Progressive Corp

NYSE:PGR

Earnings Call Transcript 2018-Q3

[Operator]

Welcome to The Progressive Corporation's third quarter investor event. The company will not make detailed comments related to the quarterly results in addition to those provided in its quarterly report on Form 10-Q in the letter to shareholders, which has been posted on the company's website, and we'll use this event to respond to questions after a prepared presentation by the company. This event is available via a moderated conference call line and a live webcast with a brief delay. Webcast participants will be able to see the presentation slides live or download them from the website. Participants on the phone can access the slides from the Events page at investors.progressive.com. In the event we encounter any technical difficulties with the website transmission, webcast participants can connect through the conference call line. The dial-in information and passcode are available on the Events page at investors.progressive.com.

Acting as moderator for the event will be Julia Hornack. At this time, I will turn the event over to Ms. Hornack.

[Julia Hornack, executive]

Thank you, Chanel, and good afternoon to all. Today we will begin with a presentation on industry-leading segmentation. After that presentation, we will have Q&A with our CEO, Tricia Griffith; our CFO, John Sauerland; our guest speakers today, Pat Callahan and Sanjay Vyas. And Bill Cody, our Chief Investment Officer, will join us by phone for Q&A. This event is scheduled to last 90 minutes.

As always, discussions in this event may include forward-looking statements. These statements are based on management's current expectations and are subject to many risks and uncertainties that could cause actual events to differ materially from those discussed during the event today. Additional information concerning those risks and uncertainties is available on our 2017 annual report on Form 10-K where you will find discussions of the risk factors affecting our businesses, safe harbor statements related to forward-looking statements and other discussions of the challenges we face. These documents can be found via the Investors page on our website, progressive.com.

It is now my pleasure to introduce our CEO, Tricia Griffith.

[Susan Griffith, executive]

Good afternoon and welcome to Progressive's third quarter webcast. We have a terrific session for you today. Before we go into the topic on segmentation, I did want to reiterate a couple of the milestones I talked about in my letter.

First and foremost, we surpassed $30 billion in net written premium on a trailing 12-month period. That is phenomenal for us. What's more phenomenal is the fact that less than 3 years ago, Glenn Renwick , John Sauerland and I toured the country celebrating $20 billion in net written premium. So to put it into perspective, it took us nearly 80 years to get to $20 billion and less than 3 to get to another $10 billion, celebrating $30 billion. Phenomenal job for everyone at Progressive, and we couldn't be happier with those results.

In addition we just surpassed 20 million in overall policies in force and 13 million in auto policies in force. Why that's significant is the fact that we are growing at a very rapid pace on Robinsons in both the agency and direct channel. And part of that is you want an influx of future Robinsons. So those auto policies support, that growth, is really important because they are our future Robinsons. Having that 2-pronged point of reference to be able to grow Robinsons is important, and that's why our policy growth is important. We always talk about unit growth being a very important measure.

And lastly, we had an internal goal of getting to 1 million Robinsons, and we surpassed that as well. So a couple more milestones I won't go into, but a lot of excitement around the momentum we've had over this past couple of years. And we are ready to kind of wrap up 2018 and have great plans for 2019. Let's start with our 4 cornerstones: who we are, why we're here, where we're headed and how we're going to get there.

Who we are, our core values. That is really the underpinnings of everything we do. We have 5 core values, and all of them work together to make sure we have a successful company. Our goal is to win, but we want to win in the right way, and our core values guide us there. Who we are.

Why we're here. Our purpose statement is, true to our name, Progressive. And when I think about that, I think of the first 8 letters, Progress, always thinking differently, always questioning what we've done yesterday, should we do it the same way tomorrow, listening to our customers, listening to our employees and our shareholders and understanding how to always be forward thinking.

Our vision, where we're headed. We want to become consumers' #1 choice and destination for auto and other coverage. For us, this is just a great thing to be able to try to achieve. We got a little bit closer there this year. We know there's a lot of room that we need to get there, but we are ready for it.

And our strategy, that is really how we're going to be able to achieve our vision. Our strategy, as we've been talking about in the past several webcasts, are really based on 4 different pillars. And you've seen all of these and what we've spent the time doing is really dissecting each of the pillars, so you can understand our business model -- again, in order to understand how we're going to achieve our vision. In fact, in the next quarter, we'll have our Chief Human Resource Officer, Lori Niederst, talk about our people and our culture, which is so key to everything that we do.

Today, we're going to do a little bit deeper dive in the pillar you see with the blue background. So if you recall in the first quarter, we had Kiara Berglund and Mike Esposito come up and talk about the balance of operational efficiency and claims accuracy, always trying to get that near-perfect balance of cost and quality. Today, our guests are going to be talking about offering competitive prices driven by industry-leading segmentation.

Let's start with our overarching goal, our operational goal for the company. And this is something that we've had in place for as long as I can remember, but nearly 50 years since it has been written about in every one of our annual reports since we went public in 1971. Our goal is to grow as fast as we can at or below a 96 combined ratio. The only caveat to that is we won't grow if we don't think we can service our customers in the way they have come to expect.

