 |
|
 |  |  |  |
Social intelligence is also changing with the advent of the internet and platforms such as social media, with many acting on a Facebook ‘like’ or Twitter recommendation than an advertisement compiled by experts. The key question amongst all this change is how to digest, interpret and use this exponential increase in information. Predictive analytics, and the actuarial communities’ use of predictive analytics, could hold the answer in digesting and comprehending this increase of information, and could be the solution to how big data has transformed personal lines insurance into a race for survival. Ivan Clarke sat down with Alex Poracchia, Partner at Deloitte, to discuss predictive analytics.
Before going into predictive analytics, Alex, could you provide some thoughts on the market at the moment?
I see the market today as being made of three core players: intermediaries; carriers and clients. Intermediaries can include brokers, but also comparison websites. These companies are significantly investing in tools to capture, analyse and compare carrier data and preferences. It can be used to guide risk appetite, but also clients. It is this extra intelligence that makes them more relevant and competitive to the clients and carriers. Those intermediaries are, in fact, responding to a client need: clarity and relevance. Clients are becoming overwhelmed with choice, options and price. This will only increase as the channels for communication grow, particularly in terms of social media. Just to give you a few examples, every minute, according to Deloitte research, over 170,000 Tweets are posted; 570,000,000 corporate emails are exchanged; and brands and organisations receive in excess of 34,000 ‘likes’ on Facebook. This is a new source of information that can influence how people buy products. Now, from the carriers’ point of view, this is a big data challenge for organisations and something that shouldn’t be overlooked. There are currently two trends in addressing those challenges. The first one relates to “breaking the silo” and, the second one is, “knowing my client”. Breaking the silo is about organisations making the most of their own data by addressing the functional and legacy system barriers, or “the silos”. “Knowing my client” is more about catching up on the intermediaries by developing their own business intelligence, capturing data points on intermediaries and / or clients. They are also trying to put together more direct channels, particularly for SMEs. Differentiation will be critical for carriers and offering new products will be one way to get an edge. It is a very competitive and crowded environment, where everyone is trying to gain a broader position in terms of what to offer. Advantage will be won by those who compete on insight gained from data. So coming back to analytics, what do you mean by predictive analytics? It is the application of advanced mathematics to large or enhanced data sets while using new technologies to increase and improve your forecasting and predictive capabilities. It can be used to solve business problems with better accuracy and in real-time through fact-based decision making. 
Click image for larger view
Most of the advances in predictive analytics to date have come from motor and small commercial lines where do you see these sophisticated analytics techniques being utilised in the future? Analytics will provide new opportunities for businesses to enhance how they sell from business to business; business to business to customer; and, business to customers to customers. We shouldn’t underestimate people talking to one another – what products people recommend to one another is a powerful endorsement – it’s information that can be predicted using analytics techniques. But, in insurance, we’re not there yet. And what are the leading sources of competitive advantage derived through the use of sophisticated analytics? Operational excellence is the first competitive advantage. In the insurance industry, people will be familiar with analytics being applied to identify new risk drivers, for pricing purpose, or claims indicators to identify those that are likely to be fraudulent. Those can then be used to define a new, more effective, course of action. But, this ‘next best action’ doesn’t just apply to risk; it can be relevant for highlighting who might be the next client and how you can tailor offers to that individual. Using analytics to identify adjacent or new clients and/or products is the second and greater competitive advantage. The credit card industry in the 1990s is a good example of how it can work in practice. They looked at how people used their credit cards abroad; they analysed when, where and how people spent their money. By co-branding with a number of airlines, they were able to expand the offers to the next group of customers, therefore broadening their client base. The insurance industry is not there yet. It’s more anecdotal; it’s more in pilot and development stage than ready to go to market. Whilst it’s clear that analytics can improve the profitability of a carriers customer mix what happens when all carriers are employing analytics? First, it’s going to take time to get there. However, when analytics is widely applied and data quality and coverage has reached a consistent stage, I believe that clients will see more individualised and adjusted products, and we will certainly see product and client innovation. Telematics is definitely one of those examples where you can see the value in gaining and understanding client behaviour or risk. In the US, certain carriers have proposed telematics products for young drivers, whereby they can provide a report on the individual’s driving habits, whether speed, g-forces or braking. Parents of drivers can use this information to intervene when there’s been aggressive driving, but equally good drivers can be rewarded. It’s a win-win situation. How do you feel predictive analytics compares to the more traditional use of loss and underwriting data? In a way it’s pretty much the same; it’s just more of it, but with new tools and techniques that can provide information in real-time. The difference between the old method and the new method of predictive analytics is that we can also start without a hypothesis. In previous actuarial methods of pricing, you have to make a few assumptions