Dr. Yinglin Zhang on the indispensable role of data analytics in life and health insurance

00:00:00: Replay by Hannover Re. Expert insights on life and health insurance.

00:00:08: How data analytics can boost your life and health insurance business, what to consider, and why you cannot just rely on artificial intelligence.

00:00:24: This is RePlay by Hannover Re. I'm your host Susanne Loomis and today we'll talk to Yinglin Zhang about data analytics.

00:00:31: This podcast is for general information only, no professional advice and is subject to change. You can find the full disclaimer in our show notes.

00:00:42: Welcome to our podcast, Yinglin.

00:00:44: Thanks Susanne, happy to be here.

00:00:46: Yinglin is general manager of our life and health data analytics team here in Hannover and holds a PhD in financial mathematics. She previously worked in the primary insurance sector and brings quite a unique perspective to our conversation.

00:00:59: More on that a little bit later because first I want to know, data analytics is a big deal in the insurance industry. So tell us why is that and why are we talking about this today?

00:01:10: Well, it seems like a hot topic coming up only in the last few years, but the concept of data analytics is actually not new for insurance at all.

00:01:18: It's about analyzing data for predicting the unknown and/or the future and to support business decision making.

00:01:24: Now, let's think about the concept of insurance business.

00:01:27: Data is actually the only asset of this industry. Analysis and modeling of data is its foundation.

00:01:33: It's a highly regulated industry, so insurers have always been familiar with data protection laws and legal regulations around data management.

00:01:41: But what's new today is the combination with modern technology for data collection and storage, which unlock new opportunities.

00:01:48: Okay, so essentially analyzing data has always been central to the insurance business and significant transformations have taken place over the last few decades.

00:01:58: Indeed, a lot has changed. I can share a little story of a colleague which I respect a lot.

00:02:03: She has been working at Hannover Re in 40 years.

00:02:06: Wow.

00:02:07: Four zero, no joke. And she told me as she started her career, she wrote every calculation by hand on a paper.

00:02:15: Wow, can you imagine that today?

00:02:17: That's incredible. It's impossible for me to imagine that.

00:02:20: And as the company introduced the computer, she was one of the first to use Excel for calculations, and that was high-tech at that time.

00:02:29: Later on, there was the introduction of database, and now she's witnessing how colleagues of my department store data in cloud infrastructure and code everything in the programming language R.

00:02:40: So the process of data analytics comes together with the development of technology, and now we have more efficient infrastructure and more computational capacity.

00:02:49: And where does the data come from?

00:02:51: Well, there are two categories of data which insurance company can use. One is portfolio data and one is third-party data.

00:02:58: When we talk about portfolio data, it's about data related to insurance policies and policyholders.

00:03:03: And those data are, of course, necessary to run the insurance business and to provide insurance cover in case of claims.

00:03:09: When we talk about third-party data, it's mainly about population statistics from national statistics services or from external data providers.

00:03:18: But the availability of such data can differ a lot from market to market.

00:03:22: So now we have all this data, and how does that get implemented in decision-making and achieving favorable business outcomes?

00:03:31: The major use case is the so-called insurance pricing, which is at the core of the insurance business since the very beginning.

00:03:38: It's about setting the level of premium to be paid by policyholders according to the risk exposure.

00:03:44: Let's imagine that we have a woman in mid-50s living in Paris with a certain health background who wants to buy a life insurance policy.

00:03:52: So we ask ourselves what is the level of premium to be charged in order to provide her appropriate insurance cover.

00:03:58: So at the core of this question, we ask ourselves how to forecast the future based on existing available data and how to derive drivers of life expectancy.

00:04:08: So all of this is not new, but what's new today is the level of automation and sophistication of pricing.

00:04:15: Can you give me an example for that?

00:04:17: Sure. In some markets, you have the so-called pay-as-you-live product, which you might know.

00:04:22: So let's say you have a smartwatch, which is combined with a health app on your mobile phone, and app is backed by an insurance product which decides the premium level according to your daily activities and health status.

00:04:36: So in order to launch such product, of course, you need new ways of data collection and for storage and data processing.

00:04:44: In general, data analytics is applied across the value chain of insurance, so it can be product design, pricing, accounting, risk management, until the process of underwriting and claim, and even portfolio theory.

00:04:56: In other words, insurance companies use this data to make appropriate predictions and could even personalize them.

00:05:03: So data analytics helps insurance companies to draw profitable insights that then contribute to their business success.

00:05:11: That is correct, but not only for the business of the insurance company, but the goal is also and mainly to be able to provide appropriate cover to policyholders in a long-term manner, since insurers are also service providers.

00:05:24: So now on a more personal level, Yinglin, what drew you into this topic?

00:05:30: So it's an interesting question.

00:05:33: If I think about the time when I was studying at the university, there was not even such a filled data analytics.

00:05:41: I studied pure mathematics and statistics, and I had the luck of catching always the right industry development at the right moment.

00:05:49: During the wave of quantitative finance, I did my PhD in financial mathematics, applied to insurance with a focus on stochastic modeling, and during the era of big data, I started my first job, which was in IT, and such focused on database development.

