Brought to you by
TECHNOLOGY SUPPLEMENT | AUTUMN 2016
THE ACTUARY â€¢ September 2016 www.theactuary.com
The chain attraction Blockchain technology could hold considerable disruptive powers. Gary Nuttall separates the hype from reality
Rise of the robo-investors Paolo Sironi looks at the role of robo-advice in wealth management and how artiﬁcial intelligence is transforming business
Learning by numbers Mark Lee and Matthew Evans provide an introduction to data science for actuaries and consider the opportunities for competitive advantage
Towards machine pricing Pietro Parodi looks at the development of machine learning and how actuaries can utilise its many capabilities
Driving success Rutger van der Wall examines how telematics technology can be used to understand the risks of driverless cars
Editor Richard Purcell email@example.com Supplement project editor Gemma Gregson firstname.lastname@example.org Features editors email@example.com Jeremy Lee, investment, ERM Garry Smith, banking, life (regulation) Gemma Gregson, GI, environment Stephen Hyams, pensions Sheila Harney, life, (pricing, product) reinsurance, health Yves Colomb, GI IFoA editor Alison Jiggins +44 (0)20 7632 2172 firstname.lastname@example.org Internet The Actuary: www.theactuary.com Institute and Faculty of Actuaries: www.actuaries.org.uk
Managing editor Sharon Maguire +44 (0)20 7880 6246 email@example.com Sub-editors Kathryn Manning James Richards Display sales executive Vlad Harmanescu +44 (0)20 7324 2726 firstname.lastname@example.org Senior designer Gene Cornelius Designer James Tuthill Picture editor Akin Falope
Welcome to the ﬁrst issue of Predictions, a supplement and dedicated microsite produced by The Actuary magazine, and sponsored by Willis Towers Watson, exploring the challenges and new opportunities technology presents for the ﬁnancial sector and the actuarial profession. Technology is transforming existing business models and creating brand new ones, and in doing so is changing the way we live our lives in many respects. It is making it easier for customers to access information and services, and helping markets operate more eﬃciently by creating new ways to manage supply and demand. The most successful examples of disruptive technologies are those that give the customer more control and put them at the centre of what the business does. In this issue of Predictions we look at the potential for technology to disrupt the ﬁnancial world, one that has so far seen relatively little transformation through new technology. We look at how robo-advice is already widening access to ﬁnancial advice for many customers, while blockchain could transform the way insurers manage claims, and eventually turn them into risk prevention specialists. We also ask how actuaries can respond to the emergence of new technologies by utilising new data sources, such as telematics, to prepare for the dawn of driverless cars. Meanwhile we look at how new data science techniques can help actuaries maintain a competitive edge. Finally, as an exclusive feature on our microsite, we consider the way digitalisation is inﬂuencing customer-focused strategies in insurance. Willis Towers Watson discuss how there has been a general trend for companies to focus mainly on the customer-facing elements of the end-user experience. They suggest that in fact, more work needs to be done behind the scenes to create a truly customer-centric proposition.
Production executive Rachel Young +44 (0)20 7880 6209 email@example.com Print Southernprint
Published by Redactive Media Group on behalf of the Institute and Faculty of Actuaries (IFoA) The opinions put forward in The Actuary magazine and through its commercial and advertising partners are not necessarily the opinions of the editor, the IFoA or the sponsors. No part of this publication may be reproduced, stored or transmitted in any form, or by any means, without prior written permission of the copyright owners. While every effort is made to ensure the accuracy of the content, the publisher and its contributors accept no responsibility for any material contained herein. © Institute and Faculty of Actuaries, September 2016 All rights reserved ISSN 0960-457X
PREDICTIONS Technology | Autumn 2016
I NTR O D U CTI O N
TI P O F TH E I C EB ER G : D I G ITA LI SATI O N AS A N EN A B LER O F C U STO M ER- C ENTR I C STR ATEG I ES PREDICTIONS Technology | Autumn 2016
s an indicator of how much has been written about the concept of customer centricity in insurance, I did a couple of Google searches before writing this article. I expected high numbers, but even I was somewhat taken aback by the 289,000 results returned for the search ‘customer centricity in insurance’ and the colossal 971,000 for ‘customer centricity and insurance’. The numbers are irrelevant; the point is this is a subject that has been discussed – a lot. Given this level of background noise, it’s hardly surprising to have seen many insurers joining a global shift towards digital and customer-focused strategies. In delivering these strategies, the general trend has been for companies to focus mainly on the customerfacing elements of the end-user experience, such as faster, intuitive websites and user-friendly apps. As much as those things are important, in our experience they represent a small fraction of the eﬀort needed to attain the customer-focused ambitions of many insurers. Most of the hard work has to go on behind the scenes. In this respect, the iceberg analogy, so loved by many consultants, really is appropriate. Likely stumbling blocks are the limits of legacy systems, technology, organisational structure and supporting processes, and having the right talent to execute the transformation and deliver value.
