Often these two processes are managed separately. In fact, even the way banks model these risks for capital and provisioning purposes – the cash they set aside to cover potential bad debts – can be isolated.
More can be done to bring these two elements together.
For example, some lenders are currently only prepared to use automated valuation models (AVMs) for valuations where the estimate is associated with extremely high precision, or high confidence level.
This means they can have a narrow view of the losses they would incur if the mortgage was not repaid and the bank had to resell the property.
However, if the borrower had a stellar credit history and affordability – and hence are extremely unlikely to default on their mortgage – there might be some leeway in relaxing the expected narrowness of the AVM result.
This would allow the lender to automate the valuation process in more cases.
Decisions for customers sooner
An example of this is an existing customer with an excellent credit history borrowing to renovate a rundown property in an otherwise gentrified area.
Because of the lack of transactional history on that particular property the AVM might allocate a lower confidence level to its estimate, perhaps only marginally below the bank’s threshold for instructing a full physical valuation.
The applicant is still extremely likely to repay the mortgage and the bank is extremely unlikely to have to repossess it and potentially suffer a loss.
A credit-adjusted AVM policy makes a lot of sense and could unlock additional digitalisation of the mortgage application process without impacting a lender’s risk appetite.
By bringing these decision points upstream as much as possible, banks can provide outcomes to their customers sooner in the process.
Better customer journeys
Of course, these policies need to be defined within a clear risk and governance framework and with a clear understanding of the impact on the risk management of the mortgage book.
However, if done correctly, this strategy could result in better customer journeys for some segments of banks’ customer base.
Additionally, more can be done to test the correlation between collateral and risk models.
For instance, an AVM’s confidence level is a reflection of the market liquidity of a particular area, which can also relate to a borrower’s ability to cash in and move on in case of payment difficulty.
It would be interesting for lenders to test whether the AVM’s confidence level adds predictability to their credit scoring models.
Generally, a better understanding of a bank’s customer base and segments, and tailoring of credit and valuation strategies to these segments, will result in better customer journeys and automation.