Nikolaus: Hey everyone, welcome back to another episode of Reinventing Finance. As usual, I'm not all by myself. I have the great pleasure of having Eliron with me today. How are you today?

Eliron: I'm good. Thank you. How are you?

Nikolaus: I'm very good. Thank you. I am excited to speak about, you know, what we are going to speak about, about the company, about motor insurance, about how to make it a little bit better. And maybe a good place to start would be, why don't you briefly introduce yourself and your company, and then I think we can take it from there.

Eliron: So Ravin uses technology to remotely assess physical condition of vehicles. We use mainly computer vision and deep learning algorithms to determine damages and actually calculate repair costs for both fleet companies, car dealers, and insurance companies.

Nikolaus: What would be the macro problem that you're solving with this?

Eliron: Look, I think there is, there is a problem of transparency around motor costs and repairs around the world, particularly what we've seen in the insurance industry in recent years is a dramatic inflation of repair costs, which push the motor line for those insurance companies, which has already been challenging to an even tougher environment. We see price increases, we see, you know, Germany motor insurance, CPI, I was just reading a report by Swiss Re saying it was up 23.4% year on year in March 24. And I think a lot of that is driven by the repair costs and actually less transparency around how do you repair a vehicle? What kind of processes do you take place in the background, which we can help pretty significantly.

Nikolaus: Out of curiosity, what drives repair costs and is there some learning for me as a consumer? Maybe there is a difference whether I am insured or not insured, you know, it's because, it's like a black box. I have a guy who I go to and I trust that guy, but I have no idea about assessment and the only reason why I trust that guy is there are actually some family relations. I have a really old BMW, really old, could be a classic car if I treated it well - but which I didn't. And I had some some damage to it. I went to BMW and they just said like 4,000 euros or something. I then went to this guy and he said, listen; you don't need to replace the whole thing. We can, we can kind of repair it. And it was 400 euros or something ever since he's my guy.

Eliron: You are pinpointing exactly the right matter. And, um, the, what you've described now, the difference between what this guy quoted to you, which was probably in the hundreds of euros versus what W quotes, which was in the thousands of euros is essentially what we see on the balance sheets of insurance companies today in the billions and trillions of dollars, actually. So, you've asked what drives the dramatic cost increases, obviously macroeconomic problems like shortage in parts and labour, the crisis in Ukraine, the crisis in the Middle East, um, disruptions to the supply chain that all has a burden on supply of, uh, materials and parts for vehicles. There is also a shortage of technical labour in body shops and insurance companies. So there are less people to do the work and that puts a lot of pressure. And then in the post-COVID, uh, era, there are also actually a lot more accidents happening. Surprisingly, you wouldn't expect it, but, um, yeah, I did, I did read a report. I think it was CNBC that, uh, showed an increase in collisions, and because the cars are more expensive, and every collision is also more expensive to repair. There are a lot more components that get, uh, that get hit. Now, once you look at the problem, if you look at the, you know, the, the average repair costs and you put next to it, the logistical costs, the logistical costs start when you have an insurance claim and you just drop your car off at a body shop. Once you do that, only then the insurance company gets involved. It may receive some images and reports from the body shop, but then the ability for insurance companies to actually control the process is very limited. And the expenses start rising from that point because many of them give you a courtesy car for replacement. There is already a tow truck that needs to be paid. There is already somebody at the desk of the insurance company that needs to now process the, uh, the case. And so the admin costs also start rising. And the question you ask yourself, does it have to be like that? Is there a way because these claims are more frequent and repetitive to automate those, uh, those claims and approve repairs more quickly and, on the way, also control the cost of repair more effectively?

Nikolaus: So we've had a look at German figures. I have a bit of a closer or German use cases, and I hope I don't butcher them with this, but I've, I've heard that one of the leading car insurers, they actually set up their own proprietary network of garages, which they own, thus they drive costs. They are actually providing that as a service to others, e.g. to their competitors that they can get even better, um, margins, but ultimately it's used to decide where you go, to make your repairs. I don't know whether that is the same in each country, but you need to decide whether you have kind of free choice or not. And if you have free choice, then it is for example a higher premium. The other thing that I thought was quite interesting what you've mentioned is that they do actually want to get, ahold of the process really early on because it minimises cost because they can minimise the cost of the courtesy car. They can minimise the cost of getting it to the right repair shop. They get it potentially to the repair shop which potentially has the right materials, et cetera, because the cost clock starts ticking for the insurance company. So there seems to be a strong kind of incentive to own that supply chain. How do I now solve this process? Um, where do you guys come in? What's your take on the problem chain?

Eliron: So I'll give you an example from one of our projects. We actually work with the largest insurance company in Australia called IAG. And when we looked at the breakdown of their claims, , we identified that about 10 to 15% are total loss vehicles. So they will never actually get repaired. And then another 10 to 15% are just too small. That's on the other side, too small to be repaired. And if you manage to grab that information at the point of, um, the first notification of loss from the customer or from a tow truck driver that comes to the scene, just determining that decision and avoiding the car getting to the body shop is a major saving. You are talking about anything from hundreds to thousands of dollars per case.

Nikolaus: Ah, okay interesting, yeah, It's, it's starting it right at the beginning, right? It's, it's, um, okay.

Eliron: Yep. Now, you know, once you get those images, then you can make the decision. So you fast track, um, those two cases on the other side, you know, either a total loss or too small to claim if you like, or some sort of cash settlement. Um, so you go to your guy, instead of going to BMW, just that saving is massive. And then if the car does end up in the body shop, at least they have our...

Nikolaus: What does massive mean? Um, do you have a kind of...

Eliron: 30% lower. So you, if you paid, uh, we have cases where the insurance company would have paid, you know, uh, a thousand euros and they will now only have to pay 700 euros because they've settled it immediately and they don't have all the overhead around it.

Nikolaus: Okay. Okay.

Eliron: So this, then these numbers add up. Now, um, some cases will end up at the body shop. You will say, look, according to Ravin's repair estimate, we think this car does need to get repaired, send it to the body shop. Even then we create a benchmark for them to approve the body shop quote more quickly. So the body shop doesn't need to wait for approval. You don't need to wait for your car. It all becomes much more effective. The car gets to the body shop already with, uh, a Ravin supported, uh, approval of, of the repair and the body shop can just carry on with the repair. So on one hand, its efficiency, on the other hand, its transparency.

Nikolaus: So I think, and on those ones how easy is it for you to localise if you need to adapt locally? I would say: “ great, this works in Australia. Does this also work in Germany or Switzerland?”

Eliron: Yes. Actually it’s a great question. We've started working with insurance companies in Germany. And what we realised is that about 80% of the parts and the materials are actually quite similar to what you see in, uh, other countries like the US and Australia where we operate, uh, because there is a global supply chain. There are some unique models of vehicles that may only exist in Germany or the EU, and then you need to adopt the local operations.  And we have the right partnerships in every market to do that kind of last-mile adaptation.

Nikolaus: So if you kind of take a step, kind of triaging right in the beginning, um, hhow accurate are you on estimating certain claims and, what's required. Iit might very well be that an insurance company is willing to share data with you, if that saves it, but just kind of how would that process look like and maybe let's take that process to say “okay, we want to enter Germany or Switzerland, whatever”. what would that process be like? When could we, could we get started on, as you said, 80% straight away, how much lead time would you have? What would an ideal process look like if you wanted to, let's just start with on-board an insurance company.

Eliron: Yeah, there is a little bit of an adaptation period, uh, which we are, uh, we have already started doing for Germany, for example. So, uh, any insurance company in Germany, if you talk to me in a couple of weeks, uh, they will be better placed to actually start benefiting from the solution today. Um, and that means that our accuracy is above the 80% mark. It means it's just good enough for those clear cases. And then it gets better and better every week that we, that we work together, um, in order to adapt the model we've seen it's kind of in the hundreds of cases, not thousands of cases.

Nikolaus: And at which point does it make sense? What's the kind of normal engagement at which point does it make sense for you to, um, have a chat with an insurance company? How, how do you go? Is there like a pilot? Is there minimum engagement? You know, what would, what would that look like? How do you, how do you, how do you, how do you crack the nut?

Eliron: Yeah, there is, there is indeed a pilot period where they just try out the solution. The solution doesn't need much integration. It can work pretty much out of the box. So you are talking about a web application that you can send a link to a customer that reports an incident or a tied agent. Many insurance companies in Germany have agents that work with them. So, um, one of them, either the customer or the agents, um, they will get the link to the web application. They will do the scan. We will deliver the estimates to the insurance company. And there is a motor assessment team within the insurance company that can look at our estimate and verify that it's fit for purpose.