A Insurance customers today, whether direct end-consumers or intermediaries, are becoming increasingly demanding. Their expectations are changing at a significant pace, driven by exponential advances in technology and innovations in other industries that are dramatically raising the bar for what good customer service looks like. Heloise Rossouw examines the implications for insurers PREDICTIONS Technology | Autumn 2016
Connected foundations Over the years, insurers have typically accumulated an assortment of legacy systems that play a signiﬁcant role in their businesses operations. Often, these systems were built when data storage, for example, was unwieldy and expensive, and therefore they don’t store data in a way that supports shifting needs. Moreover, the process ﬂows that may work with these legacy systems don’t necessarily apply in the new world. The resulting assortment of old and new technologies, internal and external data and creaking processes makes it very hard to achieve the levels of IT connectivity that provide the foundation for unlocking the potential of data and technology. Fortunately, software is now available to act as an interface umbrella for the compendium of systems and technology, helping insurers to integrate their data and provide a seamless integrated customer interface across all communication channels. This connectivity is particularly important when considering the three main improvements that a connected, digital insurer typically wants to deliver: analytics; distribution and customer service; and
eﬃciency and expense management. To date, many insurers have tended or have been forced to concentrate on only one, whereas if they have the required level of connectivity, they can quite easily do it all.
Analytics arms race Enhancing analytics capability has often been the ﬁrst port of call for insurers, led by property and casualty (P&C) companies. This isn’t that surprising when the rapid development of big data and associated technologies, coupled with an explosion in potential data availability and more powerful predictive analytics tools, are already changing the face of competition in much of the insurance market. Consequently, the increased use of analytics and swift deployment of insights at both the case and portfolio levels are keys to success for a growing range of insurance businesses. Indeed, we are among a growing band of insurance insiders to refer to the situation as an ‘analytical arms race’. But the analytics on their own aren’t enough. Companies need to be able to deploy them into the business and get the right people seeing the right business intelligence and decision support materials at the point where they are making a decision.
Distribution and customer service Technology needs are closely allied to distribution and multi-channel customer service. That’s because, across industries, the balance of power at the point-of-sale is shifting towards customers. Largely thanks to technology, many expect to be able to obtain the products and services they need or want, when and how they need or want them. This applies not only to end-customers, but also increasingly to the intermediaries that represent them in personal and commercial markets. Back-oﬃce systems need to be up to the task of packaging information for the customer-facing technologies. Furthermore, expectations of customer experience and service are being shaped by the likes of online retailers, who have taken attributes such as ease of use, algorithms that can pre-empt customer needs, fast or realtime service, and individual customer recognition to new levels. Distribution and customer service nirvana for insurers is therefore increasingly likely to involve an omni-channel approach to customer connectivity, allowing customers to interact with insurers in a way that suits them. Recent examples of insurance propositions that reﬂect the changing approach include Ageas’ ‘Back Me Up’ and
pricing customer behaviour
distribution and customer service
customer connectivity omni-channel
improved efficiency and expense management
costs/expense ratio automation de-duplication of tasks
Discovery’s Vitality initiative.
Driving eﬃciency A further incentive for insurers to focus on the elements below the waterline is to improve expense management and operational eﬃciency. Relatively simple but often surprisingly hard-to-achieve goals, such as streamlining processes so that they’re not constrained by legacy systems, increased automation and removing duplication such as data re-keying, are often a vital part of success in this area. Crucially though, delivering eﬃciencies doesn’t mean framing every aspect of serving customers in the changing business environment in terms of how to cut cost. The point is that customer centricity doesn’t just hinge on eradicating ineﬃcient processes. Instead it needs to involve looking at the value and cost chains of an omni-channel business model in a completely new light. Understandably, this can seem rather daunting, but it’s deﬁnitely achievable.
The art of the possible Insurers need to recognise that, while they may not be Fintech 100 organisations, they have a lot working in their favour. For a start, most have capital with which to turn ideas into reality. Second, they have customers – and normally lots of them – providing plenty of data with which to work and develop new analytics insights. Crucially, they also have people with deep industry knowledge and relevant experience – and that includes actuaries and their analytical skills. Where adjustments may need to be made is in how teams work together. The days of actuarial, underwriting, claims, marketing and actuarial departments working in silos certainly won’t be compatible with the needs of a more customer-focused business model and improved customer experience. Actuaries, for example, are likely to need to communicate more broadly their progress in areas such as price
Figure 1 Drivers of digitalisation in insurance
optimisation, so that they can be used more widely across the business. The challenge for many insurers will be to organise the capabilities they already have diﬀerently to meet more customer-focused and cross-functional goals. Having done that, they will also need to be prepared to experiment and try new things. A useful approach may be to identify opportunities that can deliver pockets of value to the business, but with a ‘if you’re going to fail – fail fast’ mentality. This requires a culture of conﬁdence, where the management can act at pace – a situation facilitated by accurate real-time analytics and rapid monitoring.
Ready for take-oﬀ ? There are all sorts of legitimate reasons why higher levels of customer centricity have been rather slow to take oﬀ in the insurance sector, such as the innate complexity of the insurance contract and onerous regulation. But until insurers fully address the back-oﬃce implications and the delivery of real-time analytics, the beneﬁts of more customer-focused business models are likely to fall short of expectations, no matter how great the front-end proposition looks. Initiatives already taking place across large swathes of the industry can provide a ready-made vehicle to assist progress. Handling and manipulating data, applying new technologies and using analytics to improve performance – all of these represent familiar territory for insurers. Insurance companies and the actuaries working within them should be able to use their experience to rise to the challenge that developing more customer-centric business models poses. And, as these technologies become enterprisewide, actuaries will need to assume a wider role in driving the adoption of analytics tools. HELOISE ROSSOUW FIA IS A SENIOR CONSULTANT IN THE INSURANCE MANAGEMENT CONSULTANCY AT WILLIS TOWERS WATSON
PREDICTIONS Technology | Autumn 2016
H C AIN A TR CTI O he global banking sector is estimated to have invested over $1bn in blockchain technology in the last 12-18 months. In 2015 alone, it is estimated that bitcoin and blockchain start-ups raised nearly $500m in funding. In January this year, the UK government’s chief scientiﬁc oﬃcer published a report that suggested the technology would be gamechanging and would provide great opportunities for companies to develop solutions. The report also suggested a variety of potential uses for the technology, including the delivery of government services. ‘Blockchain technology’ is a term that is often used interchangeably with ‘mutual distributed ledger’, although, arguably, this is incorrect: ● A ledger is a record of transactions ● A mutual ledger is a ledger to which all participants have access ● A mutual distributed ledger is, as the name suggests, a mutual ledger that has been distributed, such that everybody involved has a complete, synchronised copy ● When the data is stored as a series of sequential blocks, each of which is cryptographically chained to the previous block, this is called a ‘blockchain’. To keep it simple, the term blockchain can be used to describe a mutual distributed ledger database that everyone has an identical copy of, and that is cryptographically secured and time-stamped.
