Risk Markets Technology Awards 2021: a different kind of virus

Tech vendors played a key role as markets – and market participants – adjusted to Covid-19

Risk.net montage

A year into the Covid-19 pandemic, it is clear that the technology sector has been one of the big beneficiaries of changing consumer and corporate behaviour. That doesn’t mean it’s been easy for tech providers to embrace the opportunity.

As an example, take the surge in cloud usage driven by a newly minted army of home-workers: Microsoft saw usage of its Teams videoconferencing and collaboration application skyrocket, as well as its virtual desktop and Xbox Live gaming, creating unprecedented demand for its Azure cloud platform on which these applications operate. In addition, Microsoft had to hive off capacity for organisations providing emergency, medical, research and other pandemic responses.

What did this mean for existing, heavy users of the platform? One of these was Milliman. The firm’s actuarial services platform, Integrate – with more than 50 clients in the cloud – can require around 150,000 cores (processing units within a chip) for its asset and liability calculations.

Microsoft organises Azure’s capacity by region, but Integrate’s peak demand is such that it cannot be met by a single Azure region, so Milliman had already engineered its platform to distribute peak loads across regions. However, the pandemic required further intervention.

“In response to the scarcity of resources brought about by cloud saturation from Covid-19, we made improvements to optimise which regions are used by Integrate,” says Pat Renzi, principal for life technology solutions at Milliman. “We distributed our workloads to regions with spare capacity and ensured our clients were able to run all their models without encountering any capacity issues.”

In the first three months of 2020, Integrate processed more than 25 million hours of actuarial analysis in the cloud.

Milliman also optimised how it allocates models to the cloud’s ‘virtual machines’ – software, hardware or combinations of the two that emulate physical computers. Instead of a one-size-fits-all approach, which can lead to significant wasted capacity, or small virtual machines struggling with large models, Integrate now dynamically allocates models to appropriately sized machines.

“It was critical to provide this elasticity without increasing waste, which required significant engineering,” says Renzi. Integrate clients typically consume around 1,800 core years each quarter, and the system is now able to provide this with only 432 core years of overhead. “In addition, we found ways to leverage that overhead for other purposes,” says Renzi.

We distributed our workloads to regions with spare capacity and ensured our clients were able to run all their models without encountering any capacity issues

Pat Renzi, Milliman

During the pandemic, Milliman has provided the option to donate ‘waste’ resources towards Covid-19 research via a distributed computing project called Folding@home.

Milliman won the ‘best use of cloud’ category in this year’s Markets Technology Awards (MTAs) for this re-engineering of its Azure footprint – but the implications of the pandemic cropped up repeatedly, throughout the 28 categories.

This article brings together some of the highlights of the MTAs. In some cases, software had to be tweaked to address the spreading disease; in others, the impact was on customer support or product development plans.

One example of the former came at Conning, where the firm had undertaken a review and update of the calibration for its economic scenario generator in 2019, prompted in part by the prolonged low interest rate environment. It reopened that work last year as the implications of the pandemic became clear.

“[The new calibration] was already in the pipeline prior to 2020, but we implemented a mechanism for producing much more tail dependence than was previously possible,” says Joe Golaszewski, vice-president for risk solutions at Conning, which won market scenario generator of the year. “This allows for the simulation of joint crisis events of the type that we have frequently observed over the last 50 years. With this enhancement, severe events across multiple asset classes and geographies can be simulated, closely mimicking real crises.”

[The new calibration] was already in the pipeline prior to 2020, but we implemented a mechanism for producing much more tail dependence than was previously possible

Joe Golaszewski, Conning

In May 2020, Conning back-tested the calibration across 20 economies, covering interest rates, equity returns, credit spreads and foreign exchange rates, and found that in 96% of cases the impact of Covid-19 was captured.

In the case of Crisil, work done with something else in mind – an attempt to standardise how banks expand the perimeter of their regulatory stress tests – is now being applied to the pandemic.

Currently, most banks’ modelling processes for scenario expansion are resource-intensive, says Stephen Knights, director for risk intelligence and solutions ecosystem at Crisil. They typically require considerable manual effort, are undertaken regularly and involve large amounts of market data and documentation. Crisil saw the opportunity to automate and mutualise parts of the process, and, with the help of five major bank clients, developed Scenario Expansion Manager (SEM) – a centralised modelling framework to design, expand, analyse and track all regulatory and internal scenarios, with the additional ability to share anonymised scenarios.