Nikolaus: Okay. So are there any other European countries that you're currently looking at for anyone listening? Or would you say, you know, Germans call me up in a few weeks - French maybe give me a few months.

Eliron: No, we are, we are open to most countries as we've seen the adaptation between countries is not as big a challenge as people think. So definitely open to any country. I think we are definitely stronger in, uh, what you call the OECD countries or that developed economies, uh; they are the ones that are experiencing a lot of the challenge. Uh, that's where we, that's where we focus.

Nikolaus: Okay. So we've talked kind of in terms of use cases, insurance claims, we have the triage we can. So, but it's effectively, it is about, it's reducing the admin is, is that the key to it? It's really reducing the admin part of the insurance claim because you,cannot fix labour shortage. You cannot fix, um, disrupted supply chains - is that the lever that you take? Or is there something else where you have claims accuracy? You know, what else is there on the insurance use case?

Eliron: So I would split the benefits to logistics and then to the repair itself. Okay. Logistics is about avoidance of, let's say, towing the car and then discovering that it's total loss and you need to tow it again to the salvage or the opportunity to settle a claim with cash to a customer. That’s the logistics part and just automates processes. But there is a big chunk of benefits in simply choosing the right repair method. So we have seen, I've just seen an example the other day of a, uh, a truck that had a few scratches on the door. And then the body shop said, I need to replace both the door and the surrounding, uh, panels because this is what I feel is right. And it was approved for repair. Now, another professional, let's just take your guy because we started with it. Your guy will be fair and say, look, I can just paint it. And by the way, I have exactly the right paint. I don't need to order, uh, something else it's approved by the OEM. So it still meets the regulation in Germany. It's, it just happens to be 50% lower repair costs. So it's not just a logistical benefit. It's ultimately choosing the most cost-effective repair method that is compliant with the law.

Nikolaus: You've also mentioned, um, fleet management and I believe on your website you mentioned sale of used cars. Can you kind of walk us through those use cases? How they compare and contrast?

Eliron: Absolutely. So fleets and rental cars, uh, our main use case there is that we can detect damages happening before and after the rental. So we give the customers a transparent view, but also the car rental company can detect if there is incremental damage to the vehicle and they can charge their customer. Also there, you know, it's very important because there are a lot of transactions happening. It's important to be efficient and automated. So focusing only on the incremental damage, analysing it, estimating the cost of repair, and then asking the customer to participate in the cost, not just randomly telling them, hey, you damaged the car as sometimes happens. And then you get this letter and you, you know, you start arguing with them, which is also not pleasant for anybody. So we are actually doing this today, uh, especially in the United States with a major car rental brands. And, um, you, we are talking about volumes of hundreds of thousands of cases per month. And then if I go to the used car space, uh, eventually a lot of fleets and dealers looking to value a used car. And part of the valuation is adapting damages on the car to the value. So let's say you have, you are going one day to sell your BMW maybe. Um, and then you will go to a dealer and they'll tell you, Oh, look, it's a great car, but Hey, you know, mechanically it's fine. But of course, I need to deduct 500 euros because of this scratch and 300 euros because of that scratch. And gradually they will peel off money from, uh, from there. Right. So here as well, we are providing the industry car sellers and car buyers, more transparency, giving them the opportunity to scan the car, analyse the damages and really giving them, a monetary value.

Nikolaus: I don't know whether the frequency varies like lease car leasing companies, when you kind of give your car back. I obviously have only had one car so far,which I've bought at the time. Um, so I don't know the process, but a friend of mine, she had a bit of a bad awakening, um, when she handed it back to the car leaser, cause something summed up, they deducted a lot of things. And then she had to pay, I think like five grand or something, uh, afterwards for various reasons, some internal, some external. Okay. But you're nodding for anyone listening …

Eliron: …Absolutely. Look the process is broken. Let's just put it out there. Leasing companies use third-party inspection, uh, people whose job is to find as much damage on the car as possible. That's what they're hired to do. If they don't find it, then they don't get paid. And so that doesn't always result in the best customer experience, let's say. And it's not always in the interest of the leasing company either, because eventually if the damage is there, they need to, they need to repair it. They need to report it before they actually sell this car onwards. So once again, if you have a tool that is able to capture the condition of a car, maybe before it gets returned, then you as a customer have the opportunity to push back and say, look, this is damaged. Uh, I'm going to repair it, don't worry about it. And, um, and this one, you know, uh, this one is not really damaged. This is actually just, uh, just, uh, dirt or something. And then you have the opportunity to take care of the car before you get charged.

Nikolaus: But you don't have a business cut-to-consumer application, right? You're pure B2B. And so the leasing company might not, might have an interest if they're looking at it transactionally to have a really large after the kind of return invoice.

Eliron: The leasing company will ask us to send the link to a customer, just like an insurance company, send a link to a customer. It works pretty much the same.

Nikolaus: Uh okay because the leasing company then has the body shop who will then inflate. Okay. Understood. Yeah, yeah, yeah.

Eliron: And they have customers bringing the car back. So our solution helps them to see the car from the customer and then tell the customer, look, you are on the hook for five grand, but you can bring it to our body shop and repair it for less. They will benefit because they get some repairs done and they will benefit from getting the car clean versus now repair it when you return it. And you also benefit because you save on those annoying charges.

Nikolaus: Okay. Yeah. Makes sense. And that's okay. That's sale of used cars. Any other use cases we haven't, you know, we should dive into?

Eliron: No, I think I'd like to go back to maybe a question that some of our listeners will have. We clearly haven't invented this space, right? People have been trying to adopt technologies for triaging and what is called straight-through-processing, STP industry. And I kind of, you know, the saving is very clear. I mentioned that 30% is the magic number you can look to save per claim. Why hasn't it happened? You are talking to insurance companies, especially in Germany. And there are very few cases, maybe between one and 5% at most that get automatically resolved. What is the reason for that? And I think a lot of people say that the technology is not yet good enough.

Nikolaus: Sure.

Eliron: I think, you know, we see three reasons for STP by technology, not widely adopted. The first one is simply lack of data or lack of images. A lot of insurance companies will simply not ask their customer to send them images. And the first time they see the car is at the body shop. There, I will dare to say it's already lost. If the car gets to the body shop and the body shop is sending the information, then I think a lot of the dynamic we see today will still happen. So you've got to have a good and intuitive technology and an incentive by a customer to provide you with images. Either you promise them more rapid repair and more rapid resolution or some sort of discount, some incentive. The second problem, even when you get the images, they're not great quality. So uploading images, and then maybe, maybe, or maybe not, they are good enough to figure out what's going on and have an estimate. So a solution needs to have a really cool, really intuitive process that still results in good enough data. And then finally, the industry that has been building automated repair estimate today, I believe, did not have the right baseline of data. They have been using the repair quotes from today's insurance claims, which means they already have the problems of the inflated repairs in them. I think that's where we decided to take a slightly different approach and all these problems. I think, you know, we are deploying two different sets of technologies to this problem. The first one is what we call Ravin Inspect. It's a web application. And instead of uploading images, it is basically scanning the car. So it creates a quality assurance on the consistency of the images. We look at every point on the vehicle from multiple angles. We also correct the user if they are not showing the right parts of the vehicle. So the chances of making mistakes or even doing fraud go down significantly.

So you need an app and it actually uses the camera rather than uploading a picture. And then I think we know this sometimes from QR codes, right? That's kind of when people wonder what the difference might be. Okay. Yeah.

Eliron: And then you avoid a lot of the issues that we see today with fraud. And maybe I'm uploading one image of one car, another image of another car. We, you know, when you scan the car, all of this goes away and we even have some fraud indicators by the way you walk around the car to say, okay, this guy is honest. It's okay. So that is the solution that we see to the quality issue.

Nikolaus: Can you speak to what the indicators are? Would we give too much away for the small population of listeners who want to conduct insurance fraud?

Eliron: Let's just say that if you think you can scan two vehicles that look the same in one scan, it will be impossible. We will find out. And indeed every suspicious behaviour, like stopping the scan for several minutes and then resuming it will be flagged as a risk. It doesn't mean we will accuse you immediately of committing fraud, but it will be flagged as a risk.

Nikolaus: Understood. Understood.

Eliron: And then, you know, on the other side, once you've got the quality images and you've got the green light that you are an honest person, how did we build our repair estimate? And the way we built our repair estimate is not necessarily by looking at the existing practices of insurance companies. It is coming from our experience in the fleet industry. It is coming from repairs carried out by people who own vehicles like you. If you want to repair your vehicle, you will find the most cost-effective way because you will go to your guy and you will repair the vehicle effectively. We think that because of that, we see that because of that, we are able to challenge the practices of let's go for the most expensive repair possible and we will adopt more of the methods that fleets use when they repair their own vehicles.

