Professional Documents
Culture Documents
Img - Applying Big Data To Risk Management PDF
Img - Applying Big Data To Risk Management PDF
APPLYING BIG
DATA TO RISK
MANAGEMENT:
Transforming Risk Management Practices
within the Financial Services Industry
Arnold Veldhoen
Arnold Veldhoen is an Associate Partner with Avantage Reply in The Netherlands. With
20 years of experience in financial services, Arnold has led a number of initiatives in
risk management and system improvements for our clients in Europe. He has lived
overseas for over a decade, managing change and integration projects throughout Asia
for the financial markets business lines at ABN AMRO, BNP Paribas and Macquarie
Bank. Arnolds product expertise lies in Rates, Currencies and Commodities that
are increasingly traded electronically and cleared centrally. At Macquarie Bank, he
held a Global Operational/Enterprise Risk role with projects completed in Chicago,
New York and London. Arnold holds a Masters in Finance from VU Amsterdam and
a postgraduate MBA from Australias business school AGSM. He is a member of the
Executive Committee of Reply Benelux/France, PRMIA Netherlands Chapter and the
Institute of Operational Risk.
Stphan De Prins
As an Associate Partner with Avantage Reply in the Benelux and France, Stphan De
Prins draws on over 17 years experience in developing and implementing cutting-
edge solutions in risk management. Stphan has devised numerous risk solutions
for leading institutions across Europe and North America. Key recent projects have
included creating a Large Exposures and Concentration Counterparty Credit solution
for a Global Custodian, and implementing a risk and regulatory reporting system for
a universal bank. Prior to joining Avantage Reply, Stphan led the development and
implementation of the IFRS Module of FRS Global.
Website: www.avantagereply.com
1
TABLE OF CONTENTS
Introduction 3
Noise or Signal 4
2
Applying Big Data to Risk Management
Introduction
Over 90% of the worlds data has been created in the last two
years 1. Forward-thinking industries and organisations have
Big Data technologies and already begun to capitalise on this gold mine. But what does the
risk management: Big Data revolution mean for risk management?
Put simply, Big Data represents the future in this field. Why? Big
! Improved predictive Data technologies can help Risk teams gain more accurate risk
power and stability of risk intelligence, drawn from a variety of data sources, in almost real-
models time. Within the financial services industry, they can allow asset
managers, banks and insurance companies to proactively detect
! More extensive coverage
potential risks, react faster and more effectively, and make robust
of real-time risk intelligence, decisions informed by thousands of risk variables. Already used
improving monitoring widely across other sectors - particularly in eCommerce - Big
of risk and reducing noise- Data is a true game changer.
to-signal ratios
Big Data is routinely defined as high-volume, high-velocity and
! Strengthened evidence high-variety information assets demanding new technological
based decision-making approaches to organisation and analysis. When applied to risk
capacities across a number management within the financial services industry, we would
of key domains add high-veracity and high-value to this list - effective analysis
of this data has the potential to drive increased accuracy and
! Significant cost savings in reliability, and offers potentially significant cost savings by
risk management combatting the risks that can cost financial institutions billions.
1. http://www-01.ibm.com/software/data/bigdata/what-is-big-data.html
2
3
Applying Big Data to Risk Management
Noise or Signal
Noise or signal
The era of data warehouses - with their For example: imagine you want to
static structures and limited interaction conduct a credit check on a new
paths - is over. The complexity and customer. Data lakes allow a risk profile
variety of sources now available to be developed based on a range of
(including social media, email, sensor data - including customer credit reports,
data, business apps, archives and spending habits, social media profiles,
documents), and the speed required and credit card repayment rates - in
for retrieval and analysis, demand seconds.
fresh new approaches. We are now
moving into the era of data lakes. Worried about fraud on the trading floor?
Rather than manually - and laboriously
- track staff trading actions, data lakes
Time is critical in the allow the retrieval of an instant snapshot
of activity, including information from
new world of risk chat room sites, mobile phones, and
management. even door swipe cards. Suspicious
If you can react to a activity can be identified and stopped
as it is happening, before incurring fines
risk faster, you have a and devastating damage to your banks
competitive advantage reputation.
Jason Hill
Executive Partner - Reply Big Data suggests big promises - but
can it live up to them? It is still very early
days. Nate Silver points out that, with
The approach to data lakes is the amount of data being generated
simple: instead of organising data daily, the noise (may well be) increasing
in a siloed, prescriptive store, you faster than the signal. There are so many
can retain all types of data together hypotheses to test, so many data sets to
in their original formats. Data lakes mine - but a relatively constant amount
store both structured data - as of objective truth 2. Determining useful
contained in relational databases and signal requires targeted strategies and
spreadsheets - and unstructured data appropriate technologies, or the sheer
- such as social media, email and text volume of data threatens to obscure
documents. It can then be used far insight and value. Our experience has
more rapidly and flexibly. This system provided us with a number of insights
allows users to run ad hoc queries, as to how Big Data might benefit the
perform cross-source navigation, and risk management sector, and how we
make analytical decisions, all based can get closer to the objective truth
on real-time information. beneath the noise.
2. Silver, Nate. 2013. The Signal and the Noise: The Art and Science of Prediction. Penguin.
4
3
Applying Big Data to Risk Management
Big data applied: Its a game changer
Credit Risk
Market Risk
Operational Risk
Compliance Risk
5
Applying Big Data to Risk Management
Big data applied: Its a game changer
Post-crisis, financial institutions are expensive purchase, a gambling clients financial reports to discover
now expected to have thorough problem. loan-servicing problems, firms
knowledge of their clients. can utilise Big Data technologies
Increasingly, forward-thinking banks Drawn from data lakes, this to detect early warning signals
- for example - will harness Big Data information can augment traditional by observing clients on-going
to develop more robust predictive data sources including financial, behaviours, and act in time.
indicators in the credit risk domain. socio-demographic, internal
New data sources - including social payments and externals loss data.
media and marketing databases - Together, the datasets can produce
can be used to gain greater a highly robust, comprehensive risk
visibility into customer behaviour. indicator.