T Blockchain technology is viewed with interest by governments and financial markets. Some believe it could hold considerable disruptive powers. Gary Nuttall provides an overview and separates the hype from the reality
Why the excitement? Blockchain is sometimes described as being ‘the next internet’ and could be a highly disruptive technology, due to a number of key features: Transfer of value without a trusted intermediary If you wish to participate in a ﬁnancial transaction, you need to use a trusted third party, such as a bank or credit card provider (unless you’re paying in cash).
PREDICTIONS Technology | Autumn 2016
With blockchain you can simply reassign the ownership of value directly, since everything is held in a single ledger. This could dramatically impact industries that have been created due to lack of trust between counterparties, for example, conveyancing, banking and escrow agents. The digital currency, bitcoin, is an example of the use of blockchain for value transfer. Smart contracts This is perhaps a poor choice of words since so-called ‘smart contracts’ are neither smart, nor contracts in the legal sense. A smart contract is self-executing computer code that runs on the blockchain. Smart contracts allow the ability to link external events identiﬁed through an external data supply such as weather reports, ﬂight delays and shipping manifests. To be able to code that automatically presents signiﬁcant opportunities to develop new products and services. The DAO, which stands for decentralised autonomous organisation, is a well-known example of using smart contracts to create an entire, self-managing eco-system. It raised in excess of $150m through crowdfunding. However, it may turn out to be a classic example of adopting technology before suﬃcient maturity has been achieved; as over $50m was quickly siphoned oﬀ by a hacker who exploited a programming error.
Immutability As each block of data is added, an algorithm, known as a cryptographic hash, is run. This produces a unique value based on the content of the block of data. This hash is then stored at the beginning of the next block and therefore forms part of the calculation of the next block’s hash. The end result is that data held in prior committed blocks cannot be amended since this would then aﬀect the hash value of the block. This would then require the algorithm to be recalculated for each subsequent block. As such, this makes it impossible to amend data once it has been written to the ledger. Cyber-resistant Every participant ‘node’ in a blockchain distribution holds a complete copy of the synchronised ledger. This means that even if multiple nodes are successfully attacked then the blockchain is still able to operate. This removes the risk of distributed denial of service (DDoS) attacks and of data being subject to ‘ransomware’ attacks. Crypto-secure Using a combination of public and private cryptographic keys, it is possible to tightly manage and control access, such that participants can only see the data that is relevant to them. Synchronised mutual ledger Many organisations spend signiﬁcant amounts of money on the reconciliation processes needed to check that data copied from one organisation’s ledger to another is consistent. If there is just one ledger, there is no data being transferred and therefore no need to reconcile. Likewise, auditing overheads are reduced since there are fewer processes to audit.
Applications Blockchain has a wide range of potential applications including identity management and asset registration. For insurance, blockchain provides opportunities in two main realms. First, it allows insurers to develop new products. Linking it with other emerging
“Blockchain is sometimes described as being ‘the next internet’ and could be a highly disruptive technology” IMAGE: VINCE FRASER
technologies such as ‘internet of things’ sensors and artiﬁcial intelligence, the potential exists to develop new parametric products that can link the automatic detection of an event to the automatic payment of a claim. In the longer-term, linking these technologies together will allow insurers to become ensurers, whereby they will be able to identify that an event is going to happen and execute a preventative action. This is akin to sending an engineer to repair a faulty bearing on a machine before it fails. Second, it provides signiﬁcant savings in operational eﬃciency by decreasing the time, cost and eﬀort expended on reconciling multiple ledgers. This will reduce the burden of audit by reducing the number of processes that require auditing, and by providing more reliable, provably immutable ledgers.
Hype or reality? The concept of ledgers, mutual ledgers, distributed databases and linking blocks of data isn’t new. Indeed, back in 2008 a white paper on the bitcoin protocol was published by an anonymous author known only as Satoshi Nakamoto. Bitcoin has seen good and bad periods; it has been linked in the past with ‘dark web’ markets, such as Silk Road, which was shut down by the FBI because of the illegal trade being facilitated therein. However, the fact that bitcoin is still around today shows how robust the protocol is. Blockchain uses open source code, and with a market cap (at the time of writing) of over $10bn, it would normally present a lucrative target for hackers. To date, although other exchanges have been subject to fraud, the bitcoin blockchain itself has remained unhacked and blockchain-based technology is now well-established. As with any emerging development, the technology presents its own challenges. In particular, performance, scalability and power consumption are known issues that are being worked on. The original bitcoin blockchain processes transactions at around 10 per minute, which makes it unsuitable for real-time ﬁnancial transactions. However, new blockchains are being developed that enable throughput of over one billion transactions per day. Currently, and this is an area moving at a great pace, there are few examples of commercial implementations at an industrial scale. However, a number of companies will soon be moving from research and development into production. Far from being just hype, therefore, blockchain is about to become a reality. GARY NUTTALL IS MANAGING DIRECTOR OF DISTLYTICS AND PROVIDES TECHNOLOGY CONSULTANCY FOR THE LONDON COMMERCIAL INSURANCE MARKET.
