Root, Inc. (ROOT) Earnings Call Transcript & Summary

Good evening, everybody. My name is Yaron Kinar. I'm Goldman Sachs' Insurance Analyst. I'm delighted to have with me today this evening, Alex Timm, Co-Founder and CEO of Root, which went public at the end of October; as well as Dan Rosenthal, its CFO. Gentlemen, thanks again for joining us this evening. I thought maybe we jump straight into Q&A.

And first one, probably for Alex. So for those in the audience that are less familiar with Root, can you maybe explain and discuss the market in which the company operates? What end user need it is aiming to solve for?

Absolutely. And thank you for having us, and thanks, everybody, for listening today. As Yaron said, I'm Alex. I'm the Co-Founder and CEO of the company. Really where Root operates and what's unique about Root is we are a technology company that is based on mobile technology that's selling auto insurance. Now what this allows us to do is basically create a product that meets consumers where they are, which is on the mobile phone, and by doing business that way, we're able to leverage all of the rich behavioral data that you can get off of the mobile phone, whether it be how fast someone is driving, what time of day they're driving, if they're texting and driving, and allows us to incorporate all of that behavioral data to basically build a better pricing and underwriting mousetrap. And we do that. And so effectively, the consumer gets a very easy, fast, transparent product experience on the mobile phone, and they get a better price because what we found is that all of this behavioral data is actually much more predictive when it comes to trying to predict whether or not someone's going to get to an auto insurance accident versus a lot of the more older school traditional ways of price insurance, whether it be credit score or age. And that's really where we're focused. And versus -- you asked -- in terms of the market, when you look at the market, most of our competitors, I mean, still a large portion of the market is still trying to figure out what the Internet means for them and are really focused on agency channels. And even those that are focused on the Internet, and we don't really see are focus on mobile and certainly aren't focused on leveraging behavioral data anywhere near to the extent we are and the end user gets basically a great product at a great price.

Got it. And with that, as you focus on mobile and you focus on telematics, does that bring about a certain type of customer? Maybe we could talk about the typical profile of the Root customer and then maybe to add on to that, are you seeing that profiles change as you launch new iterations of your UBI model and as you expand that?

Our target customer and where we really focus is on -- about 24 to 35-year-olds. So we do skew younger, folks that are more digitally native and millennials that are comfortable, certainly, sharing their data, but that are also used to technology providing them ways, the adage you sort of get what you pay for doesn't really apply to that group. They believe technology actually solves that problem. And so for us, they're a group that largely is -- they don't understand their current insurance, they're frustrated. They don't want to use their Mom and Dad's brand of insurance. They don't necessarily want to go to insurance agent by any means. They don't want to spend 15 minutes online buying anything, particularly auto insurance. And so for us -- they're also addicted to their cell phones. So for us being able to actually talk to that customer, it's a very easy and natural way for them to purchase insurance. I will say one of the things that's interesting is as we have iterated on our models and get more and more data and understood more and more customer segments, we have started to branch out a little bit more into more suburban areas. We've started to add more age diversity to our book, even more gender diversity. We used to see more male, now we're more equally male and female. And I think that's largely because as the brand of the company becomes more well known, we're seeing a lot more people come into the space. We're also, obviously, during more difficult economic times, more broadly, people shop for car insurance because it is a substantial bill for most households. And so we are seeing sort of a more diversified book of business as we've continued to scale the company over the last few years.

And I guess, if we dive a little deeper at the actual mechanics. So we use telematics, usage-based insurance. I believe you measure the customer's driving for this 3 or 4 week test drive phase, right? And I think that applies for new customers. So can and do customers then delete the app from the phone after completing the test drive will be the first question and then maybe a few follow-ons to that.

