Having launched a company in the peer-to-peer lending space, a question I often get asked is whether our aim is to make credit decisions automatically without intervention of humans. We like to talk about our technology and the speed with which we can approve or reject an application, but at the same time a quick look at our career page shows that we are hiring for several roles in our risk team. Surely automation means we don’t need to hire many risk people? So, what does our risk team do then?
At MarketInvoice we’ve never aimed to completely remove humans from the credit decision process. Our philosophy is to leverage data and technology to help our risk team make better decisions. We want to move away from humans doing low-level process work to higher value tasks such as iterating on our credit models and sector policies, as well as working with our sales teams to improve what we can offer customers in terms of the amount of funding available, and at what price.
A typical borrower on MarketInvoice is a business that might have anywhere between £1m — £10m in yearly revenues. This might be a design business working with clients like Bacardi, a fashion wholesaler selling into Miss Selfridge and John Lewis, or an engineering company working within the oil and gas industry.
So, let me give some more concrete examples of the types of risk we’re looking at when we underwrite these funding requests:
Credit risk of the borrower: Our first key job is to understand the stability of this small business borrower to see whether they are financially stable enough to be able to deliver on the goods and services they provide their customers over the next 6 months?
To do this we need to think of ways to pull the right information about the business itself. Here we look at public filings on Companies House and credit bureaus, as well as financial information that the company itself uploads via our application form or via our technical integrations with accountancy platforms (such as Sage, Xero, and Quickbooks). This information can then be sorted, parsed and analysed to form a “financial stability score” that ranks the riskiness of the borrower against other similar businesses in our portfolio.
Fraud risk: This is a significant risk in all financial services. It’s important to remember that fraud is committed by individuals and not businesses, so we’ve developed a checklist to test whether the people applying on behalf of a company are real. We look at their track record, and their online footprint both inside and outside the company to determine the reliability of these individuals.
Customer risk: Having screened the business borrower, because our model is to advance money against future revenue being paid against invoices, licences or contracts, our Risk Team must look at the customer counterparts that the borrower trades with. Here we ask two key questions:
Can the customer pay? And will they pay?
We review public information and other sources (such as credit insurance or international credit reports) to assess the creditworthiness of the borrower’s customers and their ability to settle their invoices on time. Simultaneously, we look at the historical trading history and relationship that exists between the borrower and their customer, examining various data points such as timely delivery of customer orders, payment frequency, and any customer returns.
All these data points feed into our risk model, which categorises the borrowers funding request into a risk banding. This banding will determine an appropriate return to our investors relative to level of risk taken.
Now, while in the early days of building MarketInvoice, a lot of this information above was gathered manually. Our engineers are now busy building technology that automates much of the data aggregation and pulls everything into one place for our risk team to analyse. This allows risk team members to review an application and approve or reject it within an hour, compared to several weeks for the banks. We might be able to go even faster as banks are being nudged to make business bank account data available online (through the Payment Services Directive).
Given the duration of our lending is short (typically funds are 30–90 days outstanding), our risk team can continuously learn which loans have gone into default and how well our collections processes have performed. We can then use this to make changes to our risk models.
Our data science team uses machine learning and artificial intelligence methods to learn from historical applications, loan requests, and defaults, to identify the statistically significant determinants of default or fraud attempts by borrowers. This in turn impacts how we accept or reject future applications.
Macro factors also impact this calibration exercise: for example, if we think that Brexit will affect the large commercial property developers, we can look at our advance rate against construction invoices and lower them to reflect the greater risk of invoices not being paid in full, or projects being cancelled with knock-on effects down the supply chain. This allows our models to be dynamic and fluid depending on the learnings we uncover.
We’ve discovered that in lending, risk models equal the product you are offering to customers. Yes, business borrowers expect easy-to-complete online application forms, intuitive draw-down requests and reporting, but they are most interested in the size and growth of their credit limits and the cost of funds. And these two things come directly from our risk operating model. Our risk team is therefore at the heart of product development, and crucial in the decision to roll-out new products. In February of this year we launched our MarketInvoice Pro product for larger invoice finance facilities, and this required a new risk model with a greater level of interaction between our risk team and borrowers to ensure we understood their business and ambitions. And if we want to develop more products, we would need a new risk framework to make them work.
So, as you can see there is a huge amount of work to be done within risk at MarketInvoice and we’ve managed to get this far with a very small team. As the funding volumes and client numbers go up steadily we are looking to grow this team to help deliver a world class service to existing clients as well as work with engineering, data science, and product to develop and refine our risk models for the new finance products we want to launch.
It’s also hugely satisfying to see how our funding can have a significant impact on the success of the businesses using MarketInvoice. Many businesses apply to us when they only have a few customers and a less than a million of orders on their books, and several years later, they’re turning over multi-millions and have customers all over the world. This growth comes with risk and there will be times we can’t always get comfortable with what the customer needs. It is the job of our risk team to manage this carefully to meet our customers’ needs as well as ensuring a low loss rate and acceptable returns for investors.
So, if any of the above interests you then please get in touch as we’re looking to grow our risk team significantly over the next few months. We’ve had great success with applicants from different backgrounds ranging from economic consultancy to financial services, to operations. Check out our careers page for more information.
Co-Founder of MarketFinance