Unless you are living on a remote planet in a galaxy far, far, away, you have probably used a computer to interact with an algorithm at some point today.
You use face-detection algorithms when you tag somebody on your mobile phone; you expand the scope and range of algorithms when you make an online purchase or access a cash machine; you allow algorithms to make decisions on your behalf when you check your email (and don’t look at what’s been diverted into the spam filter).
Hype around how Facebook, Google and others are using search and self-learning algorithms can make it seem as if every algorithm is cutting edge. It is not.
An algorithm is any specific method, process or set of rules that can be used to complete a clearly defined task, which could be processing data, solving a problem or achieving a goal. Self-learning algorithms take things to another level. Indeed, The Treasurer recently explored the implications for the future of treasury roles. See also ACT chief executive Colin Tyler’s blog from earlier this year, exploring how innovation will transform the treasurer’s role.
The potential of ‘smart’ self-learning algorithms is massive. At Essex University, Professor Maria Fasli is using multiple, goal-oriented, self-learning algorithms called agents to analyse big data on the behaviour of the massive and complex FX market. Her analysis plays out over four stages:
“Traditional software applications need to be told explicitly what it is that they need to accomplish and the exact steps that they have to perform. Agents need to be told what the goal is, but not how to achieve it,” says Fasli.
Being smart, they actively seek ways to satisfy this goal, reacting to changes in the environment as they occur, then modifying their course of action to accomplish their goal. “We will increasingly be delegating tasks and goals to such systems and interacting with them,” she predicts.
Many of us now depend on rule-based algorithms, personally and professionally. “They are already in use in cash pools, zero-balancing pools, notional pools, cash concentration and that sort of thing,” says Sanjay Bibekar, treasury technology lead at PwC.
Traditional software applications need to be told what it is that they need to accomplish and the steps that they have to perform. Agents need to be told what the goal is, but not how to achieve it
With cash concentration, for example, software uses algorithms to ensure that sub-accounts have a constant/zero balance and that debit and credit balances are automatically physically swept into the concentration account (when local regulation allows).
Banks and treasury management system (TMS) developers have been using algorithms for years to offer the cash management functionality in their products and services. There is nothing new about software programs that enable treasurers to complete tasks more efficiently and effectively with the help of algorithms.
Most of the time we use them without thinking about this; particularly when we are not alerted to their presence by the names of products and services and practices, such as algorithmic trading.
As algorithms have become part of popular culture and everyday language, it can seem as if they have also been endowed with near-mythical powers, but the widespread focus on cutting-edge applications can be misleading.
Computer-based trading strategies (such as high-frequency trading), which are supported by algorithms, may grab the headlines; nonetheless, numerous treasurers are using algorithms to help them to solve everyday problems and manage core treasury processes.
This does not demand the latest technology or specialist software, though such tools can help.
Once you have devised a method (aka algorithm) to complete a task, you can use productivity tools such as the spreadsheet Microsoft (MS) Excel and other software applications in the family (such as Outlook and Word) to automate this. “I write Visual Basic routines that create reports, move data around and feed it into other systems,” says interim treasurer Chris Fell.
Treasurers who are not on first-name terms with the formulas and functions in MS Excel, or able to programme using Visual Basic, may be surprised by what Fell can achieve.
“I have a spreadsheet database for FX trades, then a routine that takes the most recent line on that spreadsheet, posts individual data points into a Word program, which then produces a confirmation for matching and signature,” he explains.
An algorithm is any specific method, process or set of rules that can be used to complete a clearly defined task, which could be processing data, solving a problem or achieving a goal
Fell also has a routine to move data from Excel into Outlook, where he has set up a maturity ladder.
Like all treasurers, Fell is focused on risk and is keen to highlight the potential strengths of an off-the-shelf TMS. “I tend to work in sectors where transaction values are high and volumes are low, such as real estate and mining. A TMS is more effective when there are many more deals or you are operating an in-house bank,” he says.
Lack of control may also be an issue. “If somebody makes a mistake, they can make it look as if they didn’t, by editing the spreadsheet. With a TMS, data is more secure and you have an audit trail.”
It may be a mistake to allow your aspirations or achievements to be limited by the overblown hype that surrounds self-learning algorithms, or to become distracted by differences of opinion and interpretation around the associated language.
“When it comes to drawing lines between terms of reference like analytics, algorithms and data science, there’s a lot of semantics involved,” says James Murray, who specialises in this area at the recruiter Robert Walters.
This need not be a barrier to exploring rule-based treasury use cases. For example, analysing bank charges, comparing them with the rate card and determining whether you are paying the least you could for them; managing credit and risk analysis of cash and liquidity across multiple cash investment strategies; optimising the control, timing, price and execution of currency orders.
FX is an area where banks are already helping corporates to explore the benefits.
For example, a large corporate can control the execution and risk of a strategy to buy currency for one large (and material) transaction, with the help of rule-based algorithms.
By using bank infrastructure that enables it to: split its currency purchase into small trades; define the start and end time for the strategy and the conditions affecting this; limit the maximum strategy price by setting the frequency and timing of trades; setting the maximum spread to limit the impact of currency volatility and wide spreads on the corporate’s average rate.
Developing the complex algorithms to achieve this is a massive undertaking – about the skills and software that would be needed to make them a reality. Then there are issues such as the need for back testing, manual oversight and security.
The biggest impediment to rapid take-up may well be the treasurer’s primary concern with risk management. Nonetheless, there are still good reasons to keep a watching brief on what algorithms make possible, as this evolves.
Adam Gable, a treasury specialist with the banking software vendor Temenos, has high hopes for the future, which he shares by explaining how a corporate treasury in a global operation could exploit rule-based algorithms to manage liquidity and ensure the best return on surplus.
“Unless there is a clear strategy around this and the ability to execute the same swiftly, it is difficult to achieve optimal returns,” he says. A manual process can be repetitive, exhausting and create delays that lead to potential loss of good opportunities to invest.
“An algorithm that could assess liquidity across the group, identify surplus, assess market rates for the instruments under consideration and then identify the best asset class and provider, even suggesting currency conversions where appropriate, would certainly have legs,” he says.
But there are even more impressive possibilities shimmering on the horizon. He adds: “Introduce machine learning, which, given the time of the month and year, could predict the most optimal trade, and we are in corporate treasury utopia.”
Lesley Meall is a freelance journalist specialising in technology and finance.
This article was taken from the November 2016 issue of The Treasurer magazine. For more great insights, log in to view the full issue or sign up for eAffiliate membership
