[Transcript] Looking beyond the tech to examine the science of data & analytics with David Pardoe, AIG

Add bookmark
Craig Sharp
Craig Sharp
10/09/2014

Ahead of Insight & Analytics for Insurance in January, PEX Network sat down with David Pardoe, Head of Science (EMEA) at AIG to discuss the company's current big data strategy, and the key role that analytics, and the right analytical minds play in extracting maximum value from data-driven projects.

Craig Sharp: Could you summarise AIG’s current analytics system and strategy?

David Pardoe: AIG made a significant investment in the whole area of Science by setting up a Science team from scratch in early 2012, so we’ve been doing this for nearly three years now. The Science Team was set up by Peter Hancock, who, at the time, was CEO for the Property and Casualty part of our business. He hired Murli Buluswar as Chief Science Officer and had the Science function report directly into him. This is incredibly unusual, not just for insurance companies but for large companies of a similar scale and size as AIG.

Typically, you would find that these types of analytics or data science functions sit in the technology side of the business, or you find them sitting within a functional area. Peter knew that Science was going to be at the centre of AIG’s strategy and would reach across the whole business.

There was a lengthy presentation to the board of AIG several months ago where we communicated the impact that Science has had right across the whole of the business. They were incredibly impressed and I think it is fair to say that Peter’s investment has paid off. As Peter Hancock has now taken over as CEO for the whole of the company and has continued to have the Chief Science Officer report directly in to him the impact of Science on the whole business will continue to grow.

We now have over 140 people within the Science Team with teams in the US, UK and Japan focussed on different geographical regions and other teams that focus on distribution, claims, underwriting and other functional areas; our work continues to deliver results right across the whole breadth of AIG’s business. We are also looking to expand into other areas such as HR; employee engagement, hiring processes, employee retention etc. We continue to focus on hiring the best Science talent with experiences across many different sectors in order to apply that experience to Insurance.

Although the size of this team is comparable to similar teams in other organisations we have a clear strategy to develop Science based approaches and thinking beyond just our team. As I mentioned, the key here is that our reporting line is straight to Peter Hancock; that’s a crucial difference, and very important for what we are trying to achieve.

Craig Sharp: Do you think that’s made a huge difference to the success of the team?

David Pardoe: Without any shadow of a doubt – by having a function that reports to the CEO, we are right at the leading edge of AIG’s business strategy. As a result our focus has been on parts of the business that really make a difference and are central to overall strategy. We are not just focussing our attention in a single functional area and limiting ourselves to only thinking about that area; we can take a more holistic approach and see all aspects of how our clients, customers, brokers, agents and partners interact with AIG. We also have not constrained ourselves by thinking about what tools and technologies we want first. By looking at the business angle we can then determine what our technology solutions need to look like.

Our fundamental principles are around using data science techniques to inform decision-making and help it be more evidence-based. Our mission isn’t about "doing analytics" or "analysing data" or "providing big data platforms" – our mission is about changing the way decisions are made in the company.

Craig Sharp: What do you feel are some of the benefits that AIG have already experienced taking this approach and using data analytics in this way?

David Pardoe: As our focus is on partnering with the business we have put a lot of time and effort into understanding the complete insurance cycle, or the value chain to put it another way. We look at the whole process from how business gets submitted to us by brokers or directly from consumers, through to the decisions we make on quoting that business or not, how we decide to price it, through to the binding of that business. We then look at what subsequently happens, the servicing of that business, the servicing of that policy. That is obviously focussed on claims but also touches on things like mid-term adjustments; for example if a client wants to change the coverage.

Then you can track the business right through to the point of renewal. Do we actually want to renew the business, do we want to renew the business and increase the rates – ie, change the price. Do we want to automatically renew policies versus manual touching each and every policy which is what we have tended to do historically.

By looking at the complete cycle we can identify where Science can inform the decisions that are made and either become ever more cost efficient and generate more premium whilst ensuring that we can manage our claims and losses.

We’ve also done a lot of work around our distribution channels including segmenting our brokers. We have developed a broker quality index, which looks at the historical and predicted future performance of brokers. We can use that to manage those relationships more effectively, committing more resource and effort towards the value creating relationships.

Craig Sharp: How about with regards to the customers? What kind of direct benefits have the customers experienced as a result of your strategy?