These sources can reveal startling
information: a costly divorce, an Rather than waiting to review loan
Following several recent scandals, manual sifting through reports. For analytics can generate real-time
the industry has zero tolerance instance, we recently observed the actionable insights, s t o p p i n g
for money laundering/terrorism case of a large financial institution p o t e n t i a l m o n e y laundering
financing. Fines have been huge, where CBI in payment instructions in its tracks, whilst still allowing
recently reaching a record $8.9bn could signify the Central Bank fund transfers for crucial economic
for one French bank found to be of Ireland, Italy or Iran. All CBI and human aid to troubled regions.
working with countries subject to transactions within this firm were Big data technologies can identify
US sanctions. routed to: AML Operations staff to incidents, help draw a wider
be manually reviewed: a resource- picture, and allow a bank to raise
The high cost of money laundering consuming, potentially error-prone the alarm before its too late.
cases has prompted banks to seek task.
new ways to address the severe
limitations in current anti-money Big Data analytics can improve
laundering risk management. the existing processes in AML
operations. Its approaches allow
Traditional approaches to anti- for the advanced statistical analysis
money laundering remain of structured data, and advanced
dependent on rule-based, visualisation and statistical text
descriptive analytics to process mining of unstructured data.
structured data. This system clearly These approaches can provide a
has limitations - without automated means to quickly draw out hidden
algorithms, detecting information links between transactions and
within the wealth of data requires accounts, and uncover suspicious
laborious keyword searches and transaction patterns. Advanced
6
3
Applying Big Data to Risk Management
Big data applied: Its a game changer
Credit counterparty risk quantification Monte Carlo scenarios. However, memory implementation brings a
has become considerably more to fully simulate potential exposure of key advantages over traditional
complex. Derivatives are no longer for all path-dependent derivatives approaches: no compromises such
simply the net discounted value of each as structured products, banks may as approximations are required any
leg - rather, the banks own credit quality need to run around 100,000 Monte longer, and incremental what-if statistics
(debt valuation adjustment - DVA), that Carlo scenarios. Traditional grid-farm are possible in seconds. Overall,
of its counterparty (credit valuation technologies simply cannot cope with the solutions can produce greatly
adjustment - CVA) and unsecured such large processes at high enough improved balance sheet optimisation
funding (Funding Valuation Adjustments speeds - blades on a grid often simply and collateral management, including
- FVA) need to be taken into account. fail due to processing overload. real-time exposure simulation for new
These calculations often have to be client trades and market price/volatility
run in sets with differing pricing data (for In-memory graphics processing units changes.
example producing risk-neutral CVA) (GPU) can offer enormous benefits in
and at various frequencies for reporting the increasingly data-heavy market risk In market risk and counterparty credit
- monthly, weekly, daily, intraday. domain. risk management, volume and velocity
are driving factors. Put simply, banks
Calculating these components is Adapted from the gaming industry, with increased Monte Carlo abilities
enormously data intensive. To calculate in-memory/GPU technologies allow will be able to price path dependent
CVA at the portfolio level, large banks systems to handle large amounts of derivatives trades at better levels than
typically run between 1,000 and 5,000 data at incredibly high speeds. In- their competitors.
2
7
Applying Big Data to Risk Management
Big data applied: Its a game changer
Fraud costs banks billions. High- Operational Risk team approach detect fraud before the damage
profile rogue trading scandals - to catching rogue traders is has hit disastrous levels.
such as those caused by Kweku to manually track Dealer and
Adaboli and Jerome Kerviel - Operations staff actions that
have dominated press coverage could manipulate positions,
in recent years, and resulted in and profit and loss. However,
staggering fines. These cases fraudsters now utilise a range
can wreak havoc on the financial of technologies and strategies,
system as a whole - Nick Leesons like the private chat rooms used
fraudulent speculative trading in the recent FX scandal. New
caused the catastrophic collapse data lake technologies have the
of an entire bank. Six banks were benefit of collecting data from
recently fined $4.3bn following any imaginable source, including
the global forex probe, reflecting a not only trading systems, email,
failure to improve their controls in social media and mobile data,
the wake of Libor 3. but also HR systems, door swipe
card activity, and computer
Traditional approaches are access log files. The result is a
proving cumbersome and slow in fully comprehensive, integrated
identifying fraud. The traditional approach to data analysis that can
3. Financial Times. 12 November 2014. Regulators slap $4.3bn fines on six banks in global forex
probe. Available at http://www.ft.com/cms/s/0/aa812316-69be-11e4-9f65-00144feabdc0.html#slide0.
8
3
Applying Big Data to Risk Management
Big data applied: Its a game changer
Past, Future
and Present
While Big Data can help us
improve our reactive and predictive
capabilities, it also presents exciting
new possibilities for addressing risk
in the present. Recent developments
in cognitive technology will enable
Big Data technologies to make
reasoned, informed decisions in
real time, defending markets against
vulnerability to shocks.
9
2
Applying Big Data to Risk Management
Big data: Our Vision
10
3
Applying Big Data to Risk Management
Why Avantage Reply?
11
12
3
CONTACTS
13
Editor disclaimer: The information and views set out in this journal are those of the authors and do not necessarily reflect the official opinion of Avantage
Reply. Avantage Reply does not guarantee the accuracy of the data included in this journal. Neither Avantage Reply nor any person acting on its behalf may
be held responsible for the use which may be made of the information contained therein.
14
Visit Avantage Replys LinkedIn page