Nikolaus: Oh, that's interesting. Because, of course, you know, Hertz is going to run their body shop repair process very differently than even in insurance historic claims where all of the inflated costs are already in and you just don't know what the right baseline is. Understood. And you've mentioned two, that's the two Ravin Inspect and then it's basically the repair cost database, right? Those were the two.

Eliron: That is the Ravin Deep Detect, which is a massive brain that includes the repair estimate, but also includes things like multi-angle analysis of damage and extrapolation and consolidating several damages on the same panel to understand the depth and many different things that honestly, I don't know much about. And that's why I have smarter people in the company.

Nikolaus: And maybe that's a good point. Could you outline at which point you'll use AI or machine learning? Or is it for your big difference? Do you have a different definition of AI and machine learning for your own company? I'm not a big fan of definitions, but these currently AI seems to be more sexy than machine learning. I think three, four years ago; people were talking a lot about machine learning. Now it's obviously all AI.

Eliron: Yeah, these are both, I would say, quite beaten terms in the industry. And to me, artificial intelligence probably just means that it is doing some tasks beyond what you initially programmed it for. In our cases, we see multiple types of applications for artificial intelligence. We use it both at the very front end of the inspection. We just mentioned fraud prevention. So training networks on understanding patterns of fraud, training the system to understand what is dirt and what is damage, what is just a reflection and what is damage. You know, just like your human eye is trained to do without even thinking. And that is on the front end. And then of course, on the back end, understanding different types of damages, how they impact the vehicle, and what it means to the incident itself. Were you hit from the side or the front? What severity was the hit? What is the likelihood of the incident impacting some internal parts beyond what we see in the images? These are all questions that our AI team is actually addressing.

Nikolaus: If you were friends with an insurance CEO, what kind of advice would you give that person?

Eliron: First of all, I'd start by telling them, look, I'm sure you know this already, but you are paying anything between 200 to maybe 600 million euros in claims, in motor claims every year. Very clear. That number needs to go down. I think we all agree on that. The question is, you know, how can you quickly, without like a massive five-year project, introduce efficiency into your process such that you can automate, so reduce the logistics and the overheads related to motor claims, and control the costs of the body shops. And I'd show them a 90-day plan of doing that. And after the 90 days, we would already show some cost reductions.

Nikolaus: Is there anything else from what you've learned that you know, you as a person if you weren't working for Ravin anymore, was there any additional advice that you would give, rather than “you should check out my former colleagues at Ravin”, you know, to get a fix on your claims-ratio?

Eliron: I would say, look, there's a lot of buzz around technology. And it's also very easy to get lost. I mean, before Ravin, I had another company, and then before that company, I was actually working for Shell, for the company group. And we were doing a lot of experiments with technology. Anything from oil drilling to selling coffee in the retail shop, and all the way to the future of mobility. And I think the most important advice that I gave, you know, that was given to me, and I will give also others, is make sure you understand the business case. Don't test technology because it's cool, because everybody's talking about it. Make sure you are business case-focused. And that is the first advice. How does it save money or make you money? Seems obvious, but very difficult to do. And the second thing I would say is that perfect is the enemy of good.

Eliron: Especially in Germany, that's something, you know, there is a saying: “Übung macht in Meister” which means something like “who trains gets better”.

Nikolaus: Yeah „Übung macht in Meister“. Absolutely. Or “Es ist noch kein Meister vom Himmel gefallen” which means something like “nobody was born being a master in their field “. But yeah, we have a few of those sentences and proverbs in Germany. This was really, really interesting. Is there anything else we haven't touched that you would kind of leave us with before we wrap up?

Eliron: I enjoyed it very much. I think it's an exciting time for the industry. And I'm glad I could speak on your platform and hopefully get to some interested people that want to explore this space. But eventually it's not just about Ravin or it's not even about just images and claims and things like that. I think the whole industry is just forced to go through a major transition and I'm just happy to be part of that journey.

Nikolaus: Perfect. Thank you so much. Anyone listening, if you want to reduce your motor claims, maybe hit up Ravin and see how these guys can help you.

Thank you so much. It was a true pleasure.

Eliron: Thank you. Take care.

Nikolaus: Bye.