The great news here is that with our tremendous growth in these last 3 years, we have been able to hire -- hire in advance of needs, specifically in the CRM and claims organization. And I can tell you, the talent that is coming into the doors of Progressive -- through the doors of Progressive is phenomenal, so I'm excited about our future. For us, a 96 isn't a "solve for" variable, it's a constant. And it sets the direction for the whole company in everything that we do. It's a great balance of consumer competitiveness and attractive margins. And it helps us in certain situations because it is something very specific and it provides zero ambiguity.

As an example, in the quarter 3 of 2016 right after I took over as CEO, we went over 96. We had more cats then than we've had the year before on our commercial lines business. We had put some rates in but hadn't earned in yet because it's an annual policy. We were spending more on advertising. Probably spending a little bit more across the board. And literally that zero ambiguity was so helpful to the entire organization because we all got together and said, "Okay, what levers are we going to pull to make sure we get under that 96?" And in fact, when we talked about expenses, it caused us all to say, "Let's question everything we do as far as expenses." And we've changed since then. So even that particular exercise was a valuable one when we went over 96. Again as you know, the end of the story is we got under 96 by the end of the year.

Probably the most important part of the 96 is that it's truly ingrained in our culture. It's so core to everything that we do. In fact, as I speak to you on this webcast, I'm standing in Studio 96. We developed an in-house marketing agency several years ago that we refer to as 96 Octane. That's just a couple of examples of how embedded 96 is in our culture.

Why? Frankly because it works. This slide you're looking at now are a couple of results. You'll see results from 2015, '16 and '17 and then on a 5-year and a 10-year basis. So obviously growth and profitability, when we look at underwriting margin, this is us compared to private passenger auto. We've outpaced the industry substantially in every single time period. And as a matter of fact if you look at our 10-year results, we -- our results are on average 6.5 points, well better than our 4-point margin that our goal is. And if you look at the industry, they haven't made money in every single time frame.

We -- profit is one of our core values. We are going to make at least $0.04 and do everything we can to make at least $0.04 in every opportunity and every time frame.

Net written premium. We have grown, outpaced the industry twofold, again in every one of these time frames, so growth and profitability.

I already talked about our other measure, policies in force. So like I said last time, we are firing on all cylinders, and this is really an exciting time.

And lastly what these results end up being is double-digit comprehensive return on equity for our shareholders. So it's a great balance for everyone. We've been able to grow the company for our employees, service our customers and grow the whole company for our equity holders.

So let's get on to today. I really want to have a deep dive into segmentation. Why? Because it's really important to all these measures that you see. We have 2 guests.

Sanjay Vyas has been with us 15 years. He has been a product manager for 8 of those years in a multitude of states. And in his last 7 years, he's a general manager, leading product managers. He's going to give you a front-row view of how important that product role is at Progressive, owning your own state's P&L, really understanding it and then looking at the aggregate how it all fits together.

In addition and I wrote about this in the letter, we just recently have done a cross-functional job swap, and Sanjay is a recipient of that. So just recently he took over as the claims controller. And I'm so excited for Sanjay as well as the others to be able to get to know different parts of the company.

Job swaps at Progressive is a hallmark of what we do. We want to have talent across the company, a really good bench of people that have both depth and breadth. So I'm really excited to see what Sanjay brings to claims.

But before then, I'm going to ask Pat Callahan to come up. Pat Callahan, as you know, is the Personal Lines president. He's going to give you an overview of the product strategy, specifically segmentation pricing and how we match rate to risk.

[Patrick Callahan, executive]

Thanks, Tricia. Let's get started by jumping back to Tricia's last slide. And if you focus on the upper right-hand corner of this slide and look at Progressive versus industry, profitability and growth, you'll notice they're growing twice as fast as the industry while simultaneously delivering profit margins 8 to 10 percentage points wider. These results don't happen by accident. They are the direct result of our deliberate investment and increasingly fine segmentation and matching rate to risk. I'm thrilled to spend a little time sharing with you our investments we are making to continue to advance this segmentation and widen this moat around our Personal Lines franchise.

Our vision is to grow Progressive in pursuit of becoming consumers' #1 choice and destination for auto and other insurance. We all know that growing an insurance company is frankly not that difficult if you don't care about making money. And making money in insurance is not all that hard if you don't worry about growth. We are laser-focused on both. And as a result, it's absolutely critical that we have high operating efficiency, great claims accuracy and industry-leading segmentation.

So I'll now share a little view that we've talked about previously, that breaks down the U.S. auto and home market by distribution channel in the columns and by Progressive segment across the rows. For decades, we've been heavily focused on growing share within the simpler needs auto insurance marketplace. The $90 billion row represented by the Sams and Dianes in the chart that you're looking at. And while these simpler needs customers were our bread and butter for a long time, they trained us and honed our segmentation skills, primarily because with shorter policy life expectancy, it was critical for us to get our prices right, given they weren't with us for all that long.

