation at Nestle – Establishing an extended innovation ecosystem

2 thoughts on “Square: Using machine learning to de-risk small business lending

  1. Thank you for this interesting submission on Square Capital. While I think Square could pose a threat to traditional banks that dominate lending in the Small Business segment, it seems that online lenders share of the total lending market has stagnated in recent years at around 11%. One question is how will Square market this product to drive further industry penetration? https://www.fundera.com/blog/state-of-business-lending-q3-2017

    I think the applications of machine learning for square is going to be most applicable in taking share from businesses that already have a large presence in online lending. An example of a product I could see Square taking a significant share from is a Merchant Cash advance, such as those offered by non bank lenders such as Fora Financial. This product offers customers a large cash payment upfront, in exchange for a portion of their daily revenue.

  2. This is awesome! I definitely see how access to daily transaction data could give Square an edge over other lenders in assessing credit-worthiness. I wonder if Square also monitors the companies transactions in real time as a detection mechanism to flag potential defaults much earlier than other lenders can. The loan repayment structure also strikes me as a great way to differentiate from competitors by making it easier for the businesses to pay when they have the cash. With all of the transaction data Square gathers, I wonder if they could offer services where they aggregate and display patterns in the data to the small businesses to aid in decision making. If they were able to get buy-in from customers, they could even show businesses how they are performing against similar businesses to help managers understand the strengths and weaknesses of their business. (For example, your lunch transaction volume is 20% industry average.)

Leave a comment