Yes, absolutely. So you're absolutely right. The way it works is the consumer downloads an app, then they take a picture usually of their driver's license. From there, they don't have to enter in any information. We just say this is -- like, this is who you are. Can you confirm this is what you want to insure and who you are? And they confirm their profile with us. And then they drive for a period of 2 to 4 weeks on average. Over that time, we deliver them our coaching feedback on their driving. And then if they're a good driver, and this is very important, only if they're a good driver, they will actually get an insurance quote. So we actually take the worst 10% to 15% of folks and we tell them that we think other insurance companies probably have a better price for them. And what that allows us to do then is lower prices substantially on all of those good drivers that are coming through. Once they purchase, they can purchase the policy right in the app, we actually cancel their prior insurance for them automatically and get them a refund for any of the unused portion of their premiums. So it's very seamless. You can purchase insurance without ever touching a keyboard. And then from there, everything is managed within the app. So although you can, to your question, delete the app, we've found very few customers do. And that's largely because, one, they're sort of -- we've sort of trained them from the beginning that this is how you're going to interact with us is via an app; but two, all of the functionalities in the app, it makes it much easier, whether it be their scorecard, whether it be finding claims directly in the app, whether it be billing and payment management, they're really sort of native now to that app. So we see a very high rate in which people actually just choose to keep the app on and monitoring them.

Got it. So do you then use that data for renewals as well? Do you continue to monitor the customer in order to determine their renewal rate and segmentation score and the like?

Yes. So we certainly continue to monitor, and we continue to score. And what we do actually is we have not found the need at this time to basically go back and reprice them on renewal. And the reason for that is the way we have -- I mentioned earlier that the driving period can be anywhere from 2 weeks to 4 weeks, can go longer than that. The way we design that period is in a way that ensures that we have a very high correlation basically with our original estimate of the risk score and then our risk score at 6 month and renewal. And so if we ever were in a case where we said, "Wow, people's risk scores are really changing drastically", we actually go back and say, "Okay, what did we miss?" And so we haven't found the need to incorporate that at renewal. However, that's certainly something we think about a lot, we discuss a lot. There have been times that we have taken underwriting actions when we believe it's fraudulent or there was any fraud going on to actually get those customers out at renewal if we really need to. But in general, we don't necessarily see ourselves repricing really consistently like that.

Okay. And maybe tied to that, so I think one of differentiation or differentiating factors that you call out is the fact that you really, you own the telematics process end to end, right? You collect the data, you own the data, you own the loss experience, you own the premiums. And therefore, you can look at, I think, from the clear trends than maybe some of your peers. How does that apply -- the data collection of a specific customer that really apply to a 2 to 4-week period where maybe the loss event doesn't actually occur in that 2 to 4-week period?

Yes, absolutely. So the technology here is really not trivial. And I can't really stress that enough and that's why it's really important to build all of this in-house. So when you think of an insurance company, especially a direct insurance company, we're basically a prediction machine. And we do that by a bunch of algorithms. Those algorithms are incredibly messy and incredibly complex. We use a lot of machine learning algorithms because they're just much better than traditional models, quite frankly, at predicting the future. But once you start to try to take those algorithms and break them apart and outsource a portion of it, like most of our competitors are doing, it gets very, very tricky. And it's even worse, if you actually look, even one of our largest competitors said they take their entire traditional model and then they bolt-on -- basically, at the end, this telematics data. And that basically ignores all correlation with all of the other rating variables, right? A telematics program that tells you 16-year-olds are worse drivers than 40-year-olds is pretty worthless because I already use age, it's already in there. That's not really giving me any sort of predictive power uplift. So in order to actually fully price the insurance and using and leveraging this data, you need to have access to the underlying rating engines and the underlying rating variables and how those are incorporated. And that's not -- and do it simultaneous for all. And that's not something that we really see done, and it's really impossible to do if it's outsourced. The other thing that you mentioned is the losses. So once you have all of this data, you actually have to have real underlying ground-up losses and detailed loss data to understand, "Okay, how does all of this data ultimately correlate with whether or not somebody gets into an accident, how severe those accidents are, how does it correlate with claims fraud?" All of those things really need to be explored. And that's really hard to do if you don't have the underlying loss data. And if you look at most of the telematic service providers, their claims data is very, very thin, which is one of the reasons we said we had to become an insurance carrier. A lot of them have what they call implied claims data, where they try to predict whether or not a claim occurred and then train models on the prediction of another model, which obviously is fraught with difficulty. And so that's one of the reasons why we said we really have to in-house all of this. And what that's allowed us to do is basically control the entire technology and data science stack behind our pricing algorithms. And that then enables us to much more rapidly actually iterate on the data that we're collecting and make sure then that we are consistently improving our predictive accuracy.

