Ben Telfer:

Hello everyone. And welcome to today’s ICMIF webinar: “HUK-Coburg’s journey to create an effective telematics solution”. Today we will hear from German ICMIF member, HUK-Coburg, who will share their insights and experiences in how they have developed an effective telematics solution.
I’m very pleased to welcome Dr. Daniel John from HUK-Coburg. Daniel is head of the actuarial department for non-life insurance. Daniel, thank you for joining us today, and I’m pleased to hand over to you.

Daniel John:

Thank you. And then welcome all of you to our journey to telematics. Maybe just somewhere to add, so you will see in this presentation that people are very, very happy with our telematics solution and they are so happy that they are just beginning dancing on the street as you can see here on this picture. And you will see some pictures of China here to just illustrate this talk, but that has no special meaning, it’s just because I like China. And I think many Chinese people also like artificial intelligence, so that was what I learned in China. So to start our journey, for the beginning, I would like to mention that HUK-Coburg is Germany’s market leader in motor insurance and we have about 13 million of vehicles insured.

We started thinking about telematics maybe in 2013, I would say. And the idea was very, very easy. We thought about introducing a smart driver product. And the idea was, we have a young driver, we have a box, we install the box in the car, and then we shake this box. And from this shaking we can learn, okay, is this box driving very calm? Then we give a big discount to the customer, or the other way is, there are a dangerous shock in this box, then we probably send out the rescue. So first, thinking in 2013, but until the solution was really launched, it took some time and this first product, our first telematics product was in the market from 2016 to 2019, approximately.

It was a fixed box that you can see here, a box from Bosch. And the product was… as I’ve described, it was a telematics tariff and it was a rescue function for young drivers. Before we could start, we have to be honest, there was very much work and many challenges that had to be solved. For example, the box installation process was very complicated. We had to think, “Where can we install this little box in the car? How can we install it?” And so on. It was, “Where can we take the current?” And so on. That was very complicated. We had lots of issues with IT security. Calibration of sensors was a topic, map data was a topic, wrong speed limits was a topic.

And so there were many problems that had to be solved. So this box product was just introduced and we already began then thinking about a new product because we realized, “Okay, this box has very high cost.” The box itself was not so cheap, I would say. And also the installation procedure was tensive. And so we thought about a new product, which is now active since 2019. And this new product is now for the whole market in Germany, for the whole car insurance market in Germany. So we are offering now telematics for all of our customers. The idea is very easy. So now we have what you can see here on the left. We have a little tech. It is called a tech. It’s a sensor which you can see low on the right, the bottom on the right.

And this sensor measures acceleration. And then we also have smartphone, or your smartphone and this smartphone collects the GPS data, time and speed, and so on. And the idea is very easy, if you drive well, then you can get up to 30% of a discount. So why are we doing all this? We are doing all of this… So let me see. Yeah, no, one big… So we’re doing this because we think that’s the customer expectations of today. Everyone wants to have it online, quick products, helpful, direct, easy, entertaining, convenient, 24 hours around the clock, anywhere, mobile companion. So these are the passwords behind that.

And second point is, we want to defend our business model, which we think is very, very important because we see the car manufacturers who have direct access to car data and very easy access to such data, and therefore could provide such insurance products very easily. And last point for us, it’s a model… this new product, it’s a model for digital product, it’s something we can learn about data science. We can learn very, very much about data science here. And it’s also… maybe interesting point is that it’s an example for digital ethics. So what to do with data, how to work with such data, which we think is very important, this ethics view on it. So after all, we want to learn from this.

Maybe some facts and figures. Right now we have about 400,000 contracts in this product, in this telematics product. So these 400,000 contracts have been collected during the last two years. And you see some key numbers here. For example, there have been 240 million trips driven. We have about 64 million hours driven. We have more than three billion km driven, the maximum speed was 259 km. So you see, we are in Germany here. And one of my colleagues did this trip with this high velocity. And maybe some data volumes, we have about 430 terabytes of data, which is very, very much. And we are having now about 16,000 accidents in motor liability, just in motor liability, with the other claims it’s about 50,000.