PREDICTIONS Technology | Autumn 2016
RI E O T O B - I N V ST R he last two years have seen a period of profound business transformation. The human race is witnessing the rise of a robotic workforce, artiﬁcial intelligence and the internet of things. Moreover, big data analytics oﬀers deep business insights, while cognitive computing is evolving rapidly. Even blockchain cryptography, still in its infancy, is already under threat from quantum computing. Financial innovation, technology advances and changes in individuals’ behaviour all enforce industry change. But, most of all, ﬁnancial regulation has aﬀected the competitive landscape, and made such a transformation not only possible, but a business imperative. Regulation is the main engine for transformation, and due to the loss of reputation suﬀered by traditional players during the global ﬁnancial crisis, ﬁduciary standards have been stepped up across the globe. Transparency and suitability principles have also been strengthened in order to realign the incentives of ﬁnancial institutions with the ultimate interests of the taxable investor. Technology innovation has allowed ﬁntechs (ﬁnancial technology start-up companies) to ﬁnd solutions that are ﬁt for purpose and can lower the ever-rising compliance burden. Digital business models can replace established customers’ networks and enable ﬁntechs to reach out to investors any time, any place, with compelling, convenient and engaging investment experiences. There is no doubt banking is going digital – because today’s world is already digital. However, technology cannot be the only answer as work is required on both sides of the equation: FINance and TECHnology. Financial innovation is needed to help investors and move them outside the traditional investment pattern of buying high and selling low. This is very often based on a myopic trading strategy of buy and sell decisions that is overly reactive to news events, rather than listening for long-term trends. Robo-advisors – replacing face-to-face investment and savings advice with automated online guidance – possess three appealing features. First, they invite individuals to invest into portfolios supposedly geared toward longer-term performance; second,
PREDICTIONS Technology | Autumn 2016
Paolo Sironi looks at the role of robo-advice in wealth management and how artificial intelligence is transforming business
they oﬀer cheaper investment products, such as exchange-traded funds; and ﬁnally, passive investment strategies are preferred, in order to simplify reporting and reduce management costs. This technology has several aims: to insulate investors’ decisions (let’s be clear, not investment performance) from short-term market swings; shade investors from idiosyncratic risks; grant them longer term performance by saving on potentially excessive management fees; and, ﬁnally compound the proceeds alongside the beneﬁts of tax loss harvesting. Wealth management institutions have already started to adopt these new business models, due to depleted balance sheets, rising compliance costs and highly unbalanced cost-to-income ratios. New ventures compete in this space, as well as established players that provide automated investment solutions alongside more traditional businesses. Yet is this enough to make ﬁnancial wealth management more sound and transparent? Quite a few elements of concern remain and need to be addressed by regulators and industry participants. First, most robo-advisors have adopted rudimentary goal-based investment principles: they on-board clients by inviting them to invest according to personalised needs, such as retirement or school education. Figure 1 shows a timeline example of life events, wealth accumulation goals and wealth consumption requirements. Although notable, this is more thematics than goals as individuals’ assets and liabilities do not truly enter into the investment equation. Hence, only the evolution into more holistic approaches would allow for eﬀective personalisation of the investment experience. This limitation could be solved by techniques such as probabilistic scenario optimisation. This method avoids the limitations of modern portfolio theory and instead oﬀers long-term portfolio construction and simulation based on real products. Such techniques can be capable of encompassing ﬁxed income and derivatives and be adequate to resolve problems linked to cumulation and de-cumulation patterns such as pre-, at- and post-retirement decision-making. A further challenge comes from on-boarding
questionnaires that are way too rudimentary and do not allow digital advisors and their ﬁnal investors to engage in enriching and empowering conversations. In this regard, gamiﬁcation – the application of game design and principles – is emerging as a new digital force in the wealth management ecosystem. It allows for experiences to be created that help advisors and individuals make better decisions when confronted with ﬁnancial news and market swings. In this respect, goal-based investment can provide gamiﬁcation with the consistent mindset to gamify investments by simulating personal goals, market scenarios, and life events within a digital playground to enforce more adequate investment behaviour. However, this development will not be free from cost. It requires a change of perspective: from a traditional asset management view of market variable optimisation, to a more personalised investment modality of eliciting investors’ ambitions and fears over time. In such a transforming environment, investors’ fears and long-term aspirations would take centre stage, and goal-based investment principles will allow ﬁnancial advice and ﬁnancial planning to converge. The investment oﬀering could then be customised around clients’ ultimate goals, which would in turn generate premium services, driving proﬁtability and sustaining innovation. Financial innovation is paramount at a time when ﬁnancial advice and ﬁnancial planning are converging. While ﬁnancial advice is undergoing a process of strong commoditisation, ﬁnancial planning still grants more space to leverage technology and add value to the human relationship between planners and investors. All aspects of personal wealth today are linked to ﬁnancial markets, meaning that long-term insurance and investment planning must adopt many aspects of ﬁnancial advisory solutions. The need for better retirement advice and planning seems to be gaining momentum: the largest cohort of baby boomers is about to retire, life expectancy has never been longer, and the workforce is shrinking at a time when modern economies struggle to generate enough growth to keep the pension challenge under control. A mix of ﬁnancial and digital innovation is therefore required to facilitate the giving of roboadvice on a massive scale, so that individuals better understand the implications of investing and saving for distant goals. This is the reason why goal-based investment robo-winners will be well-positioned to outpace laggards, whether they be ﬁntechs or digital incumbents.
