Sign Out
Logged In:
Latest Blog Posts
Tab Image

Tab Image

O.R., analytics and risk

Friday, 4 Dec 2015
Jeffrey Jones

John Hopes

It may be to do with getting older, but I have recently been thinking about risk. This is, of course, in the context of O.R. and analytics where I believe there are some particular opportunities.
In the world of corporate finance, for example, a key focus is on shareholder value, which is increased by enhancing revenues, cutting costs or reducing risk. And the joy of O.R. and analytics is that they can make big contributions to all three of these. But since the financial crisis and recession there has, in my perception, been a particular and understandable emphasis for OR and analytics on the cost reduction agenda in both the public and private sectors, whether through smarter ways of distributing product, more efficient ways of deploying people on tasks, reductions in working capital including inventory or pruning branch networks. In all these cases and others O.R. has been used to identify the best ways in which the business can deliver the same for less or more for the same.
This is not to say that O.R. and analytics have been absent from revenue enhancement, for example in optimising pricing strategies, marketing spend, promotions and website interactions; it is just that it seems as if the higher priority has been cost reduction. But what about the third leg of the value stool: risk reduction? This is an area where O.R. and analytics have some unique selling points in providing tools for identifying and quantifying risks. And this is what I wanted to cover in this article, because I think it is time for us to raise the profile of O.R. and analytics in the context of risk.
First, in terms of demand, there seems to have been a management progression from cutting cost in order to survive to chasing growth in the upturn to the beginnings of a perception that risk management focus might have been lost. This has been highlighted by some headline risks including cyber-attacks, the VW emissions fiasco, Libor rigging and commodity price crashes. But across the board there are signs of a return to a need to reflect risk in decision making in a more systematic way.
As is so often the case for O.R. a lot of this is not about solving new problems, but is often more about applying techniques and methodologies that have been around for decades. There is, however, an added opportunity in the area of advanced analytics that arises from the recent data explosion and the rapidly expanding internet of things.
Starting at the more long-established end of the spectrum, my personal observation is that 20 years ago a great deal of what I did as an OR analyst and modeller had to do with risk quantification. There also seemed to be a lot going on in O.R. and risk generally in those days as championed by Chris Chapman at Southampton and others. Monte Carlo simulation and other risk modelling techniques were being used to quantify such diverse things as nuclear power station decommissioning provisions, major project cost estimates or the risk transfer associated with PFI schemes or privatisations. Indeed Monte Carlo simulation is yet another of those OR techniques that has gone so mainstream that most of its users do not think of themselves as OR analysts. For example, in complex security valuation the complexity does not have to become very great before simple approaches such as Black Scholes cannot handle the problem and Monte Carlo simulation becomes the preferred approach. And yet very few quants would think of themselves as using an O.R. technique.
In the nineties I remember running workshops for clients entitled “why are projects always late”. This started with a simple example in which two people are planning to meet at 9am, with both of them being just as likely to be up to five minute early as up to five minutes late. When asked what time, on average, the meeting will start it is amazing how many experienced project managers say nine o’clock. And this probably goes to show why meetings and projects are still always late. There are a number of O.R. techniques to allow project plans to reflect risks, options and dependencies, but they are still generally not deployed, as a result of which management make decisions with an incomplete information set.
In terms of the newer opportunities, some of the most interesting and, more importantly, high value applications of advanced analytics are in the risk domain, including using machine learning and other techniques to identify fraud or evidence of cyber-attack. Similarly, with sensors distributed through manufacturing or other operations, predictive analytics is being used to estimate risks of process or component failure. Other opportunities range from quantitative support to bank stress testing to quantifying supplier risk. The digital age is creating major new risks but is also providing far more data and new mechanisms for managing both these and more conventional risks.
But what does all this mean for the O.R. Society? Well, there seems to be less on risk management applications at our conferences than there once was, although I could be wrong. Also, on our training programme, there is one course on decision and risk analysis, so we certainly have something, but there could be more. And when it comes to special interest groups and publications, well the best that could be said is that the topic is embedded in them somewhere. But most importantly, risk is another high profile management issue where OR and analytics should have a point of view which is expressed loudly.
In conclusion, in some ways OR / analytics and risk is just an example of a more general rule. We often seem to be solving the same old problems for a new generation of decision makers. But, at the same time, there are some genuinely new and highly important problems to solve, particularly arising from the impact of digital technology and the data explosion. This is an ideal combination to ensure that OR and analytics remain relevant.


  2018 (24)
  September (1)
  2017 (15)
  2016 (14)
  2015 (8)
  2014 (5)