And yeah, we are happy with these accidents because now we can do statistics with that. Okay. So maybe also interesting for you is, with what is written here. So about 10 to 20%, all customers include telematics in case they have a specific contact with us. So it’s not sold that actively, this product, but customers are asked, “Do you want to have this product?” And 10 to 20% of all the customers take it? And I think that’s a very, very good number. So in a few years we will have one million to two million contracts in our portfolio. Okay. Yeah. Maybe that’s also interesting. You can see our customers are driving all around Europe. We have been maybe everywhere, just where we have some mountains, it’s not so easy to go.

But after all that’s… yeah, they are very interesting where they go. So some words about our system architecture, I think that could be interesting for you. First of all, what we get is a continuous stream of data from the car and it has a position. We have the speed and we have the acceleration. And these data stream is then enriched in the backend through the information about speed limits and about type of roads and to some quality measures. And maybe some expansions might be weather or traffic information. Okay. Then these data come to us and we have here our clusters, and it’s on-premise at the moment, but probably this will move to the cloud in near future.

So at the moment we have two operational clusters, Hadoop clusters and one for analytical purposes. So we use Hadoop, the Hadoop system and lots of different modules from that. And so I would say, as an actor, life has changed very much in the last year. So lots of data science know-how is needed for that. Maybe some… a little few into the fork of the algorithm, so what we do with these data. And so how does this algorithm work? The algorithm for assessing driving behavior. So what we have, we have this time series of speed. We have this time series of acceleration, braking and of cornering. And then we take this time series and we try to identify events in these data.

It’s very easy events. For the beginning, it’s just classifying data in, where have you been driving more than 10 km, faster than allowed, where 11, where 12 km and so on. So it’s just classification of events. And then if you have these events, it’s just counting these events and calculating a score, which is after all very easy, I would say. So it’s maybe even boring. It’s some kind of machine learning, it’s kind of a regression algorithm. And so I think that’s not so interesting probably, but the important point is, it correlates with the claims. So it’s really significant, so we can see the causality between driving behavior and the claims, so that’s the interesting point.

So this is the easy version I would say, but sometime has evolved since we started. And so I would say, after all maybe it’s not that easy any more. It’s not just these easy features for speed or so. But we have learned very much and now the features are much more complex and they’re much more complex because they have to show what is happening in reality. So for example, if a customer is driving on a highway, so he’s on the highway entry. So then he has to do very fast acceleration or very high acceleration. And that’s not bad, and that’s no risk. And so what we did is, that we tried to incorporate these points from reality in our algorithm and into the features. So we have now different features, which reflect this reality.

I think what is important is that such a telematics score is very, very fair rate making factor. It’s really non-discriminating if you build it in a good way. So, because you have to think about the pulling points, the target figure is just claims frequency or average claims expenditure per risk in a year, so that’s just what happens. On the claims side, we use of course scientific models. Only the behavior of the driver is what counts. And of course you can influence that by yourself. So we don’t use sociodemographic factors in the score, like age or income or occupation and so on. And so this makes it really, really fair rate making factor.

And after all it’s voluntary, you don’t need to take this telematics tariff, you can just use the traditional tariffs and you’re fine. Okay. Maybe this is what most customers are interested in. It’s the discount and the distribution of the discount, and the question, what can I expect? So many people say, “Okay, insurance, I don’t really trust them. So I suppose you don’t get a discount or very low discount only.” On the other side… and that’s interesting, most people hope they get 30% of a discount because they think they are really good drivers. And so I have to say, the reality is somewhere in the middle of that because reality, for example, your body height is normally distributed. The intelligence is normally distributed. And so it’s not difficult to conclude that also telematics discounts will be distributed normally, mainly.

Now, and the second important point for the customers is, which score is really possible? What can I achieve? And so I show you this, this is the driving style of my boss, and he achieves a score of 30, which means he gets a discount of only some small… only a very small discount. So this is my driving style, which you see here and I get a score of a hundred, of a hundred points, and which means I get a discount of 30%. And yeah, I really tried that, that’s no fake. I really tried that for one year with our smaller car. And for one year I got a score of a hundred, just to see whether this is really possible. But I have to admit, it’s not easy to achieve this score of 100, you have really to drive careful.