“This results in cost efficiencies through avoiding the potential duplication of each bank building its own set of scenario expansion models and infrastructure,” says Knights. Models are automated and reusable, reducing the amount of manual work involved in each subscribing bank’s scenario expansion processes.

Crisil has successfully concluded proofs of concept with its partner banks and is implementing at some. It is also in discussion with several other banks interested in joining the platform, while partner banks are piloting the potential of sharing Covid-19 scenarios on the platform. Crisil is also developing a climate risk stress-testing environment for SEM with models, data and benchmarking, which it aims to launch in the first half of 2021.

Remote control

Many of the changes to business practices that the pandemic has forced are here to stay, says Manav Garg, chief executive officer and founder of Eka Software Solutions. The global nature of Eka’s business – it provides software to commodity producers and traders – spurred the company to develop a remote implementation methodology for its cloud-based system well before the Covid-19 outbreak.

Early in the development process, it realised the methodology needed to include clients and its business partners as well. The approach has not only served the company well during the pandemic – it rolled the platform out to 15 customers in 2020 – but Garg believes it offers a blueprint for systems implementation in the industry post-Covid as well.

“Remote working has changed things for good,” says Garg. “With workers being remote, communication and collaboration channels will need to be amplified. The processes for trading, procurement and supply chain execution that were previously locked in a mix of manual steps, disconnected systems and inaccessible spreadsheets will need to make the shift to digital to help businesses achieve greater visibility into their operations. It is becoming increasingly clear that businesses across the board will pursue an aggressive digitisation agenda where remote implementations will be the new reality,” says Garg.

Manav Garg, Eka Software Solutions

An essential element of remote implementation is effective self-service for clients, such as setting up new users or modifying workflows. “Gone are the days when an employee could walk down to the IT desk for support,” says Garg. “In this new environment, businesses are looking for solutions that provide an intuitive experience, making it easier for users to adapt naturally. From a solution provider perspective, this means the user interface needs to be simplified.”

Similar beliefs informed an overhaul of the user interface for IBM’s OpenPages with Watson – a governance, risk and compliance (GRC) application. Again, this work started prior to the pandemic, but should now support longer-term changes to working patterns.

The original challenge was to make the software easy to use for individuals with a diverse range of roles. The software has tens of thousands of users involved in GRC activities, but only 10% are risk and compliance specialists; 80% are in the first line of defence – so the business lines – while the remaining 10% may be auditors, or other roles in the third line of defence.

“These user groups can have widely differing needs,” says Heather Gentile, director of regtech offerings at IBM, which won best user interface innovation. “Accordingly, we completely re-defined what the user experience should be for OpenPages – intuitive for the user to navigate based on their role and with the ability to support simplified collaboration across all three lines.”

IBM’s radical approach included removing thousands of pages of training and replacing it with configurable dashboards and workflows supported by natural language processing (NLP) and other artificial intelligence. For example, IBM’s Watson Moments examines users’ written description of the risks they want to assess and automatically suggests the most relevant set of categorisations, controls and policies they need to test for.

The new interface includes an always available support ‘chatbot’ that is designed to save users time when searching for answers and to present the results clearly – key requirements when so many employees are working remotely.

Libor, NLP and AI

Pandemic or not, life carried on in other respects – even when the industry might have preferred a pause. One example is the slow, but relentless, march towards the demise of the Libor family of interest rate benchmarks.

As an indication of the headache this creates for vendors, the US industry working group on Libor transition last year compiled a guide to transitioning internal systems, breaking down the work into 10 categories and 50 sub-categories – from trade capture and lifecycle management, to counterparty credit risk management, to independent price verification. Of these sub-categories, 70% affect Murex’s MX.3 platform and applications in some way.   

Alexandre Bon, group co-head of Libor and benchmark reform at Murex, calls the reform “the most impactful and challenging regulatory development we have seen”. Murex won best support for Libor reform, alongside five other categories.

Staying on top of the Libor transition work required early planning. In March 2018, Murex pulled together a large global team of multi-disciplinary experts to analyse the ramifications and to map out enhancements for its software, services and pre-packaged system configurations.

The resulting package provides support across front office, operations, risk management and finance. It includes transition mechanisms for OTC derivatives, loans and securities – allowing users to run transition scenarios by bulk or by individual trade/security. And it adds new market instruments and curves when new risk-free rate indexes become available.