Not with our expansion to focus on the broader $230 billion-plus opportunity that the bundled home and auto marketplace represents, this pricing discipline is serving us well. Because number one, it ensures that we understand our lifetime cost to serve for these new customers, and ensure that we don't have subsidization as we expand our market focus. But number two, it also enables us to bring incredibly disciplined segmentation skills to disrupt the property insurance market, much the way we did with auto insurance over the past several decades. And finally, it enables us to leverage the advantaged position we have with younger, simpler needs auto insurance customers to continue to build products and services to enable them to grow with us throughout their insurance lifetime.

We support this product expansion strategy within larger market opportunities with 5 key tactics. The first is to continuously invest to improve our matching of rate to risk. The second is to build and deploy highly competitive and stable products to meet the needs of these preferred customers. The third is to leverage verification to ensure accuracy of our rating models. The fourth is to continuously expand and enhance segmentation by leveraging third-party data, both our Snapshot usage-based insurance data and other third-party data sources. And the last key tactic here is to ensure that while we invest to improve the accuracy of our rating models, we're also keeping an eye on ease of use, to ensure that as complexity of our product models expands, we continue to deliver simple-to-use products for our agents and ultimately for all consumers. These 5 key product tactics are the foundation of what we call the virtuous cycle of risk selection.

So the U.S. auto insurance market is highly competitive. There's many players, very similar products, high information transparency and relatively low switching costs. Given the dynamics in this market, it's absolutely critical that we have a highly segmented and accurate product model. I'll walk you through this cycle starting at the top, step 1, where we design and deploy these highly segmented models and leverage our pricing discipline to have highly competitive rates in market. Following around clockwise, step 2 is these highly competitive rates enable us to drive up conversion and retention for more preferred segments.

Which takes us to step 3 where less-well segmented competitors experience adverse selection as their better customers are finding more competitive offers from Progressive. Which leaves these competitors ultimately with adverse trend. Their book looks to be increasing in loss costs. And they respond rationally by raising rates, but they raise base rates, given the lack of segmentation.

And that takes us to step 4 where higher base rates create shopping across their entire book of business and a greater shopping rate with all their customers, results in their better, more preferred customers finding more competitive rates with carriers like Progressive. And that leads us to step 5 of the cycle where ultimately, we grow premium to support fixed costs, which lowers our overall cost structure and brings us more loss experience and more data on which we feed back to the top of the cycle, which enables us to be more finely segmented and offer more accurate pricing.

Now the sub cycle that you see between steps 3 and 4 at the bottom of this is what we call that cycle of adverse selection, and that's where a competitor raises base rates because they don't necessarily know which segments are underperforming. But by raising base rates, they cause shopping amongst our entire book. And only their best profitable customers leave, creating adverse selection and requiring even greater rate increases.

Today, we're going to be focused on that top entry point of this cycle and talking about deploying our highly segmented products models, really with 2 sections to the conversation. I'm going to cover product design, and then Sanjay will cover product customization and deployment. And for purposes of this conversation, we're heavily focused on the auto insurance market. We'll talk a little bit about property and destination, but heavily focused on the auto market.

When we think about our product models, we build highly complex, engineered and architected product solutions to deliver 2 objectives: number one, match rate to risk better than other carriers; and number two, deliver on our strategic goals of having competitive rates for our core customers, while simultaneously offering competitive and stable rates for more preferred customers. We liken this to a high-rise building where not only are there foundational and structural elements to the model, but there's a significant number of operational systems that are all designed to work in concert as we attack this strategic growth opportunity.

And we'll start with the core foundational indemnity models. These are the models where we collect information about our customers in order to predict future loss cost. And our core indemnity models are honed and built on top of prior product models, such that we are constantly evolving and improving our segmentation over time. We layer on new rating variables that might be outside or internal variables more finely segmented pricing through analytical methods. Or potentially copying good ideas that we see in the marketplace. But ultimately, when we look at our product segmentation, we often see competitors who try to replicate it where they may copy the data that we use. They may copy rating elements that we use. They may even copy analytical methods in order to deliver similar segmentation.

But frankly, doing this would be similar to me buying the same shoes LeBron James plays in, and frankly, it wouldn't change my basketball game. Our core structural advantage is in decades of rate-making segmentation and culturally, a relentless pursuit of matching rate to risk.

Now we know that there's far more to the cost to serve a customer than just the indemnity or loss cost. And that's why we also layer in segmentation for acquisition expenses, operations expenses and billing in bad debt. And on top of this loss and expense segmentation, we then layer on our usage-based insurance program, which is an optional program, but highly predictive of future loss cost. And then finally, we have state-specific versions of these product models, given we operate in 51 different jurisdictions, which Sanjay will cover in more detail. But we tailor our product model to the required coverages, permissible or prohibited rating variables.

But the best segmentation and the most highly predictive rating model is only as good as the data that you input into the model. And as a result, we focus on ensuring that the data that we get from customers at inception is absolutely accurate. You can think of our customer onboarding process similar to joining an interstate toll road. And every year, millions of new customers get on the Progressive highway and most continue on unabated to their destination. More than 90% pass through what we might refer to as the E-ZPass lanes. You get on the highway and you continue on to your destination.