And then maybe another question -- or it's a question on the mechanics. So Root tracks telematics purely on a phone app, right? I think there are some other models out there where they're using a dongle, some OEMs trying to penetrate this space a little more. I guess what is the disadvantage of using a phone versus an OEM or the dongle?

Yes, that's -- yes, absolutely. The phone, although the disadvantage of the phone is it's much harder. It's noisy. I have to figure out if you're the driver, are you the passenger, are you sitting in the front seat, are you sitting in the back seat, very difficult. The upside of the phone is, it's really hard to get closer to a consumer in a digital sense, you'd argue maybe wearables, but we're doing some experimentation there, than in the phone. And so we believe that over the long arc of the universe that the behavioral data is really going to fundamentally change the insurance landscape. And one of the things that you get off of the phone is things like texting and driving. But I also will get things like, okay, you just parked your car. And it's, I don't know, 8:00 p.m., well, did you go to a bar or did you go to a restaurant? I also get you consistently charge your cellphone every night. Does it look like you're sleeping the same number of hours every night when you plug it in and unplug it? I get what other apps you have on your phone and how you leverage and use that phone. You get all of these really rich behavioral insights that you don't get off of on OBD2 devices. In terms of vehicle data, we do collect vehicle data directly off of vehicles when they are connected. Still a very small percentage of the market that is connected in that way. But if you download our app and you do have an active vehicle data, you can actually skip the test run. And then what we'll do is we'll still monitor all of those rich behavioral things we just talked about over the first 6 months of the policy and actually incorporate that in our renewal. But we like to be data agnostic and really collect data from everywhere we can. But we think the smartphone and continuing to push into more behavioral based pricing is really, really important. We think that it's really hard to get closer. I mean, a lot of consumers speak with these things. So it's really hard to get closer to the consumer than the smartphone.

Okay. And does the wide range of phone models and the ever -- the consistent launch of new models, does that create a challenge as you try to true-up and align the data?

Yes, it's huge. So this is why you need a substantial -- it's not just a data science effort that we talked about, but it's also a substantial engineering effort. The way you actually talk to these phones is whether it be battery usage, build a battery, how frequently you ping the phones, how frequently you ping the different devices, each operating system update, all of those change. And so we have a large team of engineers that's constantly working on that. iPhones, it is not so many active models. You get into the Android world and you're talking just a wide diversity of models of phones. And each of those, you have to actually have a unique strategy for. And that's difficult, and that's a lot of time -- it's very time consuming. And different phones and makes and models actually have different quality of chips, even different iPhones, whether or not you have a Sprint iPhone versus a Verizon iPhone, they aren't necessarily all created equal. And so you have to actually understand all of those differences just in the hardware and also the software. So it's an ongoing engineering effort.

And maybe final one on kind of the more the nitty gritty of the mechanics. Is the sensitivity of most phones, is it sufficient to capture the various forces that you really want to track, whether it's hard brakes, acceleration, how you take turns, and so on?

