Insurance exec: Big data allows better understanding of risk

The insurance industry is adjusting its offers on health, cars, the sharing economy and other services because of the growing amount of personal data becoming available.

Zurich Insurance’s CTO Robert Dickie told euractiv.com in an interview that big data is changing the way insurance works and making the industry more competitive.

Robert Dickie is the Chief Operations and Technology Officer of Zurich Insurance.

Dickie spoke to euractiv.com Editor Frédéric Simon.

The emergence of big data puts insurance companies in a position where it can know a great deal more about its clients than ever before, for example, with health monitoring applications or driving behaviour. What kind of challenges is this raising for a company like Zurich?

Insurance is based on the solidarity principle. Data allows us to pinpoint better ways of identifying price and risk management. This is a good thing, for us and the consumer. It means we explain the risks we write and price them better.

At its most extreme, shall we say, it allows us to be more selective on those risks and set portfolios that are more attractive.

When it comes to the sectors you mentioned, the insurance industry already has access to a lot of data. Again, we see this as a way of getting richer veins of data in order to provide better risk assessment. Of course, the concern is what this does to the solidarity principle and how we manage risks across the entire population.

It allows you, in theory, to discriminate a lot more than before, in the sense of pricing.

I certainly wouldn’t use the word discriminate, as that suggests we are being negatively selective. We are trying to understand the risks that we are looking to underwrite. It’s a selectivity of risk.

Insurance companies are aggregations of risk portfolios. Data helps us improve that. After all, in the insurance industry, we are in competition with others to provide risk assessment, balanced with price. It will be an aid to competition.

So you see this as an opportunity to be more selective on how you define profiles for different sorts of clients?

Yes, because we are trying to price risks better for clients and support them in mitigating it. Insurance is intended to allow people to go about their lives without the worry of unmitigated risk. Big data allows us to do that.

More data allows you, in theory, to be more selective. But some current legislation could prevent you from actually doing so, for example, laws on gender.

That’s a distinct possibility. That is the discussion between the industry and regulators. We all have a shared interest in the solidarity principle. Take the Japanese market as an example. Quick cover was extremely expensive and difficult to come by, so the government has its own fund, which insurers are a part of.

The data we are talking about should aid the dialogue between regulators, the government and the individual organisations. In a competitive market, there is always going to be someone that will underwrite a risk and see it as attractive.

Better data and all the participants being regulated, to me, seems like a good vehicle to see how the market is being provided and that there is adequate competition.

Back to health. If more is known about an individual, there’s the possibility that people are going to be priced out and become uninsurable. How do you see that being solved in a big data world?

I’d challenge one assumption there. The consumer always owns the data. I don’t think people are going to be compelled to provide their data. We are stewards of data, not the owners.

Of course the attitude to data varies around the world. Personally, I think there will be a relationship between those people who are willing to share their data and those that are not. Of course, being prepared to share data, with mutual consent, could lead to different pricing, not necessarily better naturally.

People would only be willing to share if they think they are going to get a better price, one would assume…

I’d agree with that, but again that’s an assumption. Some people simply don’t like sharing their data. Their data is worth more to them than the savings they could potentially make. But I agree with your logic. I think that there will be a separate market for those willing to share and those that aren’t.

How do you think the regulators will react to this? You’re essentially talking about a two-speed market.

I think it would be multi-speed actually. But you’d have to ask the regulators. I think that people are still at a stage where these models are still being tested out. I think it represents challenges to regulators certainly.

On data ownership, some people wonder whether the clients should benefit from the emergence of big data by selling their data themselves. Would you welcome this?

On a personal level, more data awareness among clients is always better. It helps the dialogue greatly. In the past, many insurance models were set up as mutual.

One model that would be interesting would be a return to that. People are prepared to share amongst groups, the basis of a mutual. Spreading risks and looking after each other. In terms of emerging models, it’s a virtualisation and rethinking of the mutual model. I think that could be attractive to regulators.

When it comes to motoring, how are you preparing for the developments there? You’ll no longer be signing contracts with individuals, it’ll be with companies instead. 

Yes, we’re talking to individual customers and car companies. Why? Because it’s moving from an individualised contract to something similar to how we buy finance on vehicles currently. There has to be clear risk accountability.

Who is accountable for the risk? The machine? The satellite link? This is the kind of stuff that needs to be figured out and what the regulators are still working on. It’s only by testing and trialling that this will move into production.

The sandbox idea is a good way of getting this done. The technology is going to be there, it almost is already. Some carmakers are already ready to roll this out. Tesla is racking up the data, Uber is compiling dataset on dataset. The tech problem is going to be solved. It’s human acceptance that will be the issue and the regulation behind it. It’s also going to take a decade to change the current stock.

Depending on the technology car manufacturers opt for, owners of certain car brands are going to be charged less than owners of other brands, because of safety. Is that the logic?

Again, if we are getting feedback on accident history etc., then the answer is going to have to be yes. But it is in the carmakers’ interest to up safety, as their sales would be affected otherwise.

Turning to a different topic, what are the specific challenges you see when we talk about the sharing economy? Do you see this as a new market opportunity?

Yes we do. We’ve been speaking to Uber and AirBnb. When it comes to the former, the challenge is at what point does the car become a commercial vehicle driven to transport valuable cargo, i.e. humans, where is the transition? It should be technically easy to identify. When the driver is driving or the purposes of commercial enterprise.

So you would have to apply different rates at different times of the day…

Correct. That’s how it works with telematics at the moment. The difference I would highlight is that in the past it was a blanket time period covered by insurance, this can now be broken down and priced seconds on minutes instead of year on year.

Would you say this is manageable? Do you need supercomputers to manage this data?

Well it’s data, so it can be captured. I’d describe it as an ‘atomic view’ of insurance.

The key thing in computing is not to put the hardware in before what we understand what we are trying to architecturally construct. And it’ll be about the right algorithm that’ll allow us to do dynamic pricing, not supercomputers. We can get that through the cloud.

Now, is any insurance company ready to do this currently? I would doubt it. It’s something we have to get our head around completely. The Climate Corporation is an example I would cite in which data and the right algorithm were connected. They identified attractive datasets and attractive risks.

We spoke about Uber, but not much about AirBnb. Any specific challenges there for insurers? I guess you must see this as a new market opportunity?

I think it’s great. It comes back to data. I’m not sure that the user, the provider or the insurer are always aware of the risk that is being underwritten. It’s one of the unseen elements of the sharing model and that it’ll only take a few occurrences for it to form a part of the insurance form when you first apply for it for your property.