However, there's about less than 10% for whom there is mismatching data at inception. And we ask those customers to stop at the cash pay lane, so that we can confirm the accuracy of information that they're providing. And that's what this is all about, accurate information so that our models can predict accurate future loss cost, which ultimately delivers fair pricing for all of our customers.

This verification really happens in 2 steps, mostly focused on new customers, where at inception, a new customer is filling out a quote form, and we may find data that doesn't match with what we pull from third-party sources. We'll stop the quote, ask them to call at that point.

For other customers, again a very small percentage of our customers during the first 60 days are what we refer to as a free-look period, we are able to run additional reporting against the data the customers provide in order to ensure that we have the most accurate data to price that policy on a going-forward basis. This accuracy enables us to avoid subsidization across our book of business and ultimately to continue to invest in a more accurate and complex product model, while concentrating the expense associated with manual follow-up on a very small subsegment of customers for whom the payoff is much higher.

So when we think about our product model, I want to share a little bit more about our most recent development around 8.5, our latest auto product. Insurance is a highly competitive and winner-take-all marketplace, in that the carriers who have low operating cost, high claims accuracy and industry-leading segmentation can offer the most competitive pricing, while also delivering their target profit margin. And we know given the competitive nature, that nothing is standing still. So it's incumbent on us to continue to invest, to improve segmentation constantly.

We do this through multiple sources. We may find new sources of external data that are predictive. We may find new ways to analyze internal data. Or frankly as I mentioned before, we may see a disruptor Insuretech or an incumbent carrier actually have different segmentation that we are not too proud to copy and frankly to improve upon when we layer it into our segmentation scheme.

So let's talk a little bit more about a couple of those key sources of segmentation that we are in the process of deploying. The first one I'll talk about is coming with our 8.5 product which is deploying to market currently. And that's for refined segmentation based on prior insurance. So when we were building 8.5, one of the objectives was we wanted to improve the accuracy of the rating model without increasing the complexity for our customers or our systems. And in pursuit of that, we started looking at prior insurance, which has been a predictive rating variable for a long time for us and most other carriers. We found subsegments or more finite and granular information in the prior insurance data we were already pulling about applicants for insurance, that enabled us to segment the market and discover that even though our 8.4 product was performing really well in market, as you can see on the screen, the yellow bars were segments that were overpriced. The red bars were segments that were underpriced in our 8.4 product. So as you can imagine, leveraging this pricing insight in our segmentation enabled us to lower prices on the yellow segments, ultimately becoming more competitive there, and frankly to raise prices on those red segments to deliver the target profit margin, whilst slowing down growth in these underpriced segments.

Second area that I'd love to share just a little more detail on finding new insights from existing data. With the increasing digitization of customer interactions and experiences, we collect billions of customer interactions. And we were able to mine those customer interactions in order to identify segments and combinations that are predictive of future cost to serve. And by identifying this segmentation and predictive power, we were then able to look at our incoming customers.

And at any point in time, about 10% of the new customers to Progressive aren't really brand new to Progressive. At some prior point in time, they had a Progressive policy, but also had a lapse in judgment and left us to buy insurance elsewhere. Now when those customers come back to Progressive, see the error of their ways and come back looking for protection, we are now able to price more accurately on those customers, which enables us also to avoid segmentation -- avoid subsidization across the book and to ensure accuracy for every single one of our customers, which we believe is just a fair way to run the business.

The final source of segmentation that I'll touch on is within usage-based insurance. And it's been almost 2 decades now since we pioneered this rating segmentation. And as we brought it to market, it has continued to be our single most predictive rating variable. Given that, it's our goal to continue to invest in usage-based insurance to drive up the accuracy and ultimately the adoption of UBI or Snapshot within our product suite.

Customers up until about 2 years ago would opt-in to the Snapshot program. We would send a hardware device that plugged into the OBD port on their car. That hardware device would collect vehicle operation data, transmit it to us, and we would use it in pricing the policy.

Now in pursuit of increasing adoption, in the fall of '16, we launched an app-based version or a software-based version of UBI, which collected all the same valuable data on how the vehicle was operated, but also for the first time, gave us an incredibly rich data set, based on how the vehicle was being operated at the same time as the mobile device was being used. And over the past 2 years, we've been able to collect more than 1.5 billion miles of data, where vehicle operation and mobile device operation have happened simultaneously. And spoiler alert, not surprising, distracted driving does drive up insurance loss cost. With this rich data set, we have started using it in pricing, because despite the academic surveys or studies, this for the first time gave us a direct correlation between vehicle operation, device operation and insurance loss cost. And we're pricing on it now in just one state, but in the process of rolling that out. And in that state, we're using how much the device is used, how the device is used, and ultimately why the device is being used. What's really exciting here is not only do we have a pricing model that includes this distracted driving, but that enables us to use some economic leverage to potentially curtail or slow this pretty risky driving behavior.

Finally we talk about increasing adoption of UBI. Because we currently offer both a device-based and a software-based, we're able to optimize our presentation to get more customers to opt-in to the app-based or software-based usage presentation. And we do that to take advantage of the advanced segmentation that I just mentioned. Given a choice, more customers opt for the app-based version. And we are investing right now to continue to increase both availability and ultimately eligibility of all customers to get the app-based version.