Yes, absolutely, it is. But it's not trivial, and a lot of people have not -- a lot of folks have worked on this problem, and it's very, very difficult. And that's really one of the things that has differentiated us the most is that we have cross-validated these findings that we can actually accurately predict things like hard brakes using smartphone data, which is very difficult. It requires reaping the smartphone dozens of times per second to actually understand this. We've done lots of experimentation, whether the phone is in the driver seat or a bag. But these phones have really, really rich sensors on them. And you can get lots of data off of them very efficiently, actually. However, they're really, really noisy. So it takes a lot of ingenuity, and a little bit of luck actually, to actually develop a lot of these algorithms that can accurately predict the underlying physical events on the vehicle. And there's lots that are trying, and almost nobody does it well.

So now if we take a step back, I think we touched on a couple of these points, but maybe we try to put this all together. Why is UBI, usage-based insurance, so important to auto insurance? One. And two, how is Root differentiated considering the growing use of UBI across the auto insurance industry?

Yes, absolutely. So UBI is so important for a variety of reasons: One, it's just more predictive. And this is usage-based experience, telematics, behavioral-based systems. As we get more and more high-fidelity data, our predictions are just much, much better. You can imagine that -- it's not hard to imagine why either, right? Of course, if I know whether or not you're tailgating, you're probably going to do a better job at knowing whether or not you're going to get into an auto accident than if you -- than if I just know your credit score. And so that's the bar we have to beat, right, because that was the last real innovation in this space, which is pretty remarkable. And so it's not so hard to imagine that a lot of this behavioral data, whether you're going to bars at night, what you're doing, is going to be much more predictable with those accidents. The other thing, though, is that it's actually pretty beneficial for society. When you think about the way auto insurance is priced, it's effectively a tax on the [Audio Gap]. We sort of take the worst ZIP codes, the worst credits for folks, those who are at least able to actually afford to pay for auto insurance, and we charge them the most. This, however -- and by the way, a lot of those factors, you don't really have much control over. The consumer can't really go and say, "Oh, I'm going to change my credit score", at least overnight, it can take them years. Whereas this is just saying, "Hey, don't text and drive. Hey, don't drive drunk. Do things that are responsible that help society and you'll be rewarded." And that's incredibly, incredibly important. And that's why it's so important to get added and to leverage it. The hard part and the difficult part is most of our competitors are well over 100 years old. Lots of them are even still recruiting COBOL programmers, operating on mainframe systems having completely outsourced, even just the most basic functions of their technology. There are companies like [indiscernible]. So when we start to talk about the fundamental change in the nature of data from sort of a you're going to go and just pull up the data, enter it into a linear model and see how that predicts losses to suddenly truly high-frequency data that's coming in and using machine learning algorithms to real-time price. That's incredibly difficult of a transition to make. It really requires almost an entire reinvention of the company. And that's one of the reasons why we started Root is because we saw this massive opportunity to really start the first actual technology-based insurance company and to do so leveraging behavioral data. And we think it's going to -- we think the long arc of the universe again is in the direction of highly customized products and we think in financial services, with better pricing that's actually [Audio Gap] on something that we can control, and that makes sense to them. And we think behavioral data and machine learning is really the way to get there.

And I think you're working on version 4 of your UBI model now. Can you maybe talk about some of the key differences between version 4 and the current model?

Yes, absolutely. So this is -- and I love this because this is one of the most -- honestly, it's been one of the most fascinating things about starting the company. When we started the company, we had no data. And so you sort of go out there and you guess more or less on -- as part of the hardware story in data science company having data when you start, you sort of take a best guess. And not surprisingly, that guess is pretty wrong, at times really wrong, at times not so wrong. And then what you do is you start to iterate as you collect more and more data on the pricing algorithms and on your algorithms. And as we've done that, what we've seen is we've seen, basically, the loss ratio has come down, we've seen performance massively improve, particularly in customer segments and in states that we've been in for quite a while. And so there's this massive benefit to actually returning that model and continuing to grow in order to better -- to basically better our price. Our 4.0 model, which is right now in R&D and is moving into production very soon, which is our version after our UBI 3.14 model, which we named after our adoration for pie at the company, we're a bit nerdy. But our 4.0 model, some of the key differences there is we've actually incorporated a lot more interactions. And when you think of the number of variables that we use to price, the interactions between all of them, you get into very high dimensionality very quickly and in order to actually produce reliable results and massive amount of data. And as we've grown substantially and retrained our model substantially, we've been able to add more and more complexity into our telematics models and our pricing models which have shown to materialize. I mean, we're not talking small improvements anymore in the iteration of its models, we're talking giant improvements, some step change improvements to these models. And that's really what was key in 4.0.

