Capital is the lifeblood of a business. For companies that make physical things, such as fashion brands, accessing the right capital at the right time is crucial. Both direct to consumer and wholesale brands have to pay for products to be manufactured before the end seller— the customer or the retailer—pays for them. This capital hole, which is entirely man-made, requires many brands to look for outside capital to finance their operations.

Commercial credit, often known as factoring, allows many brands to exist that otherwise could not, given the capital constraints and delays inherent in running a physical goods business. Even so, factoring firms often charge high—and sometimes predatory—rates to brands, anywhere from below 5% to upwards of 15% interest rates on loans. The cost and complexity of securing this credit places an immense burden on emerging fashion brands, which often have few other options and bow to factoring firms as a result.

Emerging brands are often at the whim of these firms, which need to ensure that the brand is viable if it’s selling direct to customers, or that the retailers it’s selling to are credit-worthy. Although the analysis factoring firms and banks do can range from personal to highly financial, the barriers for emerging brands to both secure credit and sustain a business on it are very real.

This might be changing. A new crop of startups and tech companies are plowing ahead with algorithmic underwriting in place of human underwriting, moving towards a new type of financial landscape. Two startups in the space, Fellow, an API for working capital, and Affirm, which allows consumers to finance online purchases incredibly easily, are pioneering this new model, and Amazon has been working in the space since 2011.

Programatic lending

The bet of programatic lending is that algorithms can take over the heavy lifting that goes into underwriting working capital, while also providing more sound risk analysis than humans are able to. Fellow, which just launched out of startup incubator Y Combinator, allows companies to easily integrate it into their financing workflow to auto-underwrite and finance invoices. A company would use Fellow to programmatically secure loans every time a new invoice is generated and then deposit that money into the right account so the company can continue operating. This presumably removes the friction it takes to find banks and convince bankers that the business is credit-worthy. The startup is brand new, and it’s way too early to know if it works, but the idea is quite powerful.

Affirm, on the other hand, has been around for a while and has raised $420 million in venture capital. Founded by PayPal co-founder Max Levchin, the company allows consumers to buy big-ticket items such as electronics and furniture and spread out the payment over a range of months with relatively low interest rates. Affirm pays the merchant at the point of purchase and then the consumer pays back Affirm over a period of time.

Affirm’s risk analysis is entirely programmatic, and no longer relies on outdated metrics such as FICO scores or human intuition. Anyone who has tried to apply for a credit card with little credit to their name knows that building up credit is traditionally a massive chicken and egg problem. Affirm, on the other hand, uses algorithms to look at a variety of factors to evaluate the customer for a given transaction beyond a FICO score. By only asking for a customer’s name, email, mobile phone number, date of birth, and the last four digits of her social security number, Affirm can make a decision, all in a matter of seconds.

Amazon has also experimented with programatic factoring, under an initiative called Amazon Lending. The program extends loans to qualified Amazon sellers and uses their selling history on the platform to underwrite the loan. From a piece in the Wall Street Journal:

Merchants who spoke to The Wall Street Journal said they were offered loans ranging from $1,000 to $38,000 apiece, with interest rates from less than 1% (for one of them) to 13.9% (for most who were interviewed). Small-business credit-card interest rates typically range from 13% to 19%.

Banks and factoring firms charge high interest rates because the rates are a function of assumed risk. The riskier a loan is, the higher the interest rate will be to hedge for that risk. The promise of algorithms is that they will allow for much better predictive analysis, which gives a clearer picture of the loan’s true risk to price it accordingly. These rates would be much lower since the chances that the applicant would pay the loan back have increased. The complexity of the systems powering this micro-lending platform is fascinating in their own right, but the benefit to consumers—more credit for more people who traditionally would not have access to it—is also noble.

Algorithms and bias

But the move to algorithms comes with complications. The tech industry loves to advertise the transparency and objectivity of algorithms. The problem, however, is that humans write algorithms, and since humans are biased, so are algorithms and platforms. The two most recent and prominent examples are the people and algorithms behind Facebook’s news feed and the accusations that Airbnb’s platform is enabling racism when people of color want to rent homes. The supposed equality of these platforms and algorithms is anything but certain, and the lack of transparency as to what factors into these systems doesn’t help.

This is all to say that the move towards more algorithms and platforms managing the financial underpinnings of a business, and our financial system more broadly, is not inherently good or care-free. Yes, technologies like Fellow and platforms like Affirm are decreasing friction and widening the pool of people and businesses with access to capital. But the reasoning behind why some people get access and other’s don’t still remains blurry.

The tradeoff of a computer making a decision to accept or deny someone a line of credit in less than a second is that the rationale behind the decision is mostly opaque. The recourse for an applicant gets much more convoluted. The irony, of course, is that many of these decisions will end up back with a human to make the final call when the applicant challenges a computer-generated verdict.

I have no doubt that the move towards algorithms will have a net positive impact on companies that need working capital, including fashion brands. But to realize this net positive impact, companies will have to pressure tech platforms to be ultra transparent about what goes into an algorithm and the reasoning behind every decision. Failing to do this risks putting the final decision right back into the hands of big institutions and legacy firms that have no interest in small businesses. But getting this right means unlocking an entire new class of emerging businesses that could not have existed in the old financial landscape.