So as you can imagine, this segmentation has incredible predictive power, and we're rapidly deploying 8.5 to market. Today we're in 16 states that represent about 40% of Progressive premium. And by year-end, we'll be in 23 states representing about 2/3 of our premium.

To date, we've been monitoring performance of 8.5. And while we were thrilled with how 8.4 performed in market, 8.5 is outperforming across all key performance indicators. And one of those that we look at regularly is, "Is our product model advancing our Destination Era strategy?" And for that, we look at preferred customer conversion, pre-8.5 and post-8.5. And on the slide, you'll see significant lift in conversion across a wide variety of preferred customer metrics, whether it's prior insurance, home ownership, better credit, clean drivers, multicar, across the gamut, we are converting better.

And that delivers value in 3 primary ways: number one, longer policy life expectancy; number two, greater propensity to bundle with Progressive, better credit, future homeowners or current homeowners; and number three, fueling that virtuous cycle of risk selection by enabling us to collect more data on these preferred customers that we can use to price our policies and out-segment the competition going forward.

So the key for us at this point is to get this segmentation to market as fast as possible. Now I'm not a patient guy. And when I talked to you a few years ago, we talked about accelerating our speed to market through what we called our pace initiative. And at that point, we had cut the time to deliver a model upgrade to market just about in half, and we continue to invest to do it quicker. As an example, for 8.5, which we're currently deploying, we are able to reduce the model build cycle time and significantly increase our deployment capacity in pursuit of getting this segmentation to market faster.

But what I get really excited about is additional investments that we've been making to ensure that we can develop and deploy product models almost in parallel. 8.5 is rolling out. 8.6 is in development and tests for launch next year. And 8.7 is already being built from a requirements and expectation perspective. So we feel phenomenal about not only our current segmentation, but this robust pipeline of future segmentation to fuel that virtuous cycle of risk selection.

So let's take a look at how our product models are performing in market. We'll look at loss and LAE for Progressive in the orange line and for the industry in the black line, and this is over the most recent 4 or 4.5-year period. You'll note that we are out-segmenting the competition by about 5 points from a loss and LAE perspective back in 2014. And our investment in product development and risk selection widened that gap to 13 or 14 points as we got into '16 and '17. And while you note the industry has been seeing a lower loss and LAE recently due to both extreme rate take and some benign trends, the gap persists.

Now what we really like to see when we look at selection and what we're doing in the marketplace, we look at rate take. So the rate change for Progressive is in the blue line. The rate change for the industry is purple. During the same period of time while we're deploying 8.3, 8.4 and 8.5, we've had to take half the rate increases that the industry has. That means current Progressive customers retain better through more competitive rates, and new customers coming in the door find more competitive rates relative to alternatives in market.

So we feel great about the segmentation and how it's helping loss and LAE, but let's take a broader look to understand what it's doing to drive Progressive growth, by looking at a 15-year history of Progressive's auto policies in force. And this is a period of time when we more than doubled the business from about 6 million policies to more than 13 million. And if we zoom in on the period that we just talked about, that more recent period of time when we were deploying 8.3, 8.4 and 8.5, or the period when we grew from about 9 million to beyond the 13 million policies we're at today.

If we look at how this growth has accelerated, using our million PIF growth milestones as a benchmark, you'll see that it took us 30 months to go from 9 million to 10 million policies in force. Yet after deploying 8.3 in the middle of 2015, it took us half that time to go from 10 million to 11 million. Deploying 8.4, it took us half the time again to go from 11 million to 12 million. And our most recent million policies from 12 million to 13 million, we did it in about 6 months’ time frame.

Now if we zoom back out to look at how that acceleration overlays with our product deployment, it is absolutely clear to us that investing in segmentation and speed to market for product upgrades is creating growth and driving growth for the company. Now what's particularly rewarding is when we overlay our combined ratio during the same period. And for the past from 2007 on, we've been rock solid within 92 to 95 combined ratio. So our operational goal of growing as fast as possible, subject only to that 96, is absolutely being delivered by our investment in risk selection and segmentation.

Now with that said, competitive auto is one piece of our Destination Era strategy, but we also know we have to have highly competitive property insurance offerings in order to meet the needs of the Robinsons. And that's why since taking the majority stake ownership within ASI, or what we call Progressive Home, we've invested to ensure that we are leveraging the property insurance expertise from the Progressive Home division in concert with the pricing and segmentation discipline that Progressive has honed over 80 years. This combination of property expertise and pricing segmentation is enabling us to leverage new solution methodologies, identify and deploy new variables and ultimately have much more accurate and competitive property insurance products.

But competitive property and competitive auto on their own are not the only benefit of making the property insurance investment. There is a strategic benefit of having a holistic data set that for the first time for us, merges the auto-rating elements and loss experience with property-rating elements and loss experience. And having this holistic data set enables us to look across products. And it's that cross-product segmentation that we fully expect will drive our segmentation into the future. Here's a quick example just of auto-rating elements that are highly predictive of property loss experience that we are in the process of building data feeds to ensure that we can get to market.