Okay. So maybe we shift gears a little bit to talk about results. Maybe we start with top line. So Root's auto policies in force nearly tripled in 2020 compared to year-end 2019. What have been some of the key growth drivers for that growth? I guess that came out where -- but what have been some of the key growth drivers? What are the key growth drivers for the company's policies in force and direct written premiums looking at the next couple of years?

Yes, absolutely. So really, the key drivers behind Root's growth is we never thought we could -- the one thing about our competitors is that they're just way better than us as mass media advertising. These guys are advertising machines. So we never really thought we'd be able to drive a lot of growth by sort of running directly at them and sort of out GEICO-ing GEICO, out Progressive-ing Progressive in terms of advertising. And so what we really did is we focused on the product. And that ended up driving quite a bit of growth for us, actually. We -- as we've iterated, as we basically removed more friction from the product, as we have improved pricing, what we've seen is that conversion has gone up quite a bit, and that's allowed us to then expand and invest more marketing dollars in growth because as that -- those marketing dollars become more efficient, that basically apparently helps us grow faster. And so that's really been a huge driver. It's just continuing to make pricing better and better, which is the #1 purchase -- factor in purchasing decisions in this space, as well as continuing to invest more in product. And that's been a large driver of our growth to date. The other thing, we've also added multiple states. We're only in 30 states today. We want to be national, and we will be national in the coming months. And we think that, that will drive substantial growth. We've also added more product lines. We now sell renters insurance and home insurance, and that certainly has opened up new marketing segments and new channel segments for us, which has really allowed us to then expand them further. So a lot of the growth has been driven by product. A lot of the growth has been driven by adding new product lines in new customer segments, but as well as geographic expansion, which we think we still have a long way to go to really tap out on all that, all those growth opportunities.

Okay. And then if we look at 2020, and maybe I'll bring Dan into the conversation here as well, had I thought of a COVID environment, social distancing environment, had you asked me what that would mean for a mobile-first direct oriented company that's also offering cost savings through telematics, I would have thought that we'd see growth just accelerating and explode. And we've actually seen a slowdown sequentially, I think over the course of 2020. Can you talk about the drivers for that slowdown? Is that just really a function of reduced marketing spend during the lot of the year? Are there other drivers that led to that?

Great question, Yaron. And again, thank you for having us today and giving us the chance all day to meet with a number of investors and obviously host us today. We -- Alex and I have talked a lot about this and going back to the decision-making that we made as a management team in March, April, May, June. The reality is, when we went to work from home, which for us was March 13, none of us really knew how this was going to play out. Unemployment levels quickly skyrocketed, regulators started intervening. We weren't sure, there was a day or 2 where we wondered if regulators were just going to make auto insurance free for consumers for a period of time. And so there was just a lot of uncertainty. You had a lot of carriers start to offer large bonuses or rebates or what have you. So against that backdrop, we really made a conscious decision to be prudent stewards of capital, and not understanding where consumer behavior would be. So we took our $36 million of marketing spend in the first quarter, which was really heavily focused on January and February, less March, and we reduced it to $18 million in the second quarter, really fundamental shift. Now looking backward, devil's advocate, should we have done that? I don't know. I still think it was the right move, given that uncertainty. And given this really other important piece that ultimately, regulators did not allow price changes and certainly, price increases for a significant period of 2020. And so for us, because of what Alex talked about, one of our advantages is that constant iteration of the pricing model. Being able to refine -- based upon the data that we're ingesting, refine that model, move through version 3.0, 3.14, 4.0, et cetera, that's a really important part of our business model. And during the time period where regulators where restricting or prohibiting that from being able to occur, I think it's an appropriate decision that we made to curtail our marketing spend, and it certainly had an impact on our growth. Now a lot of companies would take the growth levels that we experienced in 2020 because it was really significant growth but we know we can have even more special growth as we enter 2021. So we're looking forward to reporting out our Q4 results, but even more importantly, to talking about how we view 2021 because our marketing has been really only recently has ramped up, as you know, and the fourth quarter, from a seasonality perspective, is always the quarter where we experience, frankly, the least amount of growth because we're competing against retail spend in the digital marketing space. And on top of it, we all know what the presidential election amounted to from spending in this particular fourth quarter. So I think stay tuned. We're excited to come forward at the end of February and talk more about our growth plans for 2021 and why we're really confident in our existing states as well as with the state expansion that Alex referred to.

Okay. Makes sense. I look forward to hearing more. One set of kind of recurring questions that I get when I talk about Root is around the company's retention rates. And I think you've noted on several occasions that the retention rate is a function of a less season book. So maybe we'll start off with talking about why seasoning matters and can the company get to industry average as far as retention rates are concerned over the long term?

Great question. And Yaron, you're right. The context, it feels like we've been at this a while. But obviously, we only became a public company on October 28. And if I think back to all the conversations that we had, in August, in September, Alex and I did more than 50 testing-the-waters meetings and in October, a roadshow of more than 50 investor meetings, probably 85% of the questions related to growth, telematics, marketing differentiation. And then we started trading on the worst week in the market since March. As we got to the end of October, we were in a blackout period for all of November, and we woke up and 85% of the questions in December and today are on loss ratio and retention. So we're excited about that. Like, I want to be very clear, we get it. We get that we're a 5-year old company. We're really excited about our telematics. We think they're differentiated. And we think as we come to that earnings call at the end of February, we're excited to share more proof points from a lot of what Alex described as we matured in certain states around retention and loss ratio. So we're excited to bring it to that. That said, we also recognize that we're a young company, and it takes time. And for us, about half of our premium comes from renewal customers as opposed to new customers. And you're right, frankly, a seasoned book for a legacy carrier would be closer to 85% of their premium would come from renewal customers, is generally going to drive a lower loss ratio and greater retention, a greater retention feeding on itself, not just for years but for even decades. So I think some of this is truly a function of time as we continue to trend and improve our retention rates. I think the other piece and I'm sure Alex will get into this too, there are a lot of things that we're doing around retention because I wouldn't tell you. Yaron, it's not like we're sitting in settled with where we are by any means. We see this continuing to improve, as we feed our customer segmentation and the mix shift in terms of who we're bidding for customers, as we minimize some of the pricing shocks through the portfolio, which obviously has a huge impact on retention. But also, I just think it's a matter of improving our customer engagement, continuing to improve the product and putting it in perspective. And some of that will come from a function of time over that time period.

So isn't there like a point where retention just naturally picks up like very significantly? Is it the second renewal, the fourth renewal, the tenth renewal? At what point do you have a sense to step function if it even exists in terms of seeing the retention improvement?

Great question. I think when you smooth out for price shocks and some of the other aspects, getting through that first term in that first term renewal is the key instance. And part of that is our business model. Alex talked about it, we underwrite out 10% to 15% of our customers because of -- based upon how they drive. We -- ultimately, we believe 10% to 15% of drivers make up close to half of insurance-related costs. And so for us, during that test period, particularly for our day 0 product flow, it's binding with a customer, having them go through the test drive. And then there's a part of our business model, a piece of our customer segment is actually underwriting them out. Now that hurts our retention number in theory because we've taken a customer through a period of 15, 30 or even 60 days and then underwritten them out, but we think it's the right thing to do for the long-term business model. So what I'm suggesting is whether it's that, whether it's non-payment cancellations, whether it's fraud that we identify, there is some noise in the first term, but as we work through that first term retention, you see whether it's industry averages or what we're seeing at Root, the second term renewal is automatically at a meaningfully higher percentage than that first term renewal, and we expect that as our book seasons, that will continue to improve.