Figure 1 An example timeline for life events
caring years de-cumulate
wea d e - c u m lt h u la t io n
th weal th w gro
save and spend
ial nc ges a n fi llen a ch
buy house first income student loan college
IMAGE: VINCE FRASER
wedding children retirement behaviour
PAOLO SIRONI IS A THOUGHT LEADER AT IBM, AND THE AUTHOR OF FINTECH INNOVATION: FROM ROBO ADVISORS TO GOAL BASED INVESTING AND GAMIFICATION, PUBLISHED BY WILEY (2016)
PREDICTIONS Technology | Autumn 2016
DATA S C I E N C E
L A IN YN B RS ace recognition, fraud detection and spam ﬁlters are just a few examples of the applications of data science, a catch-all term encompassing big data, machine learning, data mining and predictive analytics. An ever-increasing supply of data, and powerful modern computers that are able to exploit and analyse it, has led to the growth of the data science ﬁeld. At its core is the concept of gaining insight from data, be it big or small. Data science techniques are employed in a wide variety of industries, from fashion retail to hedge funds. We use the term ‘big data’ to refer to large collections of data, potentially from diverse sources, that is often unstructured, relying on text, pictures, or geographical positions, rather than ﬁxed ﬁelds as found in more traditional data sets. But data science is not just about big data. Having big data may well require the use of machine learning technology to extract useful information; however, a lack of big data does not preclude the use of machine learning algorithms.
F Mark Lee and Matthew Evans provide an introduction to data science for actuaries and consider how it can provide insurance companies with a competitive advantage
PREDICTIONS Technology | Autumn 2016
Machine learning ‘Machine learning’ is the process by which a computer learns by being exposed to data, generally by using an algorithm that optimises some mathematical function of that data. Once the domain of computer scientists in large research organisations, machine learning is now available to everyone through free, open-source toolboxes provided for programming languages such as R and Python. These languages have a comparatively easy learning curve and come with many functions that are built in or available to download, enabling the user to perform sophisticated tasks with ease. This functionality is invaluable to actuaries as it means the exercise is more one of data manipulation and analysis of the output than computer programming. With courses available that provide the fundamentals needed to explore the ﬁeld, data science has never been so accessible to actuaries.
There are two fundamental categories of machine learning. ‘Supervised’ learning algorithms are in the business of prediction, while ‘unsupervised’ learning focuses on understanding the structure behind a data set. Imagine trying to categorise pictures of cats and dogs. Starting from a database of such photos, each labelled either ‘cat’ or ‘dog’, supervised learning involves the creation of a predictive model that exploits the information contained in the labels. The model will make predictions by taking an unlabelled, previously unseen pet photo and deciding whether it is a picture of a cat or a dog. Unsupervised learning takes a diﬀerent approach. Running an unsupervised algorithm on a set of unlabelled photos returns a grouping of photos that are most similar. That grouping might be a separation into pictures of cats and dogs, but equally could be a separation by pet size or colour. The exact results will be determined by the parameters governing the algorithm. Although the unsupervised learning algorithm may not be able to predict pet species, this is not a failing of the algorithm since it was not supplied with the information contained in the labels. A successful unsupervised algorithm will provide information about the relationships between the pictures. It is then up to the user to interpret the information appropriately – after all, pet species is not the only information in the pictures, and for some uses, maybe pet size is more important.
Actuarial applications Figure 1 shows some machine-learning algorithms. One such algorithm, a generalised linear model (GLM), has been used by actuaries in personal lines pricing for years. GLMs can be thought of as prototypical supervised learning algorithms. Given a set of prior claim frequencies and severities, a GLM algorithm creates a model that predicts, for a new policy, how likely it is that a claim will occur and how much it will cost. With modern computing power, these methods can be taken further with the use of algorithms, such as decision forests or neural networks. The ﬂexibility of these algorithms allows the ﬁtting of non-linear trends, without having to make manual assumptions. Such techniques also have the ability to identify interactions between data items that are not seen by the human eye or through the use of linear models. These ‘hidden’ interactions can then potentially be used to predict claims more eﬀectively, leading to more competitive pricing. While GLMs are often used to price personal lines, specialty lines in the London Market rely on the
expertise of underwriters. Marine pricing is one such example where there is a wealth of data, in this case on ship position and weather records. This is big data, which lacks clear structure and so can be diﬃcult to analyse. However, supervised learning algorithms such as neural networks could extract features predictive of claim patterns. The information could add an extra dimension for underwriters and may oﬀer a competitive advantage. There are also many potential applications for unsupervised learning techniques. Unsupervised algorithms can augment and replace human-labour intensive data sorting and visualisation, particularly when the number of data ﬁelds is large. For example, grouping accounts by prior loss ratio performance, enabling quick identiﬁcation of common trends or dependencies, may oﬀer management teams a valuable insight into the company.