So in aggregate when we think about combining Progressive's expertise in segmentation with the Progressive Home division expertise in property, we fully expect this will enable us to not only run highly competitive rates, but to fuel that risk selection growth that we know ultimately creates adverse selection for our competitors and will drive profitable growth for Progressive into the future.

With that said, I'll turn it over to my colleague, Sanjay, for some more details on product deployment.

[Sanjay Vyas, executive]

Thank you, Pat. My name is Sanjay Vyas. As Tricia mentioned, I've been in the product world for 15 years and just recently moved to claims. Given my experience in product, I wanted to add some additional context to what Pat just shared.

Specifically, I wanted to touch on 3 points. First, states are complex. I'll discuss how that affects how the product is implemented. Second, I'll cover what does a products -- what product managers' decisions are across the 51 jurisdictions. And last, we'll see how product managers do more than just crunch numbers.

So to the first point, state deployment is complex because of the legislative and regulatory diversity across the U.S. I'll impact this with a personal example. Recently, my wife was at Whole Foods here in Ohio, and she was parking and a pickup truck backed into our vehicle, a relatively new vehicle. So this could happen in any state. In most states, this is simple. The other driver was liable and her insurance company would pay for the repairs. But to highlight how things can be different across different states, in Michigan, the then neighboring state of Michigan, my insurance company would pay for the claims, not the other driver's. That's because Michigan laws emphasize first party for physical damage. You can appreciate then that our data should solve for Michigan differently than for other states.

And just to round that out, in the grocery store accident, thankfully nobody was hurt. But had there been injuries to any driver, the jurisdiction would matter. In about 20 states, drivers may be able to get their medical bills paid for from personal injury protection coverage or PIP as it's known. The amount of PIP coverage varies across states. In some states, it's really small dollars. In others, it can be $0.5 million or even more. By contrast in the other 30 states, there's no PIP. It'd be covered through health insurance or through the auto policy if MedPay had been purchased.

One last point highlighting the complexity that exists at the state level. Some states like California, North Carolina, maybe New York goes on that list, these states are more prescriptive about how the product should be structured. And those are some large markets. Those states require different models than the ones found in other jurisdictions. So you can see that the states vary by as much the types of yogurt that you would find at Whole Foods. And those differences are meaningful for the product design. The decision-maker who tailors a product at the state level is the state product manager. The product manager or PM has 1 to 2 states under her or his responsibility.

For example, during my PM career of 8 years, I had responsibility for Arkansas, Mississippi, Alabama, Kentucky and then finally Florida. And somewhere along the way, I appeared to have lost my hair.

The PM runs the state. They are measured first on profit and second on growth. The first day I joined Progressive, I was told that the number 96 was important. 96 is our combined ratio target. And hitting the 96 would improve the odds that my badge would work the next day. I had started on Arkansas and Mississippi, this was in 2003, and we weren't making money in Mississippi. By contrast, we were making the target margins in Arkansas. 50-50 odds that the badge would work did not sound great. So I started working on fixing the issue.

Just to give you some context for my thinking, I saw 2 facets of the problem. Looking forward, how do we get Mississippi back to a 96 when it is over 100 CR? And second, how did we get here? If I look backwards, did the segmentation in the state fail? Or were the conditions in the state such that all the carriers were losing money? So we'll cover segmentation later. First we'll talk about the rate level.

So back then I worked with an actuarial staff to assess the indicated rate mean. Next, we filed the proposed rate increase within the state. It was a state where prior approval is needed, which is true in about half of the states. The state approved the rate increase after asking a few relevant questions. So we implemented the rate changes.

And then I spent the next 6 months monitoring the results. As you know, it takes about 6 months for customers to pick up on the higher rate level. In any given month, some customers will be at the old lower rate level and some at the new higher rate level. Each day, each month, that mix changes, and that makes measurement a little tricky. And during that same time, the environment also continues to change. Frequency at a state level can change due to seasonality like weather, changes in gas prices. And then there's severity, severity can be influenced by changes to medical costs, changes to labor rates, tariffs on parts. Interpretation of results is complex at a state level, which is where the PMs focus. Of course, thankfully some of the state level variation gets washed out at the countrywide level.

To share an analogy that you might have heard before, setting rates is like surfing -- we're trying to ride the wave. Now you can get ahead of the wave and crash, if the assumption about historic and future loss trends are too low. On the other hand if the assumption trend is too high, we'll inadvertently overprice the book and come in well under 96 and choke off growth. That's like not catching the wave at all.

So working on Mississippi, I spent time reviewing numbers to see what course correction, if any, would be needed. I also reviewed our competitors' filings to ascertain whether or not the problems were associated with the general market or my particular business.

So that's the pricing of the book. It’s what we call setting the base rate. It's also -- the job also involves segmentation. Segmentation is the most significant part of what we call product. As the job title implies, product manager and product management involves advancing the segmentation. So Pat spoke about the countrywide support that drives the product. As a PM, it would be my job to adopt that product to the local market.

