AltFi.com uses cookies on this website. They help us to know a little bit about how you use our website, which improves the browsing experience and marketing - both for you and for others. They are stored locally on your device. By continuing to use this site you accept this use of cookies. Go to the Privacy and Cookies page for more information. You'll see this message only once.
Not signed in. Log in here.

Your daily download of all things alternative finance and fintech, from us at AltFi


 

A tale of two credit models




By Ryan Weeks on 17th March 2016

Ernesto Andrade, https://goo.gl/Lqy7Ym

Is there anything truly innovative about the way that alternative lenders assess credit?

 

It’s a question that seems to have been cropping up quite a bit in industry discourse of late. Peer-to-peer lending as a concept is unquestionably innovative. When Zopa pioneered the idea in 2005, retail investors had never before been able to gain direct access to consumer loans. Funding Circle later applied the model to SME debt, and voila – the spread of peer-to-peer lending begins. But many will tell you that credit assessment processes in peer-to-peer – and indeed in other forms of alternative finance – are not especially innovative. The suggestion, put briefly, is that alternative lenders in fact rely upon the very same collection of data points and data crunching techniques that the incumbent lenders have leant upon for decades. Can this be true?

 

To find out, I caught up with Greg Mrkusic of Liberis and Paul Crayston at MarketInvoice, to find out if either of them can point to any practical differences between their credit models and the tried-and-tested methods of the banks.

 

Let’s start with Liberis. Liberis is a short term, merchant cash advance provider that writes loans to small businesses off its own balance sheet. Greg Mrkusic – Marketing and Risk Director at Liberis – rather appropriately set the scene for the discussion by stating that alternative lenders certainly aspire to be different and to be innovative. But the sense is that the intention in many cases falls short of the reality. The most important factor in Liberis’ credit decisioning process is the story behind an application, and the most important source of information is credit bureau data. The platform doesn’t necessarily factor in more data than a bank does when making credit decisions – as the intention is to incorporate only the most impactful of data points into the process. Perhaps most surprising is that the platform has to date been almost entirely people driven. At the outset of 2015, 100% of Liberis’ credit decisions were executed by humans. Nothing about this model is “wrong”, per se – it simply isn’t groundbreaking. It’s battle-tested, it’s effective and it’s familiar.

 

In 12 months’ time, however, Greg believes that the process will be between 70 and 80% automated – a sharp transition indeed. The future of credit at Liberis stands to look considerably more creative. Especially when one considers that the company recently teamed up with payments giant Worldpay. Worldpay processes approximately 31 million mobile, online and in-store transactions on an average day, supporting around 400,000 merchants globally. Its customer base now enjoys seamless access to the Liberis product – under the name “Worldpay Business Finance”.

 

The way that Liberis assesses Worldpay customers is considerably more innovative than the methods it applies to other customers. Cash flow data is gold dust in the merchant cash advance space – where repayments are deducted as a pre-determined percentage of credit and debit sales. Under the Worldpay partnership, Liberis is able to suck historical payment data directly out of the network – a powerful indicator of a loan applicant’s suitability. Greg tells me that the level of access provided by the Worldpay partnership allows Liberis to more accurately target marketing activities within the network. The partnership has also opened up the potential for pre-decisioning, where borrowers are pre-approved for borrowing up to certain amounts of money. Greg also informs me that Worldpay data can help the platform to price risk more effectively.

 

Alternative finance providers tend to hold speed of decision as a key advantage over incumbent lenders. For Liberis, that advantage is accentuated by the Worldpay partnership. The platform can afford to place a greater weight of emphasis on technology for Worldpay applicants, and the range of information considered is enhanced. In summary, the process that applies to Worldpay SMEs looks a great deal more like the cutting-edge credit processes that we hear so much about when discussing alternative finance.

 

Then we come onto MarketInvoice. Paul Crayston, Head of Communications, tells me that the platform considers a broader range of data points than the banks do when making a credit decision. However, he issued the caveat that what MarketInvoice is assessing is very different to what the banks assess. The essential consideration for MarketInvoice is whether or not a specific invoice will be paid. Banks, which do not tend to offer selective invoice financing services, must instead consider the broader health of a business – bog-standard credit reference agency data will do. Not so for MarketInvoice.

 

It comes as no surprise then that MarketInvoice has integrated with a number of transaction-based networks – in much the same way that Liberis has with Worldpay. MarketInvoice counts Sage, Xero and KashFlow amongst its allies. By tapping into cloud accountancy software, the platform is able to access real-time data on the financial health of a prospective customer. Bear in mind that customers are not required to use accountancy software in order to gain access to MarketInvoice’s services, but such technology is becoming an increasingly important part of the platform’s infrastructure.  

 

We often hear that MarketInvoice uses technology to a greater extent that the majority of its peer-to-peer contemporaries. Just read Credit Analytics Lead Hendrik Brackmann’s recent AltFi News article, entitled: Why Adair Turner is Wrong About the Need to ‘Kick the Tyres’. But what does this elevated appetite for automation translate to in practical terms? Paul tells me that machine learning is an important cog in the platform’s credit machine. For MarketInvoice, machine learning is the answer to the question: “how do we best take advantage of the short-term nature of our product?” Machine learning only works if there are deep reserves of historical data available to feed the machine. More than 10,000 individual transactions have completed (i.e. been fully repaid) through MarketInvoice – perhaps the largest number in the UK’s alternative finance sector. With a couple of hundred data points per transaction, effective analysis requires considerable processing power. Paul would argue that MarketInvoice has that kind of power, and that such technology is acting to continually sharpen the platform’s credit models.

 

That’s not to say that MarketInvoice doesn’t see a role for human intervention, but the platform differs from others in that it believes data should be the primary factor in driving credit decisions.

 

Ultimately, it all boils down to product. MarketInvoice can afford to take a more innovative approach to credit because of the nature of its very narrowly defined financing niche. Similarly, merchant cash advance provider Liberis can afford to adopt a more technology-led approach to lending than a secured business lender might. That secured business lender could probably stand to be a little more colourful in its credit assessment processes than a property lending platform could be. Selective invoice finance is an unusual product. It fits with an unusual approach to credit. The secured business loan is an age-old product that requires a tried-and-tested formula for success. That formula may be improved upon by technology over time, by why ignore indicative historical data if it’s readily available?

 

In summary, the passing of time fuels innovation – and time will tell whether MarketInvoice-like innovation will take hold across the alternative finance spectrum. 

Comments

Steven Renwick

17 Mar 2016 07:48pm

Building a good model with the capability for machine learning certainly does require large amounts of data. At Satago we started by building a fully-fledged credit control platform before building the invoice finance product which enabled us to train our risk engine on over 2 million invoices representing nearly £1 Billion of paid invoices. Because of that we were able to start with our risk model already nearly completely automated - which is important for us as we can finance invoices as small as £500 - which would not be possible at scale if we were having to manually check every invoice.


Enter your name:

Enter a comment in the box below: