Stay on the cutting edge of insurance telematics, connected car technology and digital innovation in insurance with our thought leadership transcript series for webcast recordings of past events.
Hosted by our subject matter experts this webcast transcript looks at:
Unlocking the Power of Big Data with Insurance Telematics
Dr. Ben Miners, Vice President, Innovation at IMS looks at how collecting driving data can derive new insights and services that control costs while improving policyholder acquisition, retention and engagement.
[IMS] Before I get too deep into what we're doing around big data and some examples of how we use the Pentaho platform, I thought I'd take a step back and help you with some context and just talk a little bit about IMS. We're a privately held company based in Waterloo, Ontario, Canada and we've really been focused on the connected car space and obviously we see this trend that's evolution from connected car, as kind of one of the first use cases of that broader IoT space. We operate in a number of different verticals and for the purposes of today I'm going to talk about one of our largest verticals around the insurance space and the insurance telematics space, and some of the challenges they have and how we use some of the technology to be able to address it. I'm going to take a little bit of different take on it and I'm also going to talk about how that data is applicable to other areas that you might not be expecting as we talk about it. So, without further ado I'm going to jump into it. I'm going to give a bit of a background in terms of what we're seeing as general trends kind of in the insurance industry and then we'll talk about kind of how we're helping to solve for some of those things.
We're very much focused on a concept called usage based insurance or insurance telematics using information about the driving behavior, the vehicle itself. When we look at what's going on in auto insurance industry we're seeing a number of things that are happening. The first has been this whole trend around insurtech, a combination of insurance focused technology companies looking at ways they can take big data, artificial intelligence other advanced applications and start to apply them in the problem of insurance. This is kind of that space that we live in, is somewhere in that insure tech telematics space where we're starting to see that transformation. What's really interesting is considering how data rich the our business is in and seeing these macro trends, with millennial's right up to older generations just that overall openness in terms of sharing data, you think about Google Maps, obviously there's this implicit agreement that you have where you're sharing your location information and you expect to receive results back in terms of navigation and traffic data and we see these things proliferating in all sorts of different areas. We're starting to see those things already emerge in the insurance industry as well and again usage based insurance is one of those examples. Now why do insurance carriers care about usage based insurance? Well one if they're able to use different factors and it rather than just your age, sex, the zip code you live in, the type of car you drive but also how you drive and using that as a predictor in terms of the expected loss that's key. That's why insurance companies are looking at this, now we've seen again there's a wealth of data I think we collect six trillion data points on a daily basis and we have to kind of make sense of all that information on behalf of our clients. So what we're seeing is a trend where we're starting to help transform it and how we address all that big data that's coming in there and how we can use things like artificial intelligence to be able to kind of filter through all that noise and looking at how we automate some legacy processes. In terms of doing it, there hasn't been a lot of innovation from an insurance rating perspective as an example in the last thirty plus years and telematics is one of those areas that the promise is to revolutionize it. We take a step back and again when we look at this how we can help a carrier, we're looking at four different areas claims acquisition, retention and engagement and I touch upon them really quickly the claim size and this is key up to 80% of the premiums that a carrier collects gets paid out in in expense of the claim, it typically ranges somewhere between 65% and 80%. Now just by having information about a potential claim you get in an accident and knowing that that much faster you have an opportunity to reduce that by up to 20%. So there's things that we're able to do to transform that process to help reduce the amount that's paid out is key and again so we have an effective claims management solution that is able to automatically do that to be able to filter those things out and help the insurance carriers reduce what's being paid out. This means more profit at the bottom line or more cost effective rates for policyholders or hopefully both. There's also a challenge we see in terms of acquisition and again the acquisition costs can range up to $900 where we're carriers are obviously still trying to protect their existing customers but they're also trying to look at how they can cost-effectively acquire new customers. That's how they grow their overall base, and again with that, that cost what we're seeing is things that telematics because they provide better insights into how you drive. They can start to look for opportunities where they said you know what Ben is a really good driver but maybe he's not priced efficiently, now we can offer him a more aggressive price because he has the type of risk that we're willing to underwrite, and then offer me an attractive policy. The more aggressive carriers are able to offer tools like a try and by mobile telematics application that allows them to acquire new customers or to create new applications in the whole vehicle sharing space to be able to offer maybe under surface segments.
Obviously retention is critical and one of the interesting stats we found is by using telematics and providing that feedback to drivers we see an 18% to 20% increase in retention and engagement. Today auto insurance has a real challenge where the consumer only interacts with their carrier a couple of times per year, and if you think about that in your own lives, engagement is typically upon policy renewal and hopefully not during a claim. When you have that level of engagement you have a challenge there where it's easy to undervalue the service that your carrier provides and say, “you know what, I can get 20% cheaper or $100 off or some other cool feature with a different carrier why don't I switch? I’m really not that emotionally tied to that particular carrier.” Again as we see with telematics when we provide telematics to users we're able to see when somebody's checking their app and they're trying to figure out how to be a safer driver. All of a sudden we see this uptick in terms of two things, one that overall engagement level and two we see that they're actually safer and safer means less risk. As a parent I'm obviously concerned about the safety and well-being of my family. The carrier is concerned about that but they're also concerned about the bottom line and making sure that they're not paying out unnecessarily on claims and if they can help their policyholders improve that then that's key.
The other part that we see when we talk about that engagement is through somewhat telematics can do is really run behavior modification. Again there's all sorts of statistics and this is where we take that wealth of data and we look at it over time and we start to understand how we can overall improve the lives of our policyholders and the bottom line for our carriers. What we've seen is through the use of telematics after a four week period that we see an 89% improvement in the overall driving behavior. We also see either an improvement or maintenance of that safe behavior, so reinforcing those very good behaviors. Again speeding is the one that I think we're all can be guilty off, and again through awareness and making sure people aware of it that's one of the areas we see the greatest benefit in terms of the overall scores. So that is a bit of a backdrop I thought I'd shift gears a little bit and talk about what some of that big data we're collecting and what are some of the insights that are were able to drive. One of the challenges we have and in the insurance telematics industry there's a variety of ways we're able to collect data.
We can do everything from integrated to the latest car that provides a very rich set of data on the right-hand side of the screen right over to the left-hand side where we can actually have algorithms and technology resident on the consumers mobile phone and using those to infer driving behavior even who's driving the car and there spectrum in between there and they all have different merits between them. I'm not going to get into too much detail on this we actually have a whole white paper that discusses the merits of these different areas. I think the key here though is these are some of the challenges we have when we talk about collecting data. It's not a common set of data that we're collecting, there's not this homogeneity in that data and we kind of have to deal with that when we're looking at it and we're trying to apply rating algorithms and be able to understand what some of the implications are.
The other part that which is critical is around security and with some of the new legislature that's going in place in Europe in terms of what the obligations are around there some of our customers are very much focused on servicing very specific segments. We have a carrier that's focused on servicing the families of US military personnel and so obviously you want to make sure that information is absolutely secure. Even though we had lived in this era where we're willing to share information, it's absolutely critical that that we treat it and we're entrusted with that data and we protect it accordingly. One of the great things we've been recognized as kind of being head and shoulders above the competition in terms of where we are and how we handle that security, how we treat that when we're entrusted with that information. That's interesting that's critical but it also is a challenge for us when we talk about wanting access to that data and so we have these debates internally about how we can leverage the big data that we see here and how do we do in a secure way that kind of make sure that we're compliant with all the regulations and laws and rules and the obligations that our customers and their policyholders put upon us, and so it's an interesting challenge that I think many on the phone will be able to relate to in terms of I have all the data but how do I unlock it in a way that keeps us compliant.
So I'm going to step in a little bit again talking a little bit more about the data and you know this is a modified version of what you've probably seen before in terms of data insights and we all try and move from the left-hand side to the right-hand side certainly that's our goal as we best continue to invest in insights and analytics. When I talked before about the different data sources, that riskily points to that cleanse data. If you think about the difference between a car reporting information and a phone, the difference is the car is always reporting about the car, the phone is with me and it follows me on the train, it follows me on the bus, it follows me on my bike and of course in the car, whether it's me or my spouse driving. That has some interesting implications the other implications are of course I might be using an iPhone someone else might be using an Android phone they might be of different versions different equality levels and being able to normalize that information across data sources and data types is critical so we invest a lot in the data cleansing. We've kind of got that level based piece but also embracing some of the diversity in the data, for example on the mobile all of a sudden we get interesting insights about how people travel, do they use a mix of car and public transit? We can get that information through a mobile based program. Again our goal is really to move to the right-hand side in terms of prescriptive element to it and again so this is where we're able to say, “hey not only is Ben a really good driver, but this is what you should do carrier about that.” Or, “it looks like Ben has had a new driver added to the policy, there's a new member in the family,” that so happens to be driving where you have the ability to detect that more than just being able to diagnose and predict that or diagnose it, it's about what do you do about that. Triggering an event to my agent or an outbound call that now it says, “call Ben and confirm that he's got the right policies in place maybe something's changed, life has changed,” and all of a sudden you want to be able to go hit him with an appropriate message that says it looks like there's an additional driver or just maybe it's a feeling call in terms of knowing if anything changed. It's an opportunity to review that policy it goes back to that engagement conversation that I mentioned at the beginning, where this is driven by data it's an opportunity for the carrier to stay relevant in the lives of their customers.
Now specifically, and again shameless plug here for Pentaho in terms of how we incorporate it, we've embedded that the business intelligence suite within a specific tool within our solution, what we call our business center and business center is really that single place where our carriers, their support staff, the people responsible for the program, the actuaries, the claims adjusters all go to the central portal to be able to kind of consume different data. A few years back we took a step back and we wanted to figure out what we wanted to invest in and how we surfaced this up and we chose the Pentaho suite embedded into our solution and it might be a little fuzzy here in some people screens but there's a whole wealth of reports here that we then are able to expose. A great example of how we had some previous reporting kind of needed reporting in our application but a great use cases we have customers been working for a number of years, we had a report that showed how often somebody unplugged a device that's installed in their vehicle. This report had been available there for some time and when we're modeling out the effectiveness of some of these tools report of that report into our BI component. That BI component to the customers allows them to very quickly look at some outliers. When they looked at the outliers they found that there was one specific vehicle or one specific policy where they saw this huge frequency of disconnects and connects. Somebody was unplugging a piece of technology we provided. In talking with this carrier and their staff there they couldn't figure it all out and they decided actually reach out to that policyholder and they talked to the policyholder and he talked to his spouse and they couldn't figure it out. It was when they talked to their teen son that they then realized is the teen son would get the car for the weekend, he got the car Friday night, would reach under the dash and unplug the device and Sunday morning he plugged it back in. That's a great example where they had access to that information it was a bit like finding a needle in a haystack and just through some visualization they're able to identify that. In this case with the policyholder, the carrier could have a conversation with a policyholder and allow them to remain engaged in the program. They need to make sure that they are keeping that device plugged in, obviously if it's going to the garage there's abnormal circumstances that's fine but by general course that should be remained in there. All of a sudden there's an opportunity for a little bit of education between parent and teen and that situation result was resolved. I think the key there is we have that data, it was just really hard to find that it, and so that's one area where that the tool set helped us kind of service that up now.
I'm going to talk about two different areas now. The first is around distracted driving and when you look at over the last 50 years, the general trends on a constant mile basis that fatalities are going down year-over-year for the last 50 years, and it's only in the last recent years that we start to see that increase. It's not because cars aren't safe or certainly with adaptive cruise control or accident avoidance technology, the vehicles themselves are safer more than ever but what is occurring obviously is distracted driving. We've been investigating and investing in this space for a number of years and wanted to pull out and surface some of some statistics here, these might be a little bit alarming, but again the data doesn't lie. The first thing when we look at a cross in aggregation that we saw that 53% of trips have some element of distracted driving, and when you boil it down because sometimes that's actually occurring when you're in a stoplight or you’re pulled over you did the right thing. There's still 36% of the trips that are occurring while you're moving and in furthermore the average element of distraction is 94 seconds long. That's a minute and a half, and you can imagine at highway speeds even a one or two seconds of distraction the results could be catastrophic. So we started the surface of data up and then we started thinking about with our data scientists and our product folks in terms of how we address it. What we've done is we've categorized distracted driving into several different levels. The first level is at a cognitive level and you can think about that when you're having a really emotionally intense conversation you're very much focused on that conversation, and not necessarily what you're doing behind the wheel. The other scenario is you drove to work and you can't remember the last three stoplights you went through, your mind wandered and that's where you have that moment of panic and think did I go through a red light? I know I didn't think I did because I didn't hear any car horns. You had that situation there's that cognitive distraction whether it's active or passive. The second level is that physical distraction. Maybe it's because I'm holding up the phone to my ear or I've got it on speakerphone and I'm talking. Maybe it's because I'm actually physically typing into the keyboard. The third one is the visual distraction. When I'm physically taking my eyes off the road and looking at the phone, now there's limits to what we are able to detect. Sometimes we're actually inferring some of those behavior sometimes we're actually getting it through some predictive algorithms on the phone, but we characterize those in a number of different levels in terms of low medium and high. Certainly there is a cognitive load in terms of having a conversation even if my phone is paired to the Bluetooth system and I'm talking and but again, that takes an element of load that is not focused necessarily on the vehicle. The medium one is where again, I'm holding the phone and it means I don't have my hands on ten and two on the steering wheel. Then of course the worst scenario is where I'm looking at or responding to a text message or searching up directions or something else so we kind of categorize it different levels and we service that back up to the user or back to the carrier and then the user again. The goal is to inform and hopefully correct some of those behaviors so that they understand what that potential risk is. Looking at some additional statistics here, at the top when we look solely at phone call types of events, and this is a subset, just while moving. On average consumers are spending more than four minutes on the phone and again there's a cognitive load now the good side is longer and duration. The flipside side to that is, there's a lot less trips we talk about relative to phone handling that occurs but these typically happen at a longer duration so less frequent but longer duration. Again these are lower typically in severity, in terms of what they are, the phone handling ones are the ones that really give me cause and pause where an hour on average it's almost a minute of distraction again while moving. What we're seeing is 28% of trips have some element of phone handling distraction at high highway speeds 1 or 2 seconds of inattentiveness because you're typing on the keyboard or looking at the screen. Typing on the keyboard typically requires your hands physically on the phone as well as your eyes on the phone, is the highest severity on it, so these statistics are very troubling. This allows us to have the data to start to do things like creating scoring algorithms, providing alerts back to the driver, notifying a parent that their teen is doing this. There's a variety of things that we're able to do from the awareness perspective and then of course working with our carrier partners the other things we can do is provide incentives for encouraging the right behavior or kind of reducing a discount or a refund or award if they’re exhibiting the negative behavior. Having that data makes us able to kind of surface that up to us and then certainly take action on them.
Now I'm going to talk about something a little bit different and I alluded to it at the beginning and so far what I've really talked about is very auto insurance centric. Now there's been a couple of events that happen in the US that allows us to exhibit some interesting information that's not specific to the auto space. Now certainly I'll draw some parallels there, but in in in late August the Greater Houston area was hit by Hurricane Harvey and then just a few weeks later in early September, Florida was impacted by hurricane Irma. We have a wealth of data that starts to allow us to make insights around what occurs and so the first two graphs we have here again on the left is Harvey and then on the right is from Hurricane Irma. You see in the green for example, the average trip activity. You can see that there's some daily variation kind of day to day kind of leading up to when the actual hurricanes hit, and then in both in red which is the wind speed and then in blue, the precipitation you can see that the vehicle traffic dropped off and then it starts to rebound afterwards by the end of August for Hurricane Harvey. We saw this similar trend happen to Hurricane Irma, not necessarily revolutionary in terms of it but again we started to look at other types of behavior and this is where we start to really segment some of that data out in terms of looking at it.
I'm going to share with you more of a complex view of this, the first one here if we look at Harvey, we can start to look at what occurred right which number of vehicles were in motion being driven for example before the hurricane that were no longer driving afterwards. You can see that there's like 27% of vehicles that were active kind of pre-Hurricane Harvey that were no longer active. That becomes kind of an interesting insight in terms of is that an indication of loss or is it because people had left their vehicle. The after period we didn't measure that for months and months afterwards, it may be that they were living with friends and family and they hadn't actually returned to where they left one of their two vehicles but it starts to give you a kind of an indication. You can also see here just again staying with Harvey up to 15% of those vehicles were still in active use, both before and during the disaster. It could be because they're trying to get to evacuation centers or trying to get out of the path of the hurricane as well as the aftermath. In Irma we decided to look at the data a little bit differently because there's this mandatory evacuation period and you can start to get these insights in terms of before and then during the mandatory evacuation period what happened. The number of people who continue to drive during Irma for this example, we don't have the statistics on the post period because what we really wanted to zero in on is something a little bit different. We wanted to look at, for example, how this data could tell you our mandatory evacuation notifications are they valid and they drive value around them. This is an area where IMS or an insurance carrier may not be directly applicable, but it starts to have some really interesting insights if you're responsible for disaster planning from that perspective or you're in a different part of the space in terms of understanding other insights you might be able to glean from it.