00:06:05: As in Germany, there was the introduction of telematic tariff for motor insurance. I was in the pricing team of primary insurance, and shortly after the introduction of Sovency 2, which is insurance regulation for risk management in Europe, I worked both in operation and modeling roles in quantitative risk management for primary insurance.

00:06:24: So without even doing it on purpose, I realized now that all what I know and all what I did contributed to the topic of data analytics insurance, since this requires at the same time statistics, stochastics, computer science, and insurance business understanding.

00:06:40: My current department, data analytics, is a unit which supports all the other business departments, and this requires, of course, knowledge of a large range of actual topics.

00:06:50: So this is where everything fell into place for you. And you have collected quite some knowledge, which you also shared with us during an interview for our recent perspectives newsletter, and we'll link to that in the show notes.

00:07:03: And you have got a very unique perspective. You know what primary insurers are looking for, and you know what re-insurers can provide.

00:07:12: So why did you decide to make the change from primary insurance to re-insurance? And then particularly, why did you decide to start working with Hannover Re?

00:07:23: I'm a person who's always happy to have new challenges. And if we think about re-insurance, that's a very special industry. So re-insurance is at the end of the risk transfer chain.

00:07:34: If primary insurers can still transfer risks to re-insurers, the analysis and modeling of risks have to be taken much more seriously by re-insurers. So this is a much bigger intellectual challenge.

00:07:46: And coming to Hannover Re was a very conscious choice from my side. I was looking explicitly for a company within the insurance sector with a very good company culture, and I even made a personal ranking for that.

00:07:58: Hannover Re was at the very top of the ranking. And of course, I'm very happy with my choice now.

00:08:03: Yeah, that is great, and that is quite a data dream.

00:08:05: very analytical and very much on brand.

00:08:08: And I can imagine that having that insight from your time

00:08:11: in primary insurance really helps understanding the needs

00:08:14: of our clients.

00:08:16: Can you let us know what some of the biggest challenges are

00:08:20: that primary insurers face and how you address them using

00:08:24: data analytics?

00:08:25: I would say there are three major aspects.

00:08:28: One is the legal perspective.

00:08:30: That's about the responsible use of data according to regulations

00:08:34: including how to avoid reinforcement of social biases.

00:08:38: One, of course, is the IT perspective.

00:08:41: So it's about fixing the basics and finding an infrastructure

00:08:44: which balances flexibility and efficiency in data collection,

00:08:48: including validation of data quality and also data storage

00:08:51: and processing.

00:08:52: The last one, which is, of course, the most important one,

00:08:55: is about model and business perspective.

00:08:57: So it's about explainability for insurance core business

00:09:01: and the creation of business value by using data analytics.

00:09:05: What do you mean by that?

00:09:06: I said that insurance business is solely based on data

00:09:09: and analysis of data.

00:09:11: So the financial impact for that is very high

00:09:14: for insurance companies.

00:09:15: It's fundamentally important to understand the underlying

00:09:18: business and the entire data and modeling process.

00:09:21: I always tell my colleagues that every model can be fit

00:09:24: to every data set.

00:09:26: So it's not about always choosing the most fancy

00:09:28: and complicated model like going to a shopping mall.

00:09:31: Nice.

00:09:32: And how would Hannover Re collaborate with clients to tackle this?

00:09:36: We provide data analytics service to our clients,

00:09:39: especially on the topic of explainability and creation

00:09:42: of business value.

00:09:43: Our clients benefit from our specialists

00:09:46: to know how in modeling and analyzing complex risks

00:09:49: in initial markets.

00:09:50: So together with us, our clients can better select risks,

00:09:53: improve customer experience, and expand their current business.

00:09:57: They can also validate their model or third party models

00:10:00: together with us.

00:10:01: This can also include AI models.

00:10:03: Of course, ideally a good partnership with our clients

00:10:06: always leads to a mutual growth.

00:10:09: That's why we are keen to help our clients to better actively

00:10:12: steer their current business by providing explainable data

00:10:15: analytics and actionable suggestions.

00:10:17: OK.

00:10:18: So Hannover Re works closely with clients to turn challenges

00:10:22: into opportunities.

00:10:23: Absolutely.

00:10:24: And partnering with our clients is the key to success for all of us.

00:10:28: That is true.

00:10:30: So there's also the relation between data analytics

00:10:33: and artificial intelligence, which is an important point.

00:10:37: Let's dive a little bit deeper here.

00:10:39: So what is artificial intelligence?

00:10:43: The definition of artificial intelligence

00:10:46: is a quite philosophical question.

00:10:49: I can't really answer it personally,

00:10:51: but if we look at the definition of the European AI Act,

00:10:55: everything which involves a certain level of automation

00:10:58: and modeling can be somehow classified as AI.

00:11:01: And what role does it play in data analytics?

00:11:04: Well, AI in the context of data analytics is often mentioned

00:11:09: in the sense of complex black box models.

00:11:12: Can you explain black box models?