of cause to effect: for example, is the size of the engine on a car, or the value of the property related to the cost of the risk occurring? With predictive analytics, we have so much more information and techniques that allow us to be far more accurate without traditional constraints. In other words, you can let the data “talk” rather than having to interpret it. How has the use of predictive analytics in the insurance industry grown or developed and what are the key drivers for that growth? We should recognise this is a growing area for the industry, but it’s not a level playing field if you compare it to banking or credit cards. A lot of this is down to systems legacy, so we’re seeing more investment to replace ledgers and getting more out of their data and systems. The drivers will be profit focussed. Analytics allows you to make sense of the explosion of data, so that better, fact-based decisions can be made on common questions and challenges. How do I optimise my processes? Do I want a shared service platform? How do I get the most optimal geographical footprint? All of those rely on the same predictive analytics techniques. And what information can be used in a predictive model? And how can that information be used? To be fair any. There are two types of data: structured or unstructured. Structured would be anything fixed, from the name of your client, size of the car, if that person has a garage, where they live, and so on. But, unstructured data could include anything on social media or sentiment that’s captured about a particular product or company. It might be open data, the availability for which is growing in the UK, thanks to the Government’s data.gov.uk site. Since launching in 2009, it now holds over 9,000 freely available data sets that can be used. Overlaying this information together paints a much clearer picture, unlocking insights not before seen and making it easier to predict future behaviours. Predictive analytics looks beyond actuarial data to personal data as indicators of risk. Does that use of personal data have implications for data privacy and privacy concerns? I would say yes, clearly. Stepping back for a second, we need to recognise that institutions and regulations have been designed quite some time ago, so they don’t necessarily fit in today’s world. However, updates are occurring to protect the digital rights and privacy of the individuals, for example, recent EU cookie legislation that came into effect last year. Transparency will be key and it’s likely we may see more rules and regulations emerging. The most successful companies today have taken a cautious approach to exploring data use, but approach will change from region to region. Take, for instance, individual credit scoring in the US, it’s well established, but it isn’t in the UK despite it being a very good indicator of people’s behaviours and risk profile. So, how and what you can do with people’s data will always need to be considered. The underlying question always remains the same: just because you can use someone’s personal data, does it mean you should?
How is accuracy determined in predictive analytics?
By using the same testing and sampling techniques, familiar to the actuarial community. There are two possible options. The first would be to take a portion of your historical portfolio, test your predictions and then realise (based on that portfolio) what has been the experience versus your prediction. Today the market is much slicker and efficient. Therefore, the second would be to test directly in the market. New pricing structures and new products can be tested on a very small sample of clients or for a short period to gain real time feedback on what works (or doesn’t). For example, you can trial offers on comparison websites or a carrier’s website for a few days or weeks to see how clients react and how this compares with your initial expectations. Time is a factor, too. Older methods could take anywhere between three to six months to implement. The world is much quicker now, real-time. You need to constantly adjust so that sampling is not a one-off – it’s now a technique to monitor how clients will react to offering and what your competition will do. Yes it almost needs to be instantaneous today. In what areas do you feel predictive analytics are most effective? Basically, where inefficiencies are the greatest. However, I see the insurance market investing a lot of effort and research in areas where it can drive differentiation and customer benefits. You see, insurance companies have a huge opportunity to become more customer-centric, and this is something that is at the top of the executive agenda. Traditionally, they’ve been very product and markets (or regulatory regime) driven. It’s a hat-trick of product leadership, service leadership and cost leadership. The winners will be those who can manage more than one of these dimensions at a time. (see graph) How has predictive analytics impacted the personal lines sector? What positive and negative effects have you seen? The initial impact of predictive analytics is of an operational nature, for example, when it comes to managing the risk of fraud. Without it the cost of fraudulent claims to the insurance industry and, therefore, society would be much greater. Now those operational improvements are, for me, tactical changes. I think predictive analytics will contribute significantly to more fundamental changes to come. It will help us see new players and products. Deloitte research has shown that 40% of consumers are willing to buy motor insurance from supermarkets. This indicates a market shift, one that would have a massive impact to the motor insurance or personal lines sector. Any carrier would have to rethink seriously about approaching those clients contracting, retaining or servicing them. One of the interesting things, in terms of gaining efficiencies, is being able to select better risk within groups of demographics that you might otherwise have ignored in total. Absolutely. Some motor insurers still exclude entirely underwriting young drivers (as being perceived loss making). Here, predictive analytics and modelling could help to be more tactical and identify which young drivers should be in fact underwritten. How does predictive analytics assist in pricing risk better then traditional methods? At this stage, there