Murex

Alexandre Bon, Murex

But the work isn’t over, says Bon. Much of the focus to date has been on cash products and linear derivatives. As options and swaptions come into focus, so will a host of new complications.

“As risk-free rates adoption progresses, financial institutions must manage multiplying product variations and conventions,” says Bon. “Further innovations around non-linear RFR derivatives are the next logical step, introducing new modelling challenges. Firms must also analyse value transfers associated with various fallback and transition options for Libor-referencing contracts. These transition mechanisms then require a seamless implementation addressing the technicalities of both standard and bespoke fallback arrangements, front-to-back.”

Elsewhere, fintechs are continuing to exploit artificial intelligence (AI) and its associated technologies of machine learning (ML) and NLP to solve industry problems.

Robot reading of corporate documentation has become so commonplace that companies are now deliberately making them more machine friendly. However, as the software not only extracts the relevant information, but also increasingly searches for indicators of the company’s underlying health and prospects, companies are starting to modulate what they write, avoiding words that algorithms are likely to interpret as negative sentiment, according to a recent paper by the National Bureau for Economic Research.

“Just like some companies have practiced greenwashing, it is expected that some will try to game automated textual analysis systems,” says Sylvain Forté, chief executive officer of SESAMm, which won best use of NLP.

The possibility that companies can manipulate all produced documents, publications and articles on the internet is very low

Sylvain Forté, SESAMm

Manipulating sentiment interpretation is easier said than done. Companies cannot know for sure what type of analysis will be done, how the machine learning algorithms are trained, or exactly what they are looking for in documents. SESAMm further counteracts attempts to manipulate interpretation by casting its net much wider – the neural networks in its TextReveal platform read and monitor more than three million data sources on public and private companies, financial assets and other factors.

“The possibility that companies can manipulate all produced documents, publications and articles on the internet is very low,” says Forté. “We gather as much data as we can and limit the risk by mixing information from several sources, including press or social media.”

By doing this, TextReveal seeks to spot inconsistencies and contradictions between corporate claims – such as in ethics and social responsibility policies – and the realities of their business practice. In cases of fraud, for example, whistleblowers will typically share information outside of the company’s control on social media and anonymous communities that TextReveal extracts information from, says Forté.

Ultimately, Forte believes the dangers of manipulating corporate information outweigh any potential short-term gains, because the truth “will be detected sooner or later and those caught red-handed risk paying a heavy toll for doing this”, says Forté.

The award for best use of machine learning/AI went to Tradeteq, which is seeking to make the trade finance asset class more investable.

With Brexit, there are many SME suppliers to larger businesses in the UK and EU that may be potentially affected by delays to shipping and transporting of goods

Christoph Gugelmann, Tradeteq

Currently, financing is still dominated by banks, and investors face a number of hurdles in trying to enter the market, including high operational costs, manual origination processes, lack of standardisation and, most critically, a lack of transparency. Of the $8.5 trillion in bank-intermediated trade finance in 2018, less than $100 billion was sold to non-bank financial institutions, Tradeteq estimates.

The heart of the problem is that traditional credit processes are ill-suited to the granular, short-term nature of trade finance and trade-finance specific risks such as dilution (incomplete repayments) and fraud, says Christoph Gugelmann, founder and chief executive of Tradeteq. To solve this, the company built an AI engine for its electronic trading platform that uses machine learning to create models for analysing company and transactions risk.

Unlike traditional credit scoring approaches, Tradeteq’s models are not limited to accounting data for their input and are tolerant of large proportions of missing data. This allows them to assess a far wider proportion of the market. For example, while global credit rating agencies are covering around 20,000 companies worldwide, the company’s models cover 3.5 million limited companies in the UK alone, says Gugelmann.

The transaction models are based on a recognition that in many markets registration and accounting information for small and medium-sized enterprises (SMEs) is unavailable or unreliable. Instead, Tradeteq’s AI uses supply chain flows to assess risks of each trade finance transaction.

Tradeteq includes in its assessments risk propagation up the supply chain, where a company’s inability to pay its suppliers may lead to the suppliers in turn being unable to pay their suppliers. It also monitors for incorrect or untimely deliveries down the supply chain that can result in a cascade of dilutions on trade finance receivables. Brexit is one of the scenarios where this is useful.