You don’t need to drive very slowly, but you have to look in advance very, very much. You have to anticipate what will be happening on the road. And if you do that, then you can achieve a very good score. And the important point in my eyes is that all depends on you. So you can decide how to drive and you can decide which score you will achieve by this. So 30% is a fact, it’s not an empty marketing promise.

Maybe some words about explainability. So many customers think this, or many people think such an algorithm for assessing driving behavior has to be an easy regression algorithm which is easy to explain. And many customers also ask us, “Okay, just tell us the features that go into this algorithm precisely and tell us precisely which weights these features get, and so on.

I would say, no, it must not be in regression algorithm because that does not at all reflect reality in every detail. So to show you more, or to tell you more about this, I just want to tell you a little story behind. So once upon time I had a conversation with my CEO and he told me, “Okay, I have a driver, I have a chauffer, and he’s a professional driver. And he has very, very fast acceleration and very, very harsh braking, but he never has an accident because he is a professional driver. So this professional driver would get… using such a regression algorithm, would get a low score, of course, because of his braking and acceleration. But of course he has low risk.

The question is, could you build another algorithm for assessing the driving behavior, so that this algorithm reflect exactly this idea? And so I thought about that and I tried to build such an algorithm now, and the idea is similar to the one before. So we get these driving data, we get the speed, we classify it into classes, 10 km faster than allowed, 20 km faster than allowed, and so on. We have acceleration and braking, and so on. But now we don’t use a regression, but we use a tree algorithm. And so now imagine we have a driver and we look at this driver and we look, how often does he do speeding above 20 km/hr? So if he does it often, then we will look at the next step, whether he does braking more than one G, very often or rarely.

And if he does it often, then we also look at cornering. And if cornering above one G is often, of course he will be a very risky driver, okay, that’s fine. If he does not do it so often, maybe it’s not that risky, also fine. But now we’ll look at the other branch of it. And so, he does speeding… so now we take the professional driver, the chauffer of our CEO, and this driver does not do speeding more than 2O km so often, because if he would do it, he would lose his license. But he does acceleration of one G very often. Speeding of more than 30 km, never, ever, because he would lose his license. And then braking of more than 1.5 G, he does very often, because if you accelerate fast, you have to brake very fast, because otherwise you will crash into the wall, probably.

And now we see, okay, if you can see this behavior, then maybe he is a good driver, so he has low risk. And this is just to show you that you can construct algorithms different from regression algorithms, which can reflect the reality better than the regression algorithm, so that’s what I want to show you. We see now the rest of this tree. And I would say something like that could fairer, but of course this is much more difficult to explain. Okay. So what is very modern at the moment is neural networks. And so neural networks, the question is, what can you do with them for assessing driving behavior? So my opinion is that, neural networks are not the solution for everything, but for a very narrow problem, they can be a very good solution.

Maybe in something like image recognition. To give you an impression of what neural networks can do, you here are two ladies, and what your brain does usually is, you see the images and then a neural network in your brain starts to work. And then it says to you, “Okay, this is the so-called violet lady on the left. I know her. And on the right, I see the so-called blue lady. And I also know her.” So you can recognize who it is. So for a neural network, you may ask questions such as, “Is this the blue lady?” But of course you may not ask, “Tell me everything about this person.” So that’s not a question it can answer. We did some experiments with such models to assess driving behavior. And what we learned from that is that properly black box models will not… it will not be so easy to use them to assess driving behavior.

If you just think about… we take all of our driving data, put them into this algorithm, or into this net, and then shall learn what is risky driving behavior, then probably this will not work, because it’s too much data and it will be difficult for the net to learn, “Okay, what is the important point? What is the risky driving behavior in this?” Because we just have too few claims, too few accidents. So after all, what will probably be needed at the moment at least is a human for this modeling. And this human might have maybe the following idea. So he would say, “Okay, if all of the data is too much, then I will concentrate on some parts of this data. I will cut out some pieces of this information. For example, I will cut out acceleration maneuvers, for example, accelerating from zero to 30 km/hr and I will then look at all these maneuvers, and of course, people have different driving styles, and do different styles of accelerating.”

