As the specialist lending market developed in the UK, a distinction was often drawn between the faceless, stringent decision-making machinery of high-street lenders and the approachable human beings making individual lending decisions in the niche and non-conforming sector of the market. The inevitable conclusion has been that manual underwriting is the best option for specialist lending.
Rapid developments in web–based IT over the last few years, fuelled by the need to produce FSA-compliant disclosure documentation at the point of sale, have enabled specialist lenders to speed up their decision-making processes to the point where decisions that normally took hours or even days are made in a matter of seconds. So, can we still afford to maintain the traditional idea that automated decision systems are unsuitable for the specialist lending market?
Decision automation
Lenders wanting to grow their business levels in a highly competitive market must adopt decision automation in some form, as it is the insatiable appetite for fast decisions and processing that forces the pace of change. Lenders that have traditionally pledged to underwrite every case individually now face the challenge using new technology to offer the benefits of both speed and transparency without being seen as selling out to an impersonal, machine-driven way of doing business.
As discussed in my previous article, credit scoring is a tried and tested method of setting the bar for an application to be accepted or declined, and it is highly effective in mainstream mortgage lending. Many lenders in the specialist mortgage sector now have sufficient data to construct effective application scorecards and some have systems to implement them but there is still the psychological barrier to overcome. Having spent so many years making a virtue out of the fact they don’t use credit scoring, niche and non-conforming lenders now have to win over popular support among mortgage intermediaries for a new breed of credit scorecards, turning possible hostility into approbation by demonstrating the various advantages.
Rule-based systems
One way of overcoming the dilemma – and the route we have adopted at SPML – is to approach decision automation by basing the system on the same rules used by underwriters (i.e. the lending criteria). Because decisions are based on rules – rather than on a score made up of various invisible elements – the reasons for the ‘yes’, ‘no’ and ‘maybe’ answers are much more transparent to the end user. The ability to bring elements of credit scoring into the cascade process at any given point is built in to the solution, allowing the full benefit of both methods to be employed simultaneously.
Rule-based decision systems are in everyday use and always have been. For example, when you phone a company and speak to the switchboard operator they ask you a few questions, go through a set of rules, then put you through to someone who can help you. Automated telephone answering aims to replicate the rules the human switchboard operator would follow. The fact that most of these systems go round in an infuriating loop, never coming up with the option that you need, is not the fault of automation per se. It is because the person who built the system failed to properly understand the rules that the human operator worked by and built a system that is not wholly fit for every permutation.
Similarly, when a human underwriter makes a lending decision, it is based on the rules of the lender. If these rules are then carefully built into an automated linear process, the case goes through exactly the same process that the underwriter follows. Here, we have a natural advantage over automated telephone answering in that our rules are very clearly set out and naturally lend themselves to an automated process in the form of a decision tree. As the case progresses along the tree, the clear ‘yes’ and ‘no’ decisions are automatically made, with the ‘maybe’ cases then reverting to an underwriter.
This sort of automation process has, we believe, distinct advantages for all parties to the mortgage loan. The broker and borrower benefit from speed and accuracy with all products available in a single assessment and gain clarity about the reasons for a decline – if that is the outcome. The lender benefits by being able to process the bulk of applications automatically, leaving skilled underwriters to process the more complicated and borderline cases. In addition, because the automated decision process is not subject to human error, it produces more accurate data for use in the lender’s own policy making and new product development – all of which feeds through to the benefit of borrowers and brokers.
Stuart Aitken is director of credit at SPML