So for example, I elevated the 8.0 product in Florida in 2011. So I'll share 3 examples of modifications I considered. Before I jump in, I'll point out I'm using some old examples of mine in lieu of some more recent examples from current product managers. I want to be mindful of sharing too much information about their great work with competitors.

So the first thing I needed to do was to modify the product to ensure it complies with state laws and regulations. So for example, in Florida, points for speeding tickets may not be assessed against the in-force policy. That's the rule. That's different than a lot of states. I had to alter the product to ensure compliance, but there's more to it than that. I also need to assess what other parts of the product need to be modified to make up for this loss in segmentation. For example, if you can't charge for speeds, and if speeds are more likely for 20-year-olds than for 55-year-olds, then it might be more accurate to modify the driver age factors to pick up on the segmentation loss for not being able to charge for speeds. This will make Florida different than the countrywide solution, but possibly more accurate.

The second product consideration for the PM is the territory model. The rates for Miami should be different than those for Tallahassee: pretty smart, right? But we don't price cities. We price at a more granular level and adapt by line coverage. So data can get pretty thin.

So right there, you see a value we get from our scale. I'll contrast that with what a smaller carrier would be able to do. A smaller carrier has 2 choices. First, they could use less data to derive their territory factors. That's not great. Or they can use an industry solution which provides no advantage over the other carriers in the marketplace. And if this small carrier has an innovative product based on maybe new segmentation, then there's a question about the relevancy of that industry territory model to their product.

So back to Progressive. At Progressive, our product managers will look at the historical data and take into account previous model changes. The goal is to estimate the factors at the territory level, while accounting for all the new product changes being made.

A third aspect is local market considerations. For example, I'd seen a decade ago that customers in Florida who had many prior PIPs had very high prospective loss costs. At that point in our company history, we had a policy to offer a rate for every risk. So I raised rates on the segment, but the loss ratio did not improve. It did not meet targets.

Now I could see that competitors were not writing this business. We frequently review our competitors' filings and saw that -- to see whether or not those insights are applicable. And in this case, it was. So I filed a rule with the department, saying that under those conditions, we could not offer a rate for customers with multiple prior PIPs. Since then, we've added back some segmentation, so that the story has changed and we do accept business in that segment.

So to summarize, at -- state-level details such as unique laws, territory pricing, local market considerations, are significant for the product. Those 3 examples undergird our assessment that focus at the state level is necessary.

To describe our system, I'll borrow a chart from Pat's presentation. We start with data at the countrywide level and at our size, the amount of data we have creates a scale benefit over smaller competitors. But being bigger is not the same as being better; we want to be nimble, too. To get that, we have a diverse group of individual PMs identifying local insights. And then we feed those insights back into R&D. It's very possible those state-specific insights are relevant to other states or even countrywide. We acknowledge and reward PMs who contribute outside of their own states to the broader product development.

I'll close with a comment about the product manager role. It's an awesome job. I loved it. It's numbers plus a whole lot more. Let me share a profile of the types of people we hire into this role.

We have a combination. So we start with homegrown talent. These are folks who've grown up with the data, who have connections across the company. And we also have folks with professional degrees. For example, we have someone with a PhD in robotics from Stanford. So we make sure we source the best. And we want to have diversity in the ranks to ensure that we're thinking about our business problems broadly.

So in summary, we hire business people into the PM job. We don't focus exclusively on actuarial skills or industry knowledge. Rather, we hire quantitatively minded business people.

We do that because the state profit and loss rolls up to the product manager. As you know, we are one of the few companies that reports out results on a monthly basis. That's true countrywide and also at the state level.

I was a product manager for 8 years and a general manager for 7. And over those 15 years, once a month, I usually had a hard time sleeping, because the next day, the monthly income statement would come out. Once a month, Sunday was just a little tense. I viewed myself as the owner of the business. Seeing whether we hit our 96 mattered a lot. Seeing whether the segmentation played out the way I thought it would, mattered a lot.

A great deal goes into going as fast you can at or below 96. It is more than pricing or segmentation. There are many structural elements to our business that also influence the profit and growth. To that end, as a PM, I'd be involved in many facets of the business. I'll share a couple of examples.

First, 7 years ago, as the Florida product manager, I was driving around the state. As I recall, I was visiting agents in the state, and we were rolling out the new product. So I wanted to make sure they knew about the new rate level or perhaps new questions on the quoting platform. So we just finished meetings in Miami. And we were driving up to Jacksonville, and we drove through a toll booth. And I saw a competitor had put ads at the toll booths. I'm guessing the tolling authority was selling advertising space around the toll booth. So I saw their ads and I was irritated. We have a great brand. And I wanted that smart, targeted advertising for Progressive. So I worked with our media team. We bid for those toll booth ads, making sure we were under our targeted acquisition cost. And we won those bids. Progressive ads started to be displayed at toll booths throughout Florida.

Here's another example of being involved with all facets of the business. This one's a time when I was the GM. The PMs and I would meet with claims leaders. The goal? To understand how the product actually afforded coverage. There's a lot of complexity. We want to make sure we want to get to what we call ground truth. That's what one of my claims peer would say: How are claims actually being settled? Here's a relevant example.