And so we collect these pieces of data or these acceleration maneuvers, and then we classify them using the accidents we have attached to these drivers. We can classify these maneuvers as more risky driving style or less risky driving style. So for example, you could expect that a high acceleration at the beginning would be more risky and low acceleration would be less risky. And yeah, just drivers with very slow acceleration measure also show a slow reaction. So this may once again then be risky. Okay. So that’s my opinion here, I would say. And that’s what we have learned. So you should probably avoid completely black box models. That’s what we have also discussed with the European supervisors, with EIOPA.

And also consumer protectors see such black boxes as very dangerous, most of them. But maybe such gray boxes, just using very small pieces of driving maneuvers also would be then explainable in a certain sense. And maybe it’s a good idea to use. Okay. Another very, very interesting point is giving feedback to the customers. Some people say telematics is a lot of fog and you don’t really know what you get. So benefits are unclear as a customer, you don’t know what you will get, and everything is non-transparent. So the question is, is this really true? And to illustrate this, I now would like to introduce to you Ronny Racing. Ronny Racing is a very good driver… No, I don’t know, very risky driver, I would say.

And so you see here his first trip, so he got the telematics product from HUK-Coburg and installed it in his car and then did his first trip. So you see here, he drove from Obukhov to Coburg, that’s somewhere near here. And during this trip, he got a driving score of nine, so really bad driving score. And you see also here, his forward acceleration was not that bad. His backward acceleration was really bad, so he did some very harsh braking, probably his speed was somewhere in between and sideway. So his cornering was also not that good. We also gave him some hints where he did some maneuvers, which were risky. And of course his reaction comes immediately.

This is our first customer complaint from Ronny Racing. And he says, “How can this be? This is my first trip. I’m driving extra carefully. Then this slow vehicle appears in front of me and I have to brake. Am I being punished for not tracing into this traffic obstacle? This is unacceptable. A score of nine is outrageous. 100 can only be achieved if you are obstacle yourself. I’m getting rid of this.” And we are always very, very polite and say, “Thank you for your feedback Ronny. So such feedbacks, such complaints are standard for us. We can have this every day, I would say, some of them. And if we then look into the data or into the trips, so we will see the falling… in this example, we see the falling.

At the beginning, he did a kickstart, but we didn’t took this into consideration because it was just at the beginning of the trip. And then he did some racy driving through bends. So very harsh cornering in the mountains near Coburg. And then later on, braking for the obstacle, what he tells us is, he had 200 km/hr. So we are in Germany once again, unlimited speed, from 200 km to 160 km/hr with a braking of about 6 m/s2, which is really hot braking. And last but not least, he did… at the end of the motor way, where he left the highway, he braked shortly before the exit, with 5 m/s2, from 180 km/hr to 70 km/hr, not to fly out of the curve.

You see, after all, we were right with our score. So what could he expect better as a score than nine? So it’s absolutely fitting his driving behavior, we would say. But of course he does not see it himself like that. So there are also better examples, if you adapt your driving style, you can see maybe something like that. That’s a trip from Coburg to Wolfsburg. And here we have a score of about 87 and we just see one small event. It’s a little acceleration, high lateral acceleration at the motorway entry. And then we also get very good feedback from customers who say, “Okay, I think this score is highly accurate. One day I drove with many events and the next day I tried to drive better, and I got much, much better score. So I think this is a good product.”

What you see from this is that, giving feedback to customers is a very emotional topic. It’s about pure psychology, I would say. It’s about the topic of self assessment. And to learn about that, we did little surveys. And so we asked some people around us, and we asked them, “How good drive are you?” And yeah, the answer came always quick as a shot. Most of them told us, “Okay, I should get a score of 80 out of 100, 80.” Then we said, “Okay. But maybe you think about that once again, because the average score is 50. What do you think should be your score?” And the answer was, with total conviction, “If 50 is the average, then I should get 70.” Yeah. And what you see from that is that, it’s hard to accept that you are maybe not that good driver, and telematics is something like a mirror which tells the truth.