Beyond the black box
Figure 1 Some common machine-learning algorithms
Supervised learning Generalised linear models Neural networks (MLP and CNN) Decision trees Decision forests Gradient-boosted machines
Unsupervised learning K-means clustering Hierarchical clustering Principal components analysis Mixture models Neural networks (SOM and ART)
Decision forests Neural networks
Flexible ‘black box’
Figure 2 Trade-off between flexibility of algorithm and transparency of resulting model
A common criticism levelled at machine learning is that the resulting models are too ‘black box’ like. Although a simple linear model is straightforward to understand and communicate, it is not very ﬂexible when dealing with general data that may involve non-linear relationships. In contrast, very ﬂexible supervised learning algorithms, such as decision forests or neural networks, can ﬁt to quite general data patterns but at the cost of a less transparent model (see ﬁgure 2). While it might be true that the models can be complicated, the resulting model can usually be communicated suﬃciently clearly by plotting the model predictions against the various predictive features. Furthermore, a variety of statistical techniques exist that can provide the user with the comfort that the model is robust and appropriate. Data science is not a new ﬁeld. It contains a multitude of tried and tested algorithms that have already been proven to be beneﬁcial in other industries. With the development of technology giving the everyday user the computing power to use these processes, and with the tools to use these methods being easily accessible, actuaries can now apply techniques that were once only available to data specialists. In today’s competitive environment, data science could be used to supplement the tools that actuaries already have at their disposal and provide companies with that all-important edge. MARK LEE IS A CONSULTANT AT INSIGHT RISK CONSULTING MATTHEW EVANS IS AN ACTUARIAL DIRECTOR AT INSIGHT RISK CONSULTING
IMAGE: VINCE FRASER
PREDICTIONS Technology | Autumn 2016
O A PREDICTIONS Technology | Autumn 2016
RD RICI G IMAGE: VINCE FRASER
B Pietro Parodi looks at the development of machine learning, and the impact on pricing
Machine learning as a theory of modelling The main contention of that research project – and one that has not dated – was that machine learning is not just another powerful technique that actuaries should learn. Rather, it is the only rigorous theory available on how to build models with predictive powers, whether in data-rich situations (personal lines pricing) or in sparse-data situations (London Market). The problem of data-driven risk costing (the basis of much pricing, reserving and capital modelling) is an example of supervised learning – the problem of learning the features of a model based on a sample of inputs (for example, rating factors, or the
Figure 1: A model should be as complex as neccessary – but not more complex
ack in 2009, I completed a research project for the actuarial profession on the applications of machine learning and more generally of artiﬁcial intelligence (AI) to general insurance. A brief summary of that work appeared in The Actuary in March 2011 under the title From artiﬁcial ﬁsh to underwriters. Since then, I am happy to report that the ‘actuary of the future’ piece in this magazine still features human beings and not androids. Everything else that could have happened in relation to automation, however, has started happening: a) Machine learning has become a household name among actuaries (and almost everyone else), and techniques such as the lasso or elastic net regression are not esoteric names anymore b) The notion of big data has come to the fore and its use is the main reason why machine learning has become much more eﬃcient, for example in speech recognition c) The use of new business models based on digital technology and big data (‘InsurTech’) promises to disrupt the insurance industry d) Deep learning (a machine learning technique based on many-layered artiﬁcial neural networks) has achieved superhuman ability in a variety of domain-speciﬁc tasks, from face recognition to the identiﬁcation of tumours from radiological images, and is now regularly applied to insurance problems such as fraud recognition. It really looks like, after so many ‘AI winters’ – those periods in history where funding for AI dried up in the wake of crushed expectations – we are going to have a spring that none of us can aﬀord to dismiss.
Validation set error Training set error
parameters in a severity model) and outputs (for example claims amount), with the objective of minimising the expected prediction error. Therefore, actuaries should learn machine learning not only to be hip and to be well-equipped for the onslaught of big data – but because it brings clarity of thought and the right attitude to how they go about their daily job (for example, building and calibrating pricing models and deciding when to use benchmarks). Machine learning will give you mechanisms on how to optimise the complexity of your models, resisting the push towards more complicated and supposedly more ‘realistic’ models (see Figure 1, describing the famous bias-variance tradeoﬀ ).
A pathway to automation As intellectually satisfying as this is, the prize for the adopters of machine learning and artiﬁcial intelligence in pricing is not purely theoretical – insurers’ CEOs are not (always) interested in theories of modelling. Also, while AI may provide valuable tools for fraud detection and data mining, the competitive advantage of using slightly more accurate costing is likely to be limited. The big prize is that machine learning provides a pathway towards pricing (and reserving, and capital modelling) automation. The simplest and best-known example is possibly that of rating factors selection. This has never been anything but a machine learning problem. The industry standard – generalised linear modelling augmented with a mechanism to select the right
PREDICTIONS Technology | Autumn 2016
factors – is in itself a well-known supervised learning technique. A low-hanging fruit for machine learning – well underway – is enhancing the existing industry standard with techniques such as lasso regression and cross-validation as a means to select the model with a minimum expected prediction error in a fully automated and eﬃcient fashion. So much is available oﬀ the shelf (elastic net, kernel methods, support vector machines…) to keep us busy for years. It may well be that the distinction between all these methodologies will soon become as tedious as the distinction between diﬀerent goodness-of-ﬁt metrics. A less obvious candidate for automation and machine learning applications is individual contract pricing in commercial lines (or treaty reinsurance). The standard process for this is a patchwork of tasks that are completely algorithmic (for example a Monte Carlo simulation to produce an aggregate loss distribution) and tasks where judgment is required (data checking and preparation, picking suitable frequency/severity models). A possible pathway towards automation is to re-engineer the process so those areas that require judgment are isolated, and a clear protocol to deal with these areas using available AI techniques is developed. A couple of examples: ● Data exploration and preparation can beneﬁt from the use of rule-based systems (rudimentary decision systems based on simple ﬁxed rules), natural language processing (an umbrella term for various statistical machine learning algorithms aimed at extracting information from text) and data mining. ● AI provides the natural conceptual framework for automating the selection of frequency/severity models and deciding when to resort to portfolio/market data. Where data is scarce, model selection cannot be purely data-driven, but the selection must also be informed by theoretical results (for example, using extreme value theory for large losses). Of course, full automation would not happen in one go, but in an iterative and piecemeal fashion, as is the case for driverless cars.