Got it. Got it. And then maybe this next one is for Alex. So within auto, I think there are some customer acquisition strategies and product offerings that clearly target a higher churn customer in an unabashed way, right? And they could be very successful if they're priced correctly, if you collect the right amount of fees around them as well. Do you think that, in Root's case, the mobile-first customer acquisition approach attracts a type of customer that may be less sticky inherently and therefore, needs to be priced accordingly? Or is that -- are we just talking about letting this book season and this retention resolves itself?

Yes. Really, what we're seeing today is -- and this goes for retention -- and as Dan said, of course, terms 3, 4, 5 returns -- retain way better than term 1 and 2, no matter industry-wide. So that's certainly not an issue with Root. But in terms of our acquisition strategy, our customer strategy is actually driving churn or a higher churn segment. We don't really see any evidence of that, whether it's looking at our book's average FICO score, percent of our book that owns homes or percent of our book that's minimum limits. We don't really see what I would classify as often you hear it called nonstandard, but like nonstandard customers like the General or some of those others, and as you've said, they're very unabashedly nonstandard. We don't really see that mix in our book. We have some of that, low double digits, 10%, 12% of our book is probably in that range, somewhere around there. But you still see most of our book really being -- it's not super preferred. So we're not talking about the folks who need boat insurance and home insurance and car insurance, and insure their wine collection. It's not that either. We're really more, when you look at our customer segmentation, kind of right in the middle there. And so it's more of that middle-market customer. So why churn looks the way it does is actually a large function of the fact that what I was saying before is that when we launched, we had no data, and we've had to aggressively reprice many segments of our book. Some, by the way, in a good direction, in a positive direction, where we lower rates on them, they don't go and buy 2 policies though. So you don't really see that in retention. And some in the -- where we have to increase rates on. And when we do that, you see churn spike, and then it comes back down after the rates really sort of normalize. And so as Dan said, when you look at our mature markets, that's really where you see retention, pretty much stabilized to very expected levels, very industry-esque levels. And so that's really the primary driver there. In terms of our customer, and who we're targeting, we have recently seen us trend -- throughout 2020, you did see us trend more towards a standard customer mix. That's both a function, we believe, of brand awareness as well as adding on more different marketing channels where maybe a more standard customer may be. And so we have seen, for instance, our FICO scores come up. We've seen our home percentage ownership come up. And we've seen lots of those leading indicators come up. So it doesn't look as if sort of the customer by their nature is a more churney customer than average. They are younger, which is going to be a little bit more churnier, though. So they're not -- again, they're not that ultra-preferred either. The other thing that we've done in that we've really driven retention up with some of our product experience. So for some of those more churney segment, that really churney segment that we've got of that bottom segment, we actually have shipped something called Boomerang, for instance, where consumers can sort of pause their insurance and then turn it back on at a later date very easily. And we found that, that works for that higher churn customer segment. And to your point, we -- the nice thing about retention is it's actually much easier to predict than losses. So it's much easier for me to predict when a customer comes through our funnel, how long they're going to retain with us than it actually is for me to predict even the loss ratio. And so when they come in through the funnel, there's nothing wrong necessarily with the folks that we think are going to churn rapidly, but -- because we believe we can be successful there, just means you have to price them accordingly, and you have to make sure you recoup your customer acquisition costs upfront. So that's really the most important piece to it. We don't necessarily say there's a bad customer here. We just think you have to go in sort of eyes wide open know what kind of customer you're pricing.