And giving feedback is always very difficult. And this is the dilemma, so how much feedback should we give? And that’s not easy. So the question is, should we give rather less feedback? So maybe just as in this picture, just three events, yeah. Major events to show where you did something risky or should we show more events? Like in this picture here, should we show maybe every event? Yeah. And what will happen is, if we just do some major events, then the reaction of the customer would be… for this trip here, “I only braked once to avoid an accident, and this though cannot explain a score of nine, so I’m not happy with this” would be his reaction. And on the other side, if you do also minor events and maybe all events, then the customer would say, “Okay, so many events, this is unacceptable. This is not true. This cannot be true. I don’t believe it.” And they also will not be happy.

The problem is to get a good balance between just some major events and also all events. So this is the problem. Okay. Maybe one further point, which is interesting is the point of errors. So what did we learn on our journey to telematics? So we learned, errors can’t be avoided. And this is very important to realize, so you can’t be perfect with your solution. So what we use is a good, I would say a good, but low priced, a combination of technologies. So you have your own smartphone and you have a little sensor with acceleration data. And this can deliver good data for evaluation of driving behavior. But there are so many different smartphones phone models with so many different features. So many settings, which are important for all of this to work, like GPS position, the Bluetooth has to be on the power saving mode, can make problems, and so on.

We have a very, very high complexity and we have constant changes. And so errors are part of our lives. For example, pairing is not working, so this smartphone is not connected to this little tech. This can be a problem or map matching can be wrong, or so on. So there can be many, many problems. And so what we learned from that is that we have to permanently strive for improvement, that we have to try to predict changes in technology. And we have to react very quickly and we have to do lots of testing. And after all, what is also important, we have to listen to our customers. So we have to see whether there are complaints and in which direction these complaints go and how we can solve these problems then. That’s very, very important.

Just an example to show you what can go wrong or what kind of questions we got at the beginning. So we have here a highway in Coburg, and the question is, is there a speed limit of 100 km/hr at this special point, on the highway? And to see this, we look at some trips from our customers, and we see our customer which is going with 150 km/hr, then slowing down to 100 km. And then his speed is going up again to 150 km, there where it’s 150 again. The speed limit itself which is in the map, so the map says to us there’s a speed limit of 100. So the question now, is this true, or is it not true? And to decide this, we built a machine, we have built a machine an algorithm, which can predict very well the speed limits on the street and streets.

And so what we are doing is we are looking at many, many trips at this point, and we are looking whether the customers are driving there about 150 km/hr, or whether they are driving about 100 km/hr. And then using this, we can say, “Okay, there will be speed limit or there will not be a speed limit.” And in this way, we built a map for Europe. Okay. So a hundred percent accuracy of map data will never be achievable. That’s important. And when we introduced this new telematics solution, we had some very interesting success stories, I would say. And just to give you some examples, our first success was beaming. So I did a trip here, which looks very short. So it’s a very short trip here in the bottom of the picture.

But reality is, I started several km further. And probably what I did is just beaming. But yeah, it’s just an example. We see only the end of the trip has been recorded by our technology, and that’s not good. So this will result in customer complaints, probably. Second success story is, flying. So one day I went to the school of my son. And you see here, at some point, I just took off and flew above Coburg, and you see here, this is probably the real route I took. But I don’t know how the solution just recorded a wrong way, so I could fly. And last, an example here is, cloning. You see here a trip, and then the point where the yellow line starts, that’s where I started my speed limit.

I was much faster than the speed limit. And so on the map it’s shown as a yellow way to the left, and the blue line to the right. So I split it up myself. I cloned myself and went on two ways at the same time. So also an error in map matching. So just to give you idea of what can happen, and these have been problems we had very, very often at the beginning and we had really to work hard approximately half a year to get rid of these problems. So interim conclusion at this point, no, a hundred percent perfect technology, no, a hundred percent accurate data because it’s big data. And that means uncertainty. And of course the score also is not a hundred percent perfect. Yeah. After all, it’s just kind of a forecast and it can never be a hundred percent accurate.

And also important, there’s no perfect driver. And that’s what most people think, “I’m a perfect driver of this car, so this telematics solution should also be a hundred percent perfect.” But this is not realistic, and it’s just human. So this means a hundred percent is not achievable, but I would say we have a good… we achieved a good point now, but there’s also lot of things to do in the future. Okay. Remedy at this point can only be, stay calm, accept customer feedback, build tolerant systems and work hard to evolve.