David Pardoe: The work we do provides direct benefits to our customers in many ways. One of the more obvious is the work we have done in the claims area. We have "text mined" unformatted text captured in adjuster notes and one claim reasons. We then package up this analysis and insight for our clients who can then benefit by understanding where their biggest risks are and therefore invest in things that can mitigate these risks.

Another we have looked at specifically is how investments in medical rehab can reduce workers compensation claims. This not only benefits our clients in reducing premiums but also reduces our exposure to claims.

Craig Sharp: How could, or how should companies modify their existing analytics processes to address big data?

David Pardoe: There are two fundamentals here from my perspective and from the Science Team’s perspective, both aligned to themes I talked about earlier, and both related to one another:

The first is that we are fully focussed on working in tandem with the business - that’s particularly important for insurance companies and even more so for AIG which has years and years of phenomenal experience and expertise in insurance. What data science is doing is adding to that; we are not changing what has made AIG so successful, we are building on it. The Science techniques we are using provide information and data driven evidence that enhance the decisions that are made. This touches on one of the other areas that I’ll talk about, which is that whole area of cognitive bias.

The second would be that we are not focussed on the technology alone. There is little sense in a company like AIG investing in the latest Big Data technologies and solutions unless we know exactly how they are going to be used. Whilst there is no doubt that these new technologies are extremely powerful and effective, unless there is a clear link to how the business operates, it is unlikely you would ever see a return on the, not inconsiderable, investment required. Thinking about things this way round is crucial.

Craig Sharp: So, you’d say, think beyond the technology and invest in the science and the analysis?

David Pardoe: Yes, exactly right. It is very easy to get hung up on the latest buzzwords around "Big Data" and think about them only in theoretical or abstract ways. There is simply little point in making large scale investments with no clear idea how you are going to use them. Imagine buying a Ferrari but all you end up using it for is the school run!

Craig Sharp: You mentioned cognitive bias. Could you explain cognitive bias, and can this be used to aid and strengthen a company’s strategy?

David Pardoe: Cognitive bias is a collective term that describes the biases that influence how people make decisions. This is due simply to how the human mind works. There is a whole strand of exploratory and research work around lots of different types of cognitive biases; there are literally hundreds.

Understanding how cognitive biases shape our decision making enables us to do two things – firstly it enables us to be aware of the biases that we have and adjust for them; and secondly, it enables us to go seek data and information that will negate the effect of that bias.

I’ll give you one small example. Say you asked the following question to one group of people ‘Was Florence Nightingale older or younger than 50 when she died?’ Now you ask a different group of people ‘Was Florence Nightingale older or younger than 90 when she died?’

Now you go on to ask both groups ‘How old was Florence Nightingale when she died?’

You get a very different response from the 2 groups; the first group will estimate the age when she died to be much less than the second group; yet the only difference between the two groups was the number you used in the first question.

Craig Sharp: So simply by changing the first question to whether she was older or younger than 90 rather than 50 has primed people to guess older?

David Pardoe: Exactly. The cognitive bias in effect here is anchoring. Your mind is anchored on the number that you’ve been given. When you are asked the second question your mind works by "moving away" from the anchor and it is simply the case that you never move as far away from the anchor as you should. By the way the the number could be completely arbitrary; you could even use the question ‘was she older or younger than 140 when she died’, which is clearly ridiculous. I personally ran a session with some of our internal stakeholders in Prague a couple of weeks ago; randomly split the group into two and proved that this bias affects our people. If you know you are affected by anchoring you can then consciously ensure that you move further away from the anchor than you might do subconsciously.

Another cognitive bias is the status quo bias; this is where you make decisions in the same way you have made them before. So if you decide whether a claim is fraudulent based on a subjective viewpoint of the way it is worded or how the claimant sounded when they spoke to you then you are extremely likely to use the same type of judgement every time you determine if a claim is fraudulent. You are sticking with the status quo; the same method you have always used. The reality might be that there are other reasons that point to a fraudulent claim that you have simply never considered before.

Using Science enables you to challenge this unconscious decision making bias. By building predictive models that are based on data and are therefore totally immune from cognitive bias you end up making better decisions. It is important to note that here at AIG we are certainly not espousing that Science replaces human judgement but rather using it to enhance that judgement.

[eventPDF]


RECOMMENDED