“With Brexit, there are many SME suppliers to larger businesses in the UK and EU that may be potentially affected by delays to shipping and transporting of goods, or mistakes with paperwork, or simply the build-up of traffic around Dover,” says Gugelmann. Larger exporters could see delays in goods deliveries to their overseas customers, resulting in receivables payment delays, with the delays passed onto UK SME suppliers. “Our AI tool is monitoring propagation of this kind of stress throughout the supply chain and may be able to raise an alarm before the risk is realised,” says Gugelmann.

“Trade finance processes desperately need improving, especially in this era of pandemic and trade wars,” remarked one of the judges of this year’s awards.

Risk Markets Technology Awards 2021

For the second year in succession, Murex was the big winner of this year’s awards, taking six categories – three for pricing and trading technology, with other wins for counterparty risk, Libor reform and systems support.

The other awards were shared across a mix of big, established trading and risk vendors – such as Bloomberg, FIS and Oracle – and younger, or specialist firms, such as TransFICC and commodity player Eka Software.

In total, there are 28 awards in this year’s MTAs. Entries were invited for a further seven categories, but there were either too few entries in the categories – or no compelling entrant. There was one tie, in the buy-side risk management product category, between Qontigo and RiskSpan.

An overview of the awards methodology, and a list of the judges, can be found below this year’s roll of honour.

TRADED RISK:

Market risk management product of the year: Oracle

Market liquidity risk product of the year: Bloomberg

Counterparty risk product of the year: Murex

FRONT-OFFICE REGULATION:

FRTB product of the year: ActiveViam

Regulatory reporting product of the year: SteelEye

Best support for Libor reform: Murex

PRICING/TRADING TECHNOLOGY:

Pricing and analytics: fixed income, currencies, credit: Quantifi

Pricing and analytics: structured products/cross-asset: Murex

Trading systems: commodities: Eka Software

Trading systems: equities: Murex

Trading systems: fixed income, currencies, credit: Murex

BUY-SIDE TECHNOLOGY:

Buy-side market risk management product of the year: Qontigo/RiskSpan (tie)

Performance attribution product of the year: Confluence

Best execution product of the year: BestX

Buy-side ALM product of the year: Moody’s Analytics

Market scenario generator of the year: Conning

DATA AND OTHER SPECIALIST CATEGORIES:

Market data vendor of the year: FIS

Alternative data vendor of the year: FactSet

Risk data repository and data management product of the year: Cloudera

Electronic trading support product of the year: TransFICC

Best vendor for system support and implementation: Murex

BACK-OFFICE CATEGORIES:

Central counterparty clearing support product of the year: Nasdaq

Collateral management and optimisation product of the year: Vermeg

INNOVATION CATEGORIES:

Best user interface innovation: IBM

Best use of machine learning/AI: Tradeteq

Best use of natural language processing: SESAMm

Best use of cloud: Milliman

Best modelling innovation: Crisil

---

Methodology and list of judges

Technology vendors were invited to pitch their products and services in 35 categories covering traded risk, front-office regulation, pricing and trading, buy-side technology, back office, data and other specialist areas. Candidates were required to answer a set of questions within a maximum word count about how their technology met industry needs, its differentiating factors and recent developments. More than 170 entries were received and shortlisted. 

A panel of 12 industry experts and Risk.net editorial staff reviewed the shortlisted entries, with judges recusing themselves from categories or entries where they had a conflict of interest or no direct experience. The judges individually scored and commented on the shortlisted entrants, before meeting in October to review the scores and, after discussion, make final decisions on the winners.

In all, 28 awards were granted this year. Awards were not granted if a category had not attracted enough entrants, or if the judging panel was not convinced by any of the pitches.

This year’s judging panel consisted of:

Laura Barrowman, chief information officer, Credit Suisse

Peter Burgess, independent adviser

Sid Dash, research director, Chartis Research

Clive Davidson, contributor, Risk.net

David Germain, group chief information and technology officer, RSA Group

Ian Green, independent consultant and chief executive officer, eCo Financial Technology

Ahimsa Gounden, senior specialist, market risk: Liberty, Liberty Group

Jenny Knott, CEO, Fintech Strategic Advisors

Peter Quell, head of portfolio analytics for market and credit risk, risk controlling, DZ Bank

Hugh Stewart, independent consultant

Ed Wicks, head of trading, Legal & General Investment Management

Duncan Wood, global editorial director, Risk.net