We want to evolve, and therefore we have about 20 people in our customer support for telematics. And these people answer lots of questions. We analyze the feedback in the Play Store, or app stores, and we do market research, and so on, just to know how good we are or where we can do it better. I’m doing this at the moment and we have four stars in the App Store, in the Apple App Store. And I looked at this some minutes ago, it’s even 4.1. So I think we can be satisfied with that as a feedback.

Last chapter now, it’s about the effects and chances, and that’s, I think the most interesting part, especially for me as the actuary. So the first interesting point is that telematics is effective. It’s very, very, very effective. And first point is risk differentiation is really relevant. So I show you what this means. So we have here the score from zero to 100. And then we have the relative frequency of accidents.

And we see that customers with a lower score have nine times higher accident probability than those customers with a very good score. So nine times, and that’s very, very much. When we started with telematics, our first algorithm was manually calibrated, not using statistics, it was just manually. It was our opinion, what is a risky driving behavior, and that differentiated with a factor about three. And now we have claims data, we can do statistics and data science. And we now progressed to this. So we now have a factor of nine, which is very, very good. Second insight is that the 20% of the worst drivers produce about 40% of the accidents.

That’s also showing that telematics is differentiating between good and bad drivers. What is maybe even more interesting is that also the claims ratio differentiates by score. So this means when I consider the claims ratio, this means I consider the claims amount, divided by the premiums. And the premiums is the traditional premium, which is also based on the tariff attributes. So on the risk factors, I use age in that, we use the mileage in that, and so on. And so this also reflects your risk as a driver. So this premium should be… to a certain amount, should be fitting your personal situation, your driving style. But we see… taking this telematics score as an additional driving factor, as an additional tariff factor, we see that the score differentiates, and it differentiates more than a factor of two.

This means the score really delivers additional information. And this information goes far beyond traditional pricing attributes. So I think that’s very, very interesting. And at this point, maybe I could have even mentioned, so we are, or would be already in the position to build telematics algorithm without using classical tariff attributes and only using telematics attributes. And this algorithm, or this driving assessment then would be at least as good as traditional pricing. And that’s… I think that’s very interesting.  Risk selection works, it’s also an interesting point. I don’t want to talk too much about this. The point is when we start our product, so we get customers of type A, which are good drivers and deserve the discount.

They join our telematics program, and we also have customers who are bad drivers, but have little money, and also want to join our telematics program. And so these are those which probably will then produce the bad scores. Yeah, what we see then is that the good customers, one year later will stay, and the bad customers have a much higher probability to leave. So this means that the risk selection gets better over time. And yeah, seeing this, we can say, “Okay telematics works at this point, and it will be possible to really create telematics tariff which delivers profit, which is profitable. That’s possible, and yeah, that’s the good news. What else? So also interesting, using telematics, you as a driver, if you are using telematics, you have a high motivation to change your driving behavior.

And we have seen this in a little test we did, it’s not really a test. So we had about 100 test pilots in our test fleet, and these test pilots then changed to the real product after market starch. And doing this, their score improved very much, it improved from 57 to 70. And this just means their risk to have an accident reduced about 20%. And this means, probably telematics is a motivation to change driving behavior. What we also could do using telematics is this, we could use it to construct better roads and safer roads. So we did this for Coburg here. This is the town where I live. And you see here, these little blue points, and these are points where we observe many of harsh braking or some risky maneuvers.

If you look at this map, you will realize, these are… in reality, these are dangerous points on the road. And maybe such information could also be used to make the roads safer. Last but not least telematics is effective in the sense of sustainability, in the sense of ecology because drivers with a lower score consume 10% more fuel than those with a higher score. That means telematics also differentiates with respect to fuel consumption. And now I would say we’ve come to the end. Final conclusion at this point, telematics will need continuous evolvement. I hope I could show you this.

But I think already today, we know this… telematics is very, very effective with respect to risk differentiation, with respect to risk selection, and it’s motivation for driving better. And it’s also good for sustainability and ecology. Yeah. So I think that telematics offers a clear advantage to the customers. It offers savings, it offers feedback, and by this, you can increase your safety in traffic. So I think that’s a very, very interesting product. So after all, I would say, it’s a sensible, it’s an honest, it’s a good product. And with this, I would say now, telematics is a colorful new world in my eyes, very interesting new world.

