Financial services
Feature:
Investment management – analytics
25 April 2008
With new technologies transforming the investment industry, Jacqui Griffiths looks at how firms can maximise the speed and accuracy of their investments.
Investment management is an intricate business. Research, analysis, dealing, settlement, and reporting all have to be done correctly, and mistakes can cost millions. This difficult terrain is becoming more challenging as regulations such as the Markets in Financial Instruments Directive (MiFID) transform the industry to a global marketplace where entrants of all sizes can compete.
Every millisecond that can be shaved off of the time it takes to trade is a potential step ahead of the competition. But the evolving industry is also an unfamiliar territory, as new exchanges enter the market and all players develop ever-more sophisticated ways to analyse and execute trades, and to stay informed on the activities of their rivals. In this intensely competitive landscape, successful navigation demands speed and a certain amount of stealth.
In order to achieve this, financial institutions need to give serious thought to their IT. Legacy systems simply can’t handle the volume of information from various channels fast enough to ensure competitive trading. Indeed, according to research by TowerGroup, updating or replacing legacy systems and improving infrastructure are key priorities for brokerage firms. “Firms are motivated by drivers like revenue growth, regulation, technology innovation and IT management,” says analyst Dushyant Shahrawat in TowerGroup’s report, 2007 Top 10 for brokerage firms: business drivers, strategic responses and IT priorities. “One area in which firms are seriously considering automating processes and replacing legacy systems is derivatives, because enormous volumes mean that creaking applications and infrastructure are holding the entire business to ransom. Revamping the infrastructure has become a crucial priority to support next-generation trading efforts and to leverage new approaches in technology like grid computing, server consolidation, and hardware optimisation.”
Many companies are turning to automated methods for gathering and analysing information, to enable split-second responses. “Automated trade execution has been around for many years, and was traditionally the domain of proprietary trading desks in investment banks and specialist hedge funds,” says Jean Williams, product business owner at Asset Control. “In the last few years, larger brokerage houses have made their trading algorithms available to investment managers, usually at a lower cost than manually worked orders. Algorithms have allowed brokers to reduce the number of active traders, but led to an investment in mathematically skilled analysts and programmers to develop and test these algorithms. As a result, investment managers benefit from reduced execution costs, timely execution of larger orders, enhanced trade analytics, and compliance with regulations for best execution and MiFID.”
In addition, continues Williams, some institutions use more specialised automated investment decision-making: “Many hedge funds use quantitative models that determine which instruments to trade, as well as how and when to trade them. These practices are almost never used by traditional long-only investment managers. Practitioners of automated investment decisions often cite that their quantitative methods remove any inherent bias that is introduced by manual analysis of markets, and claim that their investment decision-making process is 100 per cent repeatable. In practice, the markets are constantly changing and models always need to be modified.”
“The popularity of automated trading is due to a number of factors,” says Brian Sentance, chief executive officer of Xenomorph. “For example, there are cost savings involved: why pay an expensive trader to manually implement something that a computer can do? And once automated, trading algorithms are commercially scalable – they can be made available to more users, easily and at low cost.”
The technologies most widely used in the field of automated trading are often referred to as complex event processing – tools that can be used by business users to define event-based decision processes to help automate day-to-day activities. “In order to automate investment decision making, the challenge remains the collection of vast amounts of data from disparate sources, the cleansing and validation of that data, and presenting it to a proprietary algorithm for analysis,” adds Williams. “The quality of raw investment data is rarely sufficient to support these processes without a supplementary data management system.”
“The processing power and delivery technology are now here to deal with the ever-increasing amounts of data involved,” says Sentance. “Trading venues are becoming fragmented, as regulations such as MiFID and RegNMS increase the number of venues for trading securities, and hence the number of places where prices should be searched for best execution of a trade. In addition, the decimalisation of trading prices in the US in recent years has decreased average trade transaction size but resulted in dramatic increases in the number of transactions placed.”
“The exponential growth in the amount of data a financial firm must process and the shortening useful lifespan of any new algorithm continues to place incredible demands on capital markets firms,” says Andy Hughes, director of investment management solutions at Microsoft. “Excel has long been the standard for analysing data and constructing financial models, and Microsoft has made significant investments in new products such as Windows Compute Cluster Server 2003 and SharePoint Server 2007 with Excel Services, which, when used together, can help customers continue to build their financial models quickly, share them securely with colleagues and trading partners, and deliver calculation scalability and fault tolerance across many machines containing multiple processors.”
Of course, trading is a context-sensitive as well as a time-sensitive business. The seemingly inexorable rise of the algorithm has raised questions – for example, how will the role of humans evolve alongside these technologies, and how accurate are automated processes in the face of unexpected events?
“Increased automation has already had a significant impact on the roles of human traders,” says Williams. “Statistical testing of algorithms is frequently undertaken and the overall levels of performance are generally superior to those of a human trader. Due to their investment in automated systems, brokerage houses have fewer traders processing orders. However, additional investment is usually required in the areas of risk management, audit and compliance.”
In truth, algorithms are unlikely to automate people out of a job – rather, the human role in investment management and analytics will evolve with the technology. “Algorithms are not foolproof,” says Williams. “Unexpected events will eventually occur in any market. The best an algorithm developer can do is to try and minimise the impact of an extreme situation.”
Key to this are both rigorous testing and talented people. “Models can only be tested against situations that have occurred in the market in the past, using massive databases of price movements going back many years,” adds Williams. “Consequently, models are only as good as their analysts, programmers and testers, although certain market events are often blamed on poorly implemented models – for example, the Japanese market crash as a result of unchecked computer-generated sell orders; or the recent sub-prime credit crunch resulting from incomplete risk analysis and valuations of credit securities.”
“There is still very much a place for human intuition, creativity and ingenuity in putting together trading ideas,” says Sentance. “Certainly, there will always be new, more complex products that will be handled manually until they become standardised and more mainstream. That said, automated trading will dominate and continue to grow through all of the vanilla products and markets.”
Sentance continues: “Some of the current issues in automated trading include data quality, with the usual adage that bad data means bad decisions. Another issue is time to market with an idea – while low latency is a key technical concern in being ‘fastest’, how quickly a new trading idea can be put into production in the market is a key concern. The usability/ease of use with which traders can create, simulate, back-test and deploy a new trading model is vital in this regard. The area of back-testing seems to be under-resourced at the moment, and we believe this will be a growth area for technology.”
As Williams points out, increased automation has resulted in significant number of new tools, typically point solutions that lack integration with existing systems and processes: “Modern investment managers must develop and manage an architecture to support investment operations that not only directly addresses the management of information assets within the firm, but also complements the unique workflows of these best-of-breed solutions.”
Sentance believes that automated trading will make its mark on the shape of the industry. “There is likely to be increased consolidation in the broker market, as the battle becomes a more intellectual and technical one with stiff competition facing those institutions that do not have the resources and expertise to differentiate their offerings to clients. Additionally, through automation, more exotic instruments will gradually become accessible to a wider investment base, resulting in yet more assets to trade.”
There can be no doubt that now is an exciting transitional period for investment managers, and that this is no time for them to rest on their laurels. “In recent years, the amount of trading analysis undertaken has grown tremendously,” observes Williams. “The performance demands placed on investment managers requires them to access new markets and to trade new instruments. Although tools are increasingly available to support the analysis required for these requirements, the underlying problem still lies in accessing all the necessary data when it is often siloed within an organisation. This accessibility and deployment issue has resulted in an exponential demand for front-office focused data management solutions that can rapidly adapt to new requirements as they arise.”
This article was originally published in the Spring 2008 issue of Finance on Windows magazine.