Got it. I do want to sneak in a couple on the loss ratio before we end. So that has been other sort of recurring questions that I've gone and Dan, I think you alluded to it as well, so the company's loss ratio is higher than the industry average -- of the auto industry's average. And clearly, seasoning matters there. And I think 2020 may present a bit of a challenge in figuring out how much seasoning had an impact on improvement versus this lack of driving in a COVID environment. So from that perspective, what gives you the confidence that seasoning of your portfolio today is indeed driving loss ratios lower? Is there a way that you can track cohorts over time? Is there a way for you to try and clean out the COVID noise?

Yes, all of the above. And you're right, we've now today helped us cross more than 70 investor meetings that we've had since December 1, since our Q3 earnings. And our goal was 75. We'll pass that tomorrow. And it has been fabulous to get out and talk to investors because I think the feedback has been really positive. And we have a really clear understanding of what investors want to see. And on the loss ratio, and again, you'll see more of this specifically as we approach our earnings at the -- as we report on year-end. On the loss ratio, what we've seen is they do want to have an understanding of show me that the loss ratio is continuing, to your point. Investor feedback was very positive on what we put in the shareholder letter showing the first 9 months of '19 versus the first 9 months of '20. That gives us confidence, apples-to-apples, with pretty much the same states other than West Virginia that came in early in '20, that we're seeing a really material trend. Now no question, as you alluded to, part of that is COVID. And the good thing about us is because we're not using COBOL programmers because we're actually tracking using mobile-based telematics, we're showing all of our driving data in aggregate across our customers on our website. So any investor can go to our website and actually see when our customers are driving less or when they're driving more year-over-year. And what you see, as we parse the data is that no question, from mid-March through early June, there was significant COVID impact averaging about 15 to 17 points a month over those months, so crosses over a bit of the first quarter and some of the second quarter, no question. But even if you put that aside and you look at that same chart showing the state loss ratios year-over-year, it's significant improvement from our state management and from the seasoning of the book. And also, you'll see that as we come out with some of the proof points in late February, you'll see what I said earlier, which is renewal loss ratios continuing to trend materially better than new loss ratios, which, again, I think, is a show that as we have the right customers coming into the funnel and as we retain them, the loss ratios are continuing to improve. So we're excited to share some of that data and continue to get the investor feedback in the meantime.

Great. Okay. I'd love to continue down the loss ratio path and understand what you're doing, but I think we are almost out of time. So maybe one final question to close out. When you look at the coming year, what excites you most about the business? And maybe what do you envision Root will look like in 5 years?

Yes. I think what excites me the most is the business we've learned -- we've made a lot of mistakes, obviously, in our time. We have the scars to show for it. And I think what excites me when I look forward is just the massive growth opportunity. There really is no technology-based insurance company out there today that is really accelerating and taking market share the way that we think is possible. We really do believe, and I think we now have proof in the pudding that this behavioral data and this high-frequency data that we're collecting is much more predictive and build a much better customer experiences and product. And quite frankly, we are very confident that we have a better underwriting and pricing mousetrap. There will be some efforts to get there and going national and adding new commerce segments. And we don't believe it's going to be an easy road. But we do see right now that we're operating at a $266 billion market, and we believe we have a significant technology differentiation. And so there's really massive room for growth here and to continue to grow and continue to extend our brand. And I think in the long term, what we're looking to do is build the next historic, great technology company in insurance. And we think that, that is no small task, but we also think that it is a once-in-a-lifetime opportunity, and everybody is really excited to go and build that company and really have an impact.

Yaron, if I can just add quickly. I am just thrilled -- if you ask about the next year, I'm thrilled that 2021 is not going to bring an IPO. We can actually focus on executing on our business plan and doing exactly what Alex just talked about around the strategic vision. I'm thrilled and so excited about that to just get down to business, honestly.

Well, Alex, Dan, thank you so much for your time and insights and best of luck with the execution on the plan.

Thank you, Yaron.

Thanks for having us.