And with that, I want to say, thank you for your listening.

Ben Telfer:

Thank you very much, Daniel. I would like to congratulate you on both an excellent presentation and also your driving score of a 100 and beating your boss! So well done.

We do have time a few questions. Anyone in the audience, if you’d like to submit one, please do. And if we can’t answer it today, we can obviously follow up with Daniel after this webinar.
First question for you, Daniel, “Do customers find the impact of safer roads and contribution to a cleaner environment a factor in choosing to use telematics? Or is it purely just about how much they can save on their premiums?”

Daniel John:

I think the most important point for the customers is that they can save premiums and the discount is what counts, and that is also the important point in their decision, I would say, after all. But I think it’s also important to consider these sustainability aspects because at least in Europe or in Germany, you have very much political discussion about sustainability and ecology. And so I think it’s very important also for you as an insurance company to support that. And I think this will be good for your image, I would say. Maybe it’s not the first decision for the customers, but I think that’s what… that will be very well be recognized by politics.

Ben Telfer:

Very much so and I think there will be a shift going that way, more so in the next few years. Like you say, with all the political activism and all the focus on sustainability and climate change going on at the moment.

Next question for you, Daniel, “What were your factors in how you chose your technology partners?”

Daniel John:

The factors, how we choose them, yeah, the factors where, we did a market screening to see which solutions were possible. We had tried with some service providers or technology providers and our first try was not very successful. So we had major problems with IT security at that point. So we had to cancel this experiment, I would say. And then we chose the next partner that was then Bosch which is a very large German technology provider. So the main important point was, at the beginning, that it really worked. So to become a good quality, that was very, very important. And then secondly, what is important is the topic of cost. So you need a… I would not say cheap, but cost is a big topic in that.

Ben Telfer:

Thank you, Daniel. Another question here, “Trust and transparency seems important in sort of developing an effective telematics solution. Does that give HUK an advantage due to your proximity to the customers, your customer base being a mutual insurer?”

Daniel John:

I think that, yes, that’s an advantage. So I think… I had a lot of discussions with EIOPA, with the supervisors about this product and also consumer protectors, and so on. And so at the beginning they were very critical about that. It was much about data protection and so on, but after all they realized, “Okay, it’s a fair product. It’s really fair.” So we don’t use any discriminating factors and so on, and it can be very fair. And that’s, I think a very strong point, which supports us as a mutual insurer. And so I think that’s important. Yeah.

Ben Telfer:

I think we’ve got time for one more. It just says, “Many global insurers have pulled out of telematics as a product but HUK has obviously been very successful. What do you think have been the key factors for the success of HUK’s solution?”

Daniel John:

When we first started, we had lots of ideas about ecosystems and so on, and you can use it to make advertising and whatever. And so what we realized then is, you can forget all about that, but the important point is really that you concentrate on this risk differentiation. So that was the important point, then important was to achieve good quality in all of that, because if you want to do assessment of driving behavior, you have to be sure, otherwise you will get many complaints. It’s about the money of the customer, it’s about the discount of the customer. If they lose that discount, just because of an error, they will be very, very angry.

It’s very, very important to look at this quality points. Doing this, we were able to really develop these algorithms, which really separate the possible last part I showed you. And I think using this, you can really develop a profitable telematics product, and that’s the important point. So you see you have a widespread in your portfolio between low score and high score, at this risk differentiation topic. You have a widespread, and you can use that to set up a profitable product. And I think that’s what motor insurance is about, about risk and selection, and that works. That’s the important point.

Ben Telfer:

Thank you very much, Daniel, for joining us today and for presenting and for answering those questions.

The above text has been produced by machine transcription from the webinar recording. ICMIF has made every effort to ensure that transcriptions are as accurate as possible, however, in some cases some text may be incomplete or inaccurate due to inaudible passages or transcription errors. Listening to or watching the webinar recording will allow you to hear the full text as delivered during the webinar but this is available in English only. Our transcriptions are provided to enable members to select the language of their choosing using the dropdown menu above.