Taking a slightly different perspective on it looking at the actual driving behavior and each one of these dots represents a set of trips, again this is the data from Houston and we see before there's a kind of our baseline in terms as we see that kind of driving behavior and seeing it kind of plotted on a map and you start to see what happened during the hurricane. Obviously we see a great drop-off, kind of similar to what we saw on that Venn diagram in terms of behavior but you start to see what really occurred here. Then you start to see after as well and I don't have the overlays here for it but if you start to overlay for example where there is flood damage, you can start to see those patterns of behavior where there are a lot less driving trips in those areas where that there's a flooding and flood activity. It also starts to give you an idea in terms of where people are going, where they might have gone if you look at origin and destination information which is not the visual representation here you start to understand where people go to seek refuge. Were they going to the other areas outside of the Galveston Houston area or did they stay put? Again, we have all these insights, we have that wealth of data that not necessarily relates directly to an auto insurer but also much more applicably in other areas. Two weeks after Harvey you see this resumption of travel that's kind of the period where the flooding has largely subsided and you see things restore back to almost a normal level just prior to the hurricane.
Looking at the scenario with Hurricane Irma, Irma is a little bit different. The data we've got here shows before and then during the evacuation period and kind of following that, the other challenge there of course is the predictions that they had. They weren't sure if it the hurricane was going to make landfall on the east side of Florida and hit Miami or the path that it did through the Florida Keys and up into the west side of Florida. Before making landfall in Florida, we see this kind of driving behavior this is kind of the base case. Here we see during that evacuation period, pretty consistent with not a lot of drop-off of some of that behavior in terms of driving behavior. During the actual hurricane we see a big drop as people either got to their destinations or those that chose to try waiting it out in terms of where they were. So you see this mix in terms of statistics and how we're able to surface up the data we're kind of very fascinated with in terms of looking at that, and then what that can tell us in terms of other areas taking a step back again in different area the insights were able to provide looking at. For example, my vehicle activity, I'm a three driver two vehicle household and an insurance carrier if they're able to determine that one car is static sitting in my driveway and the other one drove down in this case, a family trip to California, then all of a sudden that becomes a really interesting insight where a vehicle is now gone on the proverbial road trip and the other vehicle static. Does that now mean the exposure on the home side is different because now one vehicle has stayed and another one is gone? It looks like it can be predictive of certain behaviors like I'm on a family road trip and that might be a drive to prompt a communication back to me, the policyholder, and say, “hey make sure friends or family have checked in on your home to make sure that that leaky faucet didn't lead into some sort of flood damage that maybe you're not covered for.” So those are the type of insights that we look at and that we're able to kind of gather not just driving specific but that impacts a home or in this case in the case of hurricane Harvey and Irma much broader in terms of predictive around it. Then of course you can see here the resumption of that overall driving behavior again looks very similar to what we saw at the beginning of or prior to Hurricane Irma. There is as I mentioned, a white paper that talks about the different data sources and the pros and cons to it, we have a wealth of information that explores these areas and then I'll leave that up here again. You can go see our page and find some of those white papers if you're in the telematics space. Furthermore, even if you're not necessarily directly in the telematics space it may be worthwhile taking a look through that information because those insights we talked about, the challenges we have are very similar right now. Obviously they may not be applicable to a vehicle but we talked about IOT and the wealth of data and the cleansing of the data across different sensors there may be a high degree of applicability where you're able to kind of take lessons that we've learned and apply them to a different problem space. I think that's that, we'll pass it back over here for the Q&A session.
[Pentaho] Great thanks so much Dr. Ben, let's see, it does look like we had a couple of questions come in and then if anyone does have any questions at the moment free to enter them into the questions panel, we'll try to get to them. I know we don't have too much time but we'll try to get to them as fast as we can. The first question is pretty easy, I can take this, is it possible to get a copy of this [webinar]? Yes absolutely, we will be sending out a recording of this presentation tomorrow, I believe, or by end of the week. You should receive a copy of it if you're registered. It does look like we have one or two other questions, so for one of them, I'll pass this over to you Dr. Ben as you are evaluating technology partners. What key criteria were you looking for?
[IMS] Great question, I think there's a number of things, one we wanted to make sure that in general we have to deal with scale security. As I mentioned, the scale security was absolutely critical on it and we wanted it that way because we have this wealth of data that's coming in from all sorts of disparate sources, we were very much looking for a solution that allowed us to kind of pull that in from those different areas where it may not be data that we collect directly. Whether that be traffic data, would be a good example of data that maybe we’re ingesting that isn't native to what IMS collects so I think those are a couple of areas which scale security and the ability to ingest the kind of variety of data that maybe isn't necessarily resident natively within our environment.
[Pentaho] Great! Thanks Dr. Ben. It looks like we had one other question which I think would be great for the rest of the audience. What challenges did you face when you were coming up with the IMS platform?
