Categorisation as a Service ® and what it can do for business lending
The core principles of risk assessment have remained unchanged through the centuries — ‘Stability, Ability and Willingness to Pay’. However, what is surprising is that despite the explosion in lending over the last 25 years and the development of new statistical techniques, greater availability of data and improved processing capabilities, lending techniques remain the same today.
The way it is, the way it’s always been
Risk departments rely heavily on credit reference agency data to demonstrate that a borrower previously had these “stability, ability and willingness” characteristics. If they paid back a loan before, they should pay one back in the future. It has always been backward looking — based on the assumption that the future is a close reflection of the past — but fails to assess a customer’s financial position at this moment and how it is trending. The industry has matured to improve pricing models and access powerful analytical tools, but they are essentially using the same techniques and core data as they did when Bill Fair and Earl Isaacs first developed statistical regression techniques back in 1953.
In truth, the availability of new data sources has not dramatically increased in the last decade. The introduction of the CallCredit credit reference agency has certainly challenged the two large incumbents Equifax and Experian. Some additional data sources such as Telecoms and sub-prime loans have been added to the mix. The quality of data provision within APIs improved alongside some new products such as Income Verification and CATO. But aside from a few warning shots from fintech startups, such as new social media risk assessors and “big data” analysts, not much has changed in the world of risk.
There has not been a significant new catalyst in the process of risk assessment, which has always simply been that if someone paid back a loan before, they’ll probably pay back again. Financial outliers such as the young, immigrants, people who have limited credit use, and those with historic impaired credit have always found the attainment of credit difficult or expensive. And the problem for business lending to small and medium sized enterprises (SMEs) is worse due to the lack of credit reference agency data. Additionally, with banks having cut back on business lending since the credit crisis due to capital, regulatory and risk appetite changes, only the biggest firms with long track records of profitability have found it easy to borrow quickly and cheaply. The historic risk assessment techniques have failed the UK’s SMEs, who can’t provide their data in an easily readable format to lenders. The unregulated nature of SME lending has helped hide what is comparatively expensive borrowing even when it can be obtained, often because lenders don’t have enough data to make quick, controlled and accurate decisions. But the competitive and regulatory landscape of business in the UK has changed, and the risk industry is desperate to change to meet this challenge.
The introduction of PSD2, and, closer to home, open banking, has the potential to change everything and put customers in control of their data as well as opening the market to new entrants. This will allow small businesses and customers to take ownership of their data; banks will become rightly just the custodians of customer data and the argument over who “owns” the customer consigned to the dustbin. Customers will be able to share their financial transaction history with third party finance providers, who will be on an equal data footing to the banks that currently have sole access to the primary transactional relationship — the current account. This will provide SMEs with a much wider pool of potential financial product providers whose credit policy and terms match their needs. The process promised to be electronic, automated, controlled and monitored. Regulatory requirements to perform enhanced credit assessment will not necessarily mean more rejections or slower application processing, but quicken innovation and choice as well as potentially improving the quality of credit decision with this additional data being used alongside traditional data.
Although PSD2 has firmly put open banking on the financial services map, access to customer’s data has always been possibly using a technique known as ‘screen scraping’. PSD2 and open banking will replace scraping, but it doesn’t solve the problem of what to do with the data once obtained, and how to understand and interpret it. Banks themselves have struggled to understand their own transactional data. A limited knowledge of whether a customer’s expenditure was cash vs. retail vs cross-border, perhaps with an industry code for the merchant, has been useful for fraud, but it has not allowed a bank to solve the question of affordability. Is a customer’s spend discretionary, or is it a monthly commitment that must be paid such as a council tax bill? Office of National Statistics estimates have tried to fill this gap, but estimates are averages and are not a true reflection of many individuals’ lives.
Castlight and Categorisation as a Service ®
Castlight Financial was born out of the Individual Voluntary Arrangement (IVA) market and a company called CreditFix. When a customer enters a voluntary arrangement a detailed assessment of their spending habits must be made to understand their true affordability, to make a fair and proportionate offer to creditors. This process historically was manual and could be challenged by the creditors. CreditFix started to keep a record of bank transaction data. If you look at your own bank statement, transaction text can be difficult to interpret. Yes, we can all remember that Starbucks is probably coffee, but what about the multitude of other transaction descriptions? CreditFix and now Castlight has a database of around two million items of data that has been classified into 25 types of credit and 130 expenditure categories. Castlight calls this CaaS — ‘Categorisation as a Service®’. Using neural network and proprietary techniques, Castlight’s CaaS can instantly categorise your spending habits — how much do you spend on your electricity bills, do you have a pension, how much do you give to charity every month, and so. This data can be trended and characteristics formed such as the size of a potential borrower’s discretionary spend in comparison to a potential loan. This depth of data aims to categorise 100% of transactions, and will continue to learn new records using the neural network combined with Castlight’s Affordability Passport TM that incorporates a customer feedback loop to further check and learn. Categorisation of less than 100% is not an option, and by simply looking for primary keywords will provide only half the answer and take us back to the inaccuracy of using estimates of a customer’s spending.
This data has helped CreditFix grow into the largest arranger of IVAs in the UK, reducing the information gathering process from weeks to just 64 minutes. It is now revolutionising mortgage broker compliance to Mortgage Conduct of Business rules from a lengthy manual process into an instant automated calculation. Loan providers are using or testing CaaS to improve lending decisions and reduce manual referrals.
Finance providers to SME businesses can use CaaS to evaluate quickly and easily a company’s banking behaviour. This includes turnover trends, regular income and high or low risk flags such as returned payment fees, over-limit charges or legal fees. There will be no onerous additional steps for an SME, just better more focussed decisions by lenders. CaaS can identify how frequently a business is paying salaries or monies owed, and whether this increasing or decreasing. Additionally, lenders can assess the financial behaviour of Directors — for example, do they run their own financial affairs in a stable manner, are they financially stretched, do they have an income stream outside the business? Castlight with CaaS will give good growing companies improved access to finance more quickly and on the terms that they deserve. Historically some SMEs have found it hard to access cheap finance due to their reduced credit footprint or trading history. This problem has infamously got worse in the last decade, only partially mitigated by the introduction of challenger banks and new fintech entrants. CaaS and access to bank transactional data will allow lenders to make a decision on these SMEs, and provide reasonably priced credit that otherwise would not be available. This will be hugely beneficial to small businesses that previously would have been denied access or paid too much — a segment estimated at 5.5 million private sector businesses at the start of 2016, according to the Federation of Small Businesses.
Castlight and CaaS, with the impetus provided by PSD2 and open banking, has given the risk assessment industry the new data needed to dramatically improve the credit assessment process. In the consumer market, comparison sites dominate lending for cards, loans and insurance, helping borrowers to access the most appropriate, quickest and cheapest credit. Open banking and Castlight’s CaaS will help these and other comparators develop SME versions, where there is a lack of credit reference agency data on business loan performance to support such decisioning.
At last, credit assessment is shifting onwards from the traditional methodology taught to graduates for the past three decades. Bill Fair and Earl Isaacs would approve.
Kevin Allen is Credit & Risk Product Manager at Castlight Financial.