Automation of the commodity trading middle office: a threat or opportunity for risk managers?
05 June 2018
We all know that disruptive technologies are coming and PwC, like our clients, is investing significant time in understanding the potential impact of robotics, Artificial Intelligence (AI) and other advanced technologies on the commodity trading and risk management landscape.
These radically different and improved ways of collecting and leveraging data and information have the potential to revolutionise these activities. Artificial Intelligence, such as machine learning, and predictive analytics will significantly improve the ways vast amounts of data are converted into real-time, high quality information.
The common mantra is that the impacts of technological advances are typically exaggerated in the short-term but underestimated in the much longer term. While that is generally true, in this case we are now seeing clients developing ‘proof of concepts’ around these technologies and applications that would have been considered ‘blue sky’ only months ago. Additionally, according to recent PwC research amongst CROs, many of the technologies described in this blog are starting to be applied or appearing on the radar of those working in the middle office causing them to think seriously about how their future operating model may change, particularly around people. Of course people will still be needed, but their skills and capabilities are likely to be very different from those possessed by the current workforce as such technologies will allow them to spend less time processing and more time analysing data.
So what are some of real practical areas of disruption and enhancement in the Commodity trading middle office?
The first is around risk and position reporting and limit monitoring. A prerequisite of automating risk and position reporting is real-time access to large amounts of structured data. ‘Data lakes’ can solve this problem as unstructured and structured information is collated in a cloud-based repository allowing for ease of data mining when needed. The difference between data lakes and ‘old fashioned’ data warehouses is that such software makes it possible to query data, held in a variety of formats, for example spreadsheets, pdf, images, etc. Hence data can be analysed at a faster pace and with greater flexibility as compared to a structured data warehouse architecture solution
Natural Language Processing (NLP) involves algorithms that are capable of reading and ‘understanding’ email and unstructured text fields, allowing the content to be categorised and specific terms, e.g. price/quantity) to be identified. This technology could be used to read a trader’s free text blotter which could then be used to validate and improve the accuracy of deal entry. Alternatively NLP could be used to read and classify settlements e-mails, identifying disputes and what they relate to allowing settlements analysts to resolve disputes faster and speed up cash collection.
Monitoring physical commodities and position reporting could also be enhanced. The Internet of Things (IoT) refers to the interconnectedness of computing devices embedded in physical objects, enabling them to send and receive data. To improve market intelligence, sensors could be connected to the IoT, collecting a wide range of data, such as on commodities in transit or harvest progress. Drones – equipped with cameras and sensors – could inspect crop conditions, inventories, ships or port congestion through capture of unprecedented levels of high-resolution imagery. For more information, see the PwC global report on the commercial applications of drone technology.
Machine learning can be used to enhance predictive analytics. Machine learning studios make use of neural networks and complex algorithms to analyse historical data to make predictions over future outcomes. There are a number of commercial applications which could be utilised across the front office – for instance by analysing the impact that weather has historically had on power markets, both in terms of power load and power production, machine learning could be used in conjunction with both long and short term weather forecasts to improve prediction of power prices. Machine learning also has extensive use cases across the middle office particularly with regards to model validation and identify correlations/risks that would not otherwise be identified. In order to get the most benefit from this it is essential that there is a feedback loop into front office models such that learnings identified by risk/middle office teams can be used to inform better decision making in the future.
The above are just a small sample of the potential implications. We’re expecting the speed of change caused by these technologies to continue to increase significantly and for more success stories to emerge and grow and those in the middle office who embrace this change to operate in a leaner, faster and more analytical manner.