The advantages of machine pricing The advantages of pricing automation would be similar to automation in other ﬁelds, but with some speciﬁc twists. 1. Machines increase the number of actuarial investigations that can be performed. Since they don’t get tired, they can price as many deals as we
PREDICTIONS Technology | Autumn 2016
want, to the desired level of detail, without needing to prioritise important work and they don’t become intractable if they receive updated information one day before the deadline. 2. Machines can improve portfolio management greatly. Background bots (pieces of software that perform tasks on behalf of a user) can maintain the claims database and the portfolio data, update pricing models and portfolio benchmarks (including exposure curves) continually. They can ensure that the benchmark curves maintain their relevance, rather than sticking to the same exposure curves as people used in the 60s because it would be too onerous to embark on regular reviews. They can ensure that the optimal number of diﬀerent benchmark curves is used, as more claims experience accumulates and it becomes possible to diﬀerentiate risks more and more. 3. Machines would be able to price contracts neutrally, without a cognitive bias. Neutrality is important – human underwriters and pricing actuaries may be able to incorporate special knowledge and wisdom in speciﬁc transactions but they will not be able to guarantee unbiasedness at a portfolio level. A pricing machine may price incorrectly but will be even-handed – and its neutrality at portfolio level can also be checked and monitored by actual vs expected analysis. The underwriters or other oﬃcers will still have the opportunity to override the machine price but this will be documented and the portfolio eﬀects of underwriting adjustments can be isolated and monitored.
Yes, but what about actuarial judgment? Automation makes everything more eﬃcient where it can be applied, but surely we still need sound actuarial judgment. Or do we? It can be argued that at the basis of judgment are experience and the knowledge of the answer to many similar cases looked at in the past, that type of ‘hunch’ that immediately tells you that a particular price or parameter is a bit oﬀ. A few things can be said about this type of judgment: a) If deﬁned as above, judgment is ominously similar to deep learning – you train an artiﬁcial neural network on a number of relevant cases and this comes up with a strategy, which can’t be articulated but gives good results;
“Specific tasks will increasingly be automated – as has been the case for decades – but this will redeﬁne actuarial jobs rather than wipe them out” b) This type of supposedly exclusively human judgment has been invoked several times in history – most famously to explain why chess software could not possibly beat the very best humans at the game, because humans would have an intuition about positions while a machine could only look ahead a number of moves. Over and over again, however, machines have proved themselves to be better at judgment in domain-speciﬁc contexts; c) This judgment is not always correct, especially where past experience is not that large – that will be true for both human and AI judgment.
So is the end nigh for actuaries? If AI is so great that it may eventually replace judgment, will our jobs still be there in 20 years (the canonical timeframe for safe prediction)? Some of the recent anxiety about jobs can be traced back to the oft-quoted paper by Frey and Osborne (2013), The future of employment: how susceptible are jobs to computerisation?, which estimates the probability of various jobs to be replaced by machines. The paper found that 47% of jobs were at high risk (60% or more) of computerisation, and put that probability for insurance underwriters at a staggering 99%. Although this datum can be put to good use for actuary vs underwriter bantering, it probably says more about the limitations of the paper’s methodology and assumptions than it does about the underwriters themselves. Speciﬁcally, the methodology fails to capture the heterogeneity of tasks performed in a speciﬁc occupation, only some of which can be automated. A subsequent study (Arntz et al. (2016), The risk of automation for jobs in OECD countries) has reﬁned the approach and put the percentage of jobs at high risk of computerisation at a more modest 9%. Nothing speciﬁcally was mentioned about underwriters (or any other jobs) but by using their task-based approach it is clear that only a small part of an underwriter’s tasks would be amenable to
automation, while most could not. The situation is not dissimilar for pricing actuaries. Speciﬁc tasks will increasingly be automated – as has been the case for decades – but this will redeﬁne actuarial jobs rather than wipe them out. The ‘lump of labour fallacy’ – that is, the idea that there is a ﬁnite amount of work to go around and automating it will reduce the need for people – applies to actuaries as well as to the labour market in general. The advent of desktop computers, spreadsheets, and programming languages hasn’t reduced the need for actuaries, but has increased dramatically the number of things that actuaries are asked to look at. This trend is likely to continue: AI techniques will make it cheaper to run actuarial investigations and therefore the demand for them is likely to increase. The actuary is in a privileged position to create and harness the technology and piece everything together. All this unless, of course, automation becomes so good that it can do the harnessing, the piecingtogether and even the theory-building that professionals do. This, however, is far-fetched. Replacing actuaries altogether is an AI-complete problem – that is, a problem equivalent to creating general intelligence. Despite some concerns that a hostile AI may soon take over the world, we’re not remotely close to the creation of a general intelligence that has consciousness and willpower and the pathway towards it is unclear. I may not have the best track record for making predictions on artiﬁcial intelligence, though. When I was working at the University of Toronto I used to shake my head in disapproval on seeing my fellow post-docs wasting their best years in Geoﬀ Hinton’s artiﬁcial neural networks lab. I, for one, was engaged in much more serious and promising theoretical research on the computational complexity of certain tasks in machine vision, building up on my PhD work. Fast forward a couple of decades: Geoﬀ Hinton is now head of AI at Google. My PhD supervisor has long had enough of this whole machine vision business and has gone back to neuroscience. As for myself… well, I’ve become a pricing actuary, so I can’t complain at all, can I? PIETRO PARODI HAS 10+ YEARS OF EXPERIENCE AS A PRICING ACTUARY AND IS AUTHOR OF PRICING IN GENERAL INSURANCE (CRC PRESS, 2014). HE WILL BE SPEAKING AT THE IFOA’S GIRO CONFERENCE, ON 20-23 SEPTEMBER
