This page contains a Flash digital edition of a book.
In Focus Risk

‘The computer says yes’

Change is coming for the credit-risk community as the speed of technological development increases

Jonathan Baum Chief credit officer, Lloyds Banking Group jonathan.baum

The recent world economic forum, in Davos, dedicated time to visionary and very practical topics spanning artificial intelligence, machine learning, and big data.

Harnessing such approaches to better understand customers, from a credit and fraud-risk perspective, and to dynamically adapt models and strategies, may to many seem a step into the unknown. However, to those seeking to maintain competitive advantage, often in cost-constrained environments, this may be a fertile area to explore; albeit with a degree of caution.

Machine learning

The mathematics behind a variety of machine-learning techniques is not new; and both academics and industry practitioners can, relatively easily, demonstrate empirical benefits of additional credit-risk insight; leveraged from daunting sounding modelling techniques such as ‘random forests’ or ‘stochastic gradient boosting’. In environments where there are ever richer sources of data, the scale of improvement becomes even more tantalising. Recent trials undertaken, suggest uplifts in model power of up to 10%, which, when applied across portfolios, begin to generate many more opportunities to meet customer needs profitably and safely.

As initiatives, such as Open Banking, further increase access to transactional data, the ability to consume it efficiently and seamlessly will continue to improve customer credit-risk decisions.

So why is the credit-risk community not already deploying machine-learning techniques, rather than the more tried and tested regression-based scorecards? Is it perhaps the absence of understanding of the models, the computational hunger


In environments where there are ever richer sources of data, the scale of improvement becomes even more tantalising

Potential improvements As analysts experiment with such methodologies, now embedded in commonly available statistical platforms, appreciation of the potential improvements is driving increased experimentation and consequential understanding of better approaches. Also, although, from a computational perspective, machine learning undoubtedly requires more powerful and efficient infrastructure to deploy and execute the models, the pace of adoption of these is ever accelerating. The incorporation of such capability into commercial credit technology is underway, and the possibility of rapid deployment via cloud technology makes it more accessible and in faster timescales. We are already on

In conclusion, we are already evolving our ability within the credit-risk community to interrogate data in a dynamic and automated fashion; the use of advanced, self-learning modelling techniques such as machine learning and AI is, in many ways, an extension of that journey.

As organisations adapt and evolve, those that do not are feted to, ultimately, fall by the wayside.

We can expect the rate of evolution and the capabilities of credit-decision support systems to advance quicker and faster in the next five years than it has in the past 30 years, and it is our job as credit professionals to plan and prepare for this evolution. CCR

March 2017

required to operate more advanced models, or a perceived lack of transparency of such models? In many ways, it is a combination, or indeed all three of these elements. So how do we overcome these barriers to realise improved machine-based credit- risk decisions and, ultimately, better risk assessment of our customers?

Can we learn from recent history when credit scorecards were new on the scene and treated with caution; or even disregarded by those who were experienced in the more traditional, manual credit- assessment processes?

a journey towards both increased customer transparency into how decisions are made, and increased regulatory oversight of the soundness of models.

Machine learning may, indeed, use more complex algorithms, but the principles of sound model governance apply equally. The man or woman on the Clapham omnibus will derive no more insight from being told than we have applied ‘logistic regression’ than ‘recursive partitioning’. Either way, firms are already developing much simpler and more friendly ways of sharing important facts with customers.

Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56