The early days were not easy for companies with new automated underwriting application programming interfaces (APIs) to sell. Attitudes surrounding the use of technology in what had traditionally been a pretty labour-intensive area (underwriting) were entrenched. Why would any lender want a ‘computer says no’ solution when they had a perfectly good team of underwriters in place?
Conversations with what became the early adopters of this transformational technology were always challenging. You would often find that while you might have persuaded senior management of the benefits, the deal would falter at the eleventh hour when they could not sell the technology internally.
There were too many defensive managers with groundless worries about their departments or jobs being made obsolete.
Part of the issue was the use of the word automated, which many misinterpreted as meaning the end of all manual input (and therefore redundancy for the underwriting team). A key element of the learning process involved telling lenders that the term 'automated underwriting' covered a very wide spectrum, and that their position on that spectrum was determined by the sort of lending they were involved in.
If most of their lending involved sums of under £200, then they might want to automate 90 per cent of their transactions (given the disproportionate cost of manually underwriting such cases). Alternatively, if they were a mortgage lender, then they may prefer to retain manual review and only automate the early stages of their screening process.
I would unpack the automated underwriting process and tell lenders to rather think of their back office as perhaps 20 tasks, of which they may well be able to automate the first 15. Also, if they did, they could then free their existing team to focus on areas that actually benefited from human input. No redundancies, just a refocusing of the team.
Eventually, the message did get through to lenders. Bit by bit.
They began to see that the technology could, in seconds, provide granular insight into prospective borrowers by interrogating line-by-line their finances; that scorecard changes could be made at the press of a button; that borrower screening did not always need costly credit bureau searches; that decisions could be 100 per cent in line with a lender’s credit policy; and that an electronic audit trail was there to protect them from mis-selling claims.
As a result of many of those conversations, we have now reached a stage where the technology is familiar to most in this industry. A sizeable number of lenders at all ends of the spectrum – from high-cost, short-term credit to mortgage – are utilising it. Using the Everett Rogers’ technology adoption lifecycle, we have moved on from ‘innovators’, through ‘early adopters’, and will shortly reach the stage of ‘early majority’.
Those that already have it have been benefiting from what is better termed assisted decisioning’s ability to deliver optimal lending outcomes (for both the borrower and the lender) and the ease with which it permits lending to be scaled according to appetite.
Such lenders are currently looking to take automated underwriting to the next stage: leverage the technology going forward so that it can deliver even greater value.
Two areas that are particularly exciting in this context concern machine learning (ML) and multi-bureau credit searches.
ML involves the use of algorithms that improve processes automatically through the mining of data. Essentially, it amounts to the formulation of best practice in an algorithm that, over time, gets better and better at doing its job.
Up to now the only way companies had access to this sort of sophistication was to pay a data analyst to go away with their raw data and, over a period of many weeks, work out what lessons could be learnt. The result was that not nearly enough data analysis was conducted by a typical lender, which is incredibly ironic given that lenders are ultimately data-driven enterprises.
Platforms such as Auto Decision Platform by LendingMetrics, however, can empower lenders to harness their data for immeasurably better-quality decision making. A lender can set a ML process in motion – swiftly deployed using ADP – that will tell them in minutes what, for example, are the best combination of predictive values on which to base a loan decision. And there is no costly data analyst to employ.
Lenders have always aspired to be able to use multi-bureau credit checks in their decisioning. Individual bureaus tend to have their strong and weak points, so a multi-bureau search makes a lot of sense. However, the majority of lenders are limited to a single search from one of the three usual providers. Given the extra cost, two credit searches are rare, three even rarer, and only if the loan is of a size that warrants it.
In this instance, platforms are, for the first time, making multi-bureau searches possible and affordable. At long last, on the basis of one contract, lenders can tap into two bureaux (in the case of LendingMetrics this is Equifax and Experian) to deliver a better-informed search. Combined with ADP, this power couple allows lenders to easily apply and adjust their multi-bureau strategies at will.
While these two facets of assisted decisioning are being enthusiastically embraced by the sector, there has been one other that has not had the speedy adoption that I first imagined. Open banking has not really gained the traction that was first originally thought, for a number of reasons.
While it does deliver real-time transaction access, unlike bureau searches, it does not show missed loan instalments or previous adverse history. It also introduces several extra pages of friction to consumers, which not all of whom will push through. Furthermore, unlike open banking, credit searches always deliver a 100 per cent response.
The only way I see this situation changing will be if the use of open banking data is compelled by regulators, as it is in countries such as Australia. And there is no indication of this being on the cards in the UK just yet.
Looking ahead to the next six or seven years of automated decisioning, I am certain that what we have seen so far is just a small sample of what is to come. The future for lending is going to be more and more about data and the technology that leverages it.
As I keep on saying to lenders, their unique selling proposition is not simply their products but also their data. Those that fully appreciate this and act accordingly are going to the ones that succeed.
Neil Williams is chief technology officer of LendingMetrics