00:11:14: So some models whose decision making processes

00:11:17: are not transparent to the user.

00:11:19: So black box.

00:11:20: And in our concept of data analytics,

00:11:22: AI model can play a role but does not have to,

00:11:25: because the modeling choice is always subordinated

00:11:28: to business problems.

00:11:29: So we always focus on explainability.

00:11:32: But we've seen some primary insurers explore AI

00:11:35: because they think they have to.

00:11:37: Got it.

00:11:38: But it's not always the best choice for everyone.

00:11:41: How do you approach artificial intelligence?

00:11:44: If we look a little bit closer,

00:11:46: the goal is never really applying AI

00:11:49: but to solve concrete business problems

00:11:51: or to create new opportunities.

00:11:53: So our approach with our clients

00:11:55: is to offer dedicated client workshops on data analytics and AI

00:11:59: and to offer data analytics services to meet client needs.

00:12:02: You mean HR Blue Box?

00:12:04: Exactly.

00:12:05: HR Blue Box is our data analytics service

00:12:07: which we offer to our insurance clients.

00:12:10: So if they have business needs

00:12:11: but do not have sufficient internal data analytics,

00:12:14: capacity to handle that, they can come to us.

00:12:17: Customers are of course always at the center of our business.

00:12:20: If we are the reinsurer, business problems of our clients

00:12:24: are often our own problems.

00:12:26: But the way how we offer our service

00:12:28: is very different from our competitors.

00:12:30: And what does Hannover Re do differently than the competition?

00:12:34: We aim always to add concrete business values

00:12:37: and we do not explicitly focus on artificial intelligence.

00:12:41: So we consciously do not offer end-to-end machine solutions

00:12:45: but rather individual solutions

00:12:47: since business needs of our clients are always individual

00:12:51: and we always add a human touch, especially for explainability.

00:12:56: And our clients' projects are always very short

00:12:59: so that we can provide actionable recommendations

00:13:02: as soon as possible.

00:13:03: The most important thing, we have a very international team.

00:13:06: That is true.

00:13:07: And you even said that you are a professional

00:13:10: and you even speak four languages, is that right?

00:13:13: Right.

00:13:14: I have lived in three countries and speak four languages fluently

00:13:18: but another colleague in my department

00:13:20: can even speak five languages fluently.

00:13:22: That culture context is very helpful to understand different expectations

00:13:26: especially also to enhance collaboration with our local offices.

00:13:30: We have a very flat structure so that we can enable quick decision-making processes

00:13:35: and all colleagues in my department are highly motivated and skilled

00:13:39: and everybody has a very high personal responsibility.

00:13:42: For every client project, we have dedicated colleagues

00:13:45: who follow the projects from A to Z.

00:13:47: So there is no functional division

00:13:49: so that we can avoid miscommunication

00:13:51: and can deliver values as quick as possible.

00:13:54: Okay, so in conclusion, the ways and opportunities to analyze data

00:13:59: have changed drastically over the last decades.

00:14:03: However, the core remains the same.

00:14:05: It's the key to making suitable predictions and to support business success.

00:14:10: And Yinglin, as a final thought,

00:14:13: coming from a re-insurance standpoint

00:14:15: and having a good understanding of the needs and challenges of primary insurance,

00:14:20: what's your best advice for them to get the most out of their data?

00:14:24: I would say four cues on basics, roses and hypes,

00:14:28: and invest in modern infrastructure and methodological development.

00:14:33: I would also emphasize that people component is even more important

00:14:37: since knowledge and technology can only be effectively used for business

00:14:41: if we have good people.

00:14:43: So strong focus on talent management and a culture of agility is essential.

00:14:48: Oh, well said.

00:14:49: So in this episode, we heard how data analytics and insurance

00:14:53: has evolved from manual calculations to sophisticated AI models.

00:14:58: Nowadays, it drives pricing strategies

00:15:01: and enables personalized products such as pay-as-you-live insurance

00:15:05: and improves overall risk management, if applied in the right way.

00:15:10: Thank you so much, Yinglin,

00:15:12: for shedding light on these transformative elements of data analytics

00:15:15: in the insurance industry.

00:15:17: Thank you, Susanne. It was great to be here.

00:15:20: And thank you for listening to RePlay by Hannover Re.

00:15:23: We hope you found this conversation helpful.

00:15:25: And if you would like to learn more about data analytics

00:15:28: and life and health free insurance,

00:15:30: check out Yinglin's interview as part of our recent Perspectives newsletter.

00:15:34: Like I said, it's linked in the show notes.

00:15:36: And don't forget to subscribe for more episodes.

00:15:38: Leave us a review if you enjoyed what you've heard

00:15:40: and stay connected with Hannover Re on LinkedIn.

00:15:43: Until next time!

00:15:45: [Music]

00:15:55: The Thoughts, Ideas, Documents and other content discussed or shared in this podcast

00:15:59: are not intended to constitute any professional advice.

00:16:02: Hannover Re and the participants are not liable for any damages resulting from the usage

00:16:06: of the content of this podcast.

00:16:08: For further information, please see the show notes.