may be a slight difference between personal lines and commercial lines. For personal lines it’s really going to take you to the next level of integration. Data sets are going beyond traditional risk factors. This helps to go beyond estimating the cost of risk (or sometime called technical price) and are now helping to assess and integrate the customer behaviours, the underwriting and negotiation processes. For commercial lines there is so much room for improvement with respect to traditional data sets and the use of traditional risk factors that it’s going to help drive the discipline further like we have seen initially in personal lines. How is predictive analytics being incorporated into the actuarial function? It’s not necessarily the panacea of actuaries; it has so far come through different channels: marketing, technology, claims and underwriting. Actuaries could be more proactive. They are a natural stakeholder to engage as it is about data and making sense from it while using advanced mathematics. The Institute and Faculty of Actuaries has adjusted its syllabus some time ago to recognise the initial steps of predictive modelling like generalised linear modelling methods and then you have an organisation like ours where predictive analytics is an integral part of our actuarial function. We recognise it is going to be a very significant part of the future. What are your predictions on the future use of predictive analytics in the insurance industry and what would the implications of widespread use be? It’s a good question. I believe it will become a standard and if you don’t embrace it now, you will be left behind. Now, predictive analytics will probably go through various phases. The first one, which we are in, is about data enhancement. It is all about seeking competitive advantage through competitive knowledge. The first phase is about answering: how do I enhance my data set? How do I make the most out of it? Who do I partner with? It may be other organisations to get that enhanced insight about my product and my client? The second phase (which in fact can take place simultaneously with the first one) is all about technology and be “real time”, multi-mobile-channels. How do I reach out to my client effectively? Which client should I focus on? How often? Which channel do I use? What products do I offer? After that, we’ll probably see new products and new distributions. We probably have a lot to learn from emerging market, where you can see developments and establishment of simpler, yet far more cost effective solutions. Will we see the opportunity to buy your insurance from a cash machine? How will Smartphone’s affect the future of insurance selling channels? These are evolutionary steps that will truly define the winners and losers if they choose to learn and adapt. So being clear on your strategy and target market first and foremost. Correct. Being clear on the strategy (and what a company wants to achieve) will be a pre-requisite to the successful use of predictive analytics and the corresponding insights.
It’s a really exciting time for the insurance business. This is a data and numbers’ business; ultimately, once you have got all the departments on board, you need to ensure consistency when data sets and formats are being used. That way, you can really focus on what you’re doing, segment and then end up with the best result to retain more of your customers.
I agree with you: it is a very exciting time. There has never been more data and better technology to date. The challenge is to identify which methodology and approach to use and realise opportunities or strategies. However, the learning curve is steep and continuous. So time to embark and become a learning organisation.
Ivan Clarke is European Principal for the Actuarial Search practice at IPS Group
|
|
|
|
| Pensions Data Science Actuary | ||
| Offices UK wide, hybrid working - Negotiable | ||
| Head of Pricing | ||
| London - Negotiable | ||
| Global Specialty Pricing Actuary | ||
| London - £95,000 Per Annum | ||
| Client-facing DC investment manager | ||
| London / hybrid 3 dpw office-based - Negotiable | ||
| Financial Risk Leader - Bermuda | ||
| Bermuda - Negotiable | ||
| Aylesbury Actuaries | ||
| Aylesbury / hybrid 3dpw office-based - Negotiable | ||
| Make an impact in protection pricing ... | ||
| London / hybrid 2 days p/w office-based - Negotiable | ||
| BPA Implementation Manager | ||
| North / hybrid 50/50 - Negotiable | ||
| Head of Reserving | ||
| London - £160,000 Per Annum | ||
| In-force Longevity Actuarial Analyst | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
| Make a difference within reinsurance ... | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
| Be at the cutting-edge of life & heal... | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
| Longevity Pricing Analyst | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
| Develop your career in life reinsuran... | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
| Protection Pricing Actuary - Life Rei... | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
| Life (Re)insurance Pricing Manager (P... | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
| Take the lead: life & health reinsura... | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
| Pricing Tools and Systems Developer | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
| Longevity Pricing Actuary | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
| Shape the future of longevity | ||
| London / hybrid 2 dpw office-based - Negotiable | ||
Be the first to contribute to our definitive actuarial reference forum. Built by actuaries for actuaries.
Copyright © 2024 Actuarial Post. Created by Zero-One Design Ltd
We are currently in the middle of a technological revolution where information is being communicated, captured and shared at an unprecedented speed and size. The big data challenge is far reaching and the implications wide, with information being more freely available means the fight to remain relevant and competitive signals significant investment in capturing even more information. This fight for information is igniting a separate revolution on the personal lines front with customers overwhelmed with choice at the point of sale or renewal, brokers gathering enormous MI intelligence and carriers who are fighting to become and remain relevant.