[IMS] Great question. I think the challenges are on data cleansing. It certainly has been a challenge, the industry has evolved from having all kinds of devices very much tied to a vehicle to these plug-in devices to mobile technology and now the pendulum has kind of swung to where there's a variety of different sources and types of technology that are being used. Being able to deal with that volume and being able to normalize and cleanse that data is one of our big challenges there. Of course the real challenge is collecting the data and is kind of interesting but transforming that into those actionable insights is critical. Again, whether it's on the auto side or they're broader in terms of aggregated driving behavior across a particular highway corridor. Though that might be really interesting to the government agency responsible for maintaining that data or investing in it, those are the type of insights that we seek to unlock with these investments.
[Pentaho] Great! Thank you Dr. Ben. This next one I'm not sure it's a little bit broader so I don't know if you want to take a stab at it we'll get a try regarding on usage based insurance. It seems the current solutions in the market are insurer specific like driving scores. Is there any trend towards standardizing it?
[IMS] That's a great question! I actually love this question. So what we see here is, if you're a large carrier you don't necessarily want to standardize that scoring, you've invested in actuarial skills you believe your models are more predictive than your competitors are and a standardized score doesn't necessarily benefit you. If you're a broker and in some markets brokers are much stronger, in Canada in the UK for example, there's a much bigger part of policies that are written by brokers. If you're a smaller carrier, you're a value play, then having standardized scores may be more attractive to you. As a smaller carrier, you don't invest in advertising. You wait to level the playing field against some of the big guys and that'd be more of a fair arena. So I see this as an area that carriers are kind of well if we do that are we just commoditizing ourselves around that and then it becomes a pricing discussion or is there still some layer that we can apply on top of that that is still more predictive, so I think that's still premature. I know certainly there's been a lot of talk about that, is there going to be a driving score similar to like a credit score? I'm not sure we're necessarily going to see that, just the differences of data sources a phone versus something tied to your vehicle very considerably and the strategies around how you use that information vary considerably.
[Pentaho] I am going to take one more question before we end this and I think I know the answer to this one, but I'm also going to pass your way Dr. Ben. Does IMS support new OEM vehicle manufacturers collecting? I think it can be data is analyzed for similar usage in sites versus servicing just insurance companies? So I think the question is do you support OEM vehicle manufacturers collect collecting data?
[IMS] Yes, the short answer is yes absolutely. We work with OEMs to collect that data and what's interesting is and obviously the person raising the question has some familiarity with it. We see a big disparity in terms of what's available. Data that is coming from the vehicles is sometimes fairly limited. Think about just getting your odometer value and your VIN, some OEMs are surfacing up a lot of more rich data and while we spent a lot of time in this webinar talking about what we do from an insurance telematics space we actually operate in a number of different verticals and having access to that data is really relevant. For example, we operate in road usage charging programs in a number of states. Having access to odometer values is critical in terms of being able to provide that. We're able to do this by having access to the CAN bus data which is really valuable when we work with some of our partners in terms of transforming something that's going on with the vehicle back to a dealer, a carrier, the policyholder themselves. Those kind of insights transform a bit of data into something that's actionable. For the policyholder, the end user or the vehicle owner being able to do that is valuable. So, the short answer is yes, we absolutely can, do and collect that information and look forward to the increasing wealth of that information and we apply it not only just from an insurance but across a number of different verticals that we participate in.
[Pentaho] Thanks Dr. Ben, I know I said only one more question but a really good one came in that was related to that in terms of being able to blend different types of information. Do you also include environmental or contextual data in your scoring models such as weather data?
[IMS] Absolutely yes, so contextual data, whether it's information related to the roads or flow of traffic, we had the opportunity to enrich our programs with all that data and again sometimes for example weather data is a great example when we collect data about a vehicle we're able to get it at a very high frequency sometimes sub-second intervals weather data you might be getting at it at a quarter of a mile radius or at a higher level certainly there is data sources that provide the temperature of the road precipitation. In some sort of micro context, those data sources get richer and I would also say more cost-effective, we absolutely enrich them in our data sources. I'll use speeding as an example, the limit might be 65 miles an hour and you're doing 70 miles an hour. Is that unsafe or is it unsafe relative to the actual traffic flow at 72 miles an hour? Well you know what, maybe we don't penalize the driver because of that is a great example of how you start to contextualize the actual driving behavior in different ways.
[Pentaho] Great well this was this was a fantastic presentation. Thanks so much Dr. Ben and thank you so much to the audience. I do know there's a few more questions we weren't able to get to, and we'll follow up by email. Really appreciate everybody listening in, thank you so much everybody.