PREDICTIONS Technology | Autumn 2016
T E L E M AT I C S
Rutger van der Wall examines how telematics technology can be used to understand the risks of driverless cars
UC E S
hether it’s motorists’ fears or fantasies about driverless cars, barely a day goes by without fresh opinion appearing in the media on the whys and wherefores of autonomous vehicles. This has recently been brought into sharper focus as the UK government has conﬁrmed it is looking to amend the Road Traﬃc Act 1988 motor insurance provisions to extend mandatory motor insurance to include product liability. In a speech to insurers earlier this year, the road transport minister warned that traditional data sources for rating insurance will become obsolete with the emergence of the connected car. If there was ever a sharp elbow prod to the insurance sector to prepare, this was it. That is not to suggest
PREDICTIONS Technology | Autumn 2016
insurers have been slow to the party, far from it. They have been acutely conscious of the liability questions and involved in discussions from the outset. However, there is an urgent need to get down to the nitty-gritty and start appreciating the value that driving behaviour data, such as speeding, braking and road familiarity, will bring in understanding liability and risk and enabling cover to be provided for driverless cars. We know that fully autonomous vehicles are not going to appear overnight – it is more realistic to consider this as a staged process. IHS Automotive, a provider of global market, industry and technical expertise, predicts that almost 76 million vehicles with some level of autonomy will be sold globally between now and 2035. This means we will see
“We have a window of opportunity to use telematics during this testing period to better understand the risks involved”
vehicles with a range of autonomous capabilities on our roads and a hybrid of liabilities for insurers and manufacturers to consider. For example, personal liability will need to be accounted for in the case of humans driving vehicles with limited automation. However, manufacturers involved in testing autonomous vehicle technology have clearly indicated they will carry liability for their vehicles when they are in fully autonomous mode. The challenge for insurers is that they have no historical data to refer to when it comes to these types of risk. So how can insurers calculate risk and develop products that support the use and adoption of driver assistance technology? Ultimately, they need to know how consumers are using this. The answer is in telematics technology. We have a window of opportunity to use telematics during this testing period to gain the necessary insights and better understand the risks involved. By working with the original equipment manufacturers (OEMs), insurers can use the mass of data being collected from
IMAGE: VINCE FRASER
vehicles to understand the eﬀectiveness of autonomous cars and how drivers interact with the capabilities of the vehicle. The data collected through telematics will be invaluable for underwriters and actuaries to determine their claims loss ratios, and also how the risk changes across the diﬀerent levels of autonomy. It will also enable insurers and manufacturers to identify, in the case of an accident, whether the technology installed worked in the way expected, and also the role that human intervention played, if any. These results can then be delivered back to the insurer for them to make a judgment on the fault of the claim and who is responsible. Fundamentally, driving data, regardless of how it is collected – whether hard wired black box, a 12V plug-in device into the cigarette lighter in the car or a driverless car – needs ﬁltering, normalising and enriching to bring value. This is where insurers need to be focusing their time right now, to fully understand the processes and possibilities. Ultimately, this will help them meet the new legal requirements by the time fully autonomous vehicles reach UK roads. In order to accelerate the learnings during this period, the industry needs to consider how it can work together for the beneﬁt of all parties. For example, a data hub could be created, which would enable all insurers and OEMs to share data and learnings. Insurers could then pitch for the data sets they require. This exchange of data would provide actuaries and underwriters with access to a large amount of quality data in order to calculate risks accurately and identify emerging trends. Insurers are already eﬀective at sharing data with central government databases, controlling personal data to eliminate fraud, and using contributory services. So this is a good model to use in the case of driverless cars. Therefore, by working together as an industry, propositions and solutions can be created to improve the success of driverless cars with all the societal beneﬁts this mode of transport will oﬀer. RUTGER VAN DER WALL IS VICE PRESIDENT BUSINESS DEVELOPMENT, AUTO INSURANCE – LEXISNEXIS
PREDICTIONS Technology | Autumn 2016
Latest thinking. Leading software. Unlocking potential. Innovative software solutions backed by insurance industry insight Our rich heritage in insurance risk, analytics and technology gives us a unique perspective. We offer more than individual products and services – we understand your business issues, allowing us to deliver software solutions so that you can unlock your full potential. To ﬁnd out more, please contact firstname.lastname@example.org
Copyright © 2016 Willis Towers Watson. All rights reserved. WTW-EU-16-ADV-3167a. Towers Watson is represented in the UK by Towers Watson Limited.