The PMs and I drove to a claims meeting, this was in New England, and we were talking with members of the Connecticut claims team. We happened to review a file where the Connecticut customer had been driving in New York City, and they had a claim. And unfortunately, the PIP, the insured, was injured. And we pay out on PIP coverage, which is established by New York law. There's a catch. Connecticut doesn't have PIP coverage. So we don't charge premium for a coverage that we don't offer. But as we've just seen, we do -- there are instances when we do pay out on PIP.

As an astute businessperson, I realized that having a coverage that we paid out on, but one we didn't collect premium on, could present an issue. Don't get me wrong, this wasn't an issue without controls. We weren't losing money per se because of this. The losses were tracking back to the Connecticut income statement, so we were rate adequate. But there are implications to the product. And as a point of reference, in a perfect world, I charge only the customers who will be traveling to New York over the next 6 months a higher rate, because that's where the exposure emerges, in New York. But I don't know that perfectly, and I certainly don't think our customers know for sure whether they're going to be traveling to New York over the next 6 months.

So how do we handle it? Well, we have a few options. We could rate on this in the liability line coverage, bodily injury. And that's what we had been doing, I learned. That would have all the customers in the state pay for these claims, and that's not unreasonable at all. But there are alternatives. One is to focus this cost on a territory. That would make sense if the exposure was localized to policies in the southwest part of the state. So what we're always trying to do is to match rate to loss costs.

This insight about PIP exposure in a non-PIP state caused us to improve our segmentation. That's the value of ground truth, of working in all facets of the business.

So those are 2 examples of where driving around the field, we learned about advertising opportunities in Florida and about unique segmentation opportunities in Connecticut. And it's really complicated and we love that. We are driven to tailor the product to our state and out-segment our competitors. We are driven to hire and develop product managers who are entrepreneurial, curious and passionate about winning the right way, by following our core values. And we are driven to go deep on every facet of the business, because we know that important insights happen locally. The point is, as PMs, we are driven to grow as fast as we can at a 96 combined ratio or better, in every single jurisdiction.

With that, I'll bring Pat up.

[Patrick Callahan, executive]

Thanks, Sanjay. Having spent about half of my Progressive career as a product manager, general manager, I can absolutely attest to the significant benefit that the combination of local market knowledge, customizing our countrywide segmentation, brings to the company overall.

So to wrap up, we talked a little bit about how we build and deploy highly complex, countrywide product models. And Sanjay shared with you how we customize those product models to the opportunities in front of our product managers in order to deliver profitable growth across the country. The combination of both our countrywide product models and local product customization, in concert with accelerating speed to market, is fueling this virtuous cycle of risk selection, which ultimately creates profitable growth for Progressive and delivers against our strategic objective of growing more preferred business and expanding into a much larger Robinson-addressable market. The combination of investments we continue to make against this target market is absolutely delivering on our vision to become consumers' #1 choice and destination for auto and other insurance.

With that, if you can give us a couple of minutes, we will set up for Q&A.

[Julia Hornack, executive]

[Operator Instructions]

[Operator]

Our first question comes from the line of Mike Zaremski of Credit Suisse.

[Michael Zaremski, analyst]

My first question is regarding your competitive advantages relative to peers. I'm curious if you feel they're materially stronger today versus just, let's just say, a couple of years ago? And if the answer is yes, if you can kind of boil down what are likely a lot of reasons, down to maybe the 1 or 2 underlying reasons you think are having the most impact.

[Susan Griffith, executive]

I would say absolutely, that we continue to try to drive a greater gap between our peers. And it's really a lot of what we talked about today. It's hard to actually talk about 2, but segmentation and what we talked about today with matching rate to risk is critically important, especially to achieve our operational goal. There's a lot of other things that go into play. So our cost structure, we think, is really important. We know that people can shop. It's easy to change. And having a competitive cost structure is very important to be able to have competitive prices. And then I think some, if I only had to do 2, I would say those. But again, our service, our local footprint, our brand, there's so many other things that come into play and it's all those things in aggregate that we believe chips away at being able to have a competitive advantage.

[Michael Zaremski, analyst]

Okay, great. And my last -- my follow-up is regarding -- thanks for mentioning 96 Octane, which is your in-house marketing agency, and I had a quick moment to go over to its website. So just curious, a lot of advertisers are speaking openly about trying to better understand and manage their digital ad spend, given the tectonic shift towards digital consumption. It's getting a lot of press. So I'm just curious if I'm barking up the right tree or not. And is this one of the reasons Progressive has grown faster than many have expected?

[Susan Griffith, executive]

Well, we, from an advertising perspective, we try to be out and about where anyone wants to shop. So a lot of that has to do with demographics. 96 Octane was really born from us understanding and being able to do a lot of the things in-house that frankly, we were paying a premium on for doing outside. So that was a little bit different. On the media side, a couple of sessions ago, we had our guests talk really about understanding efficiency of ad spend, whether it is on social media, on Instagram or on mass media. We do believe that we have an advantage here. And that's one of the reasons why we buy a majority of our advertising in-house, because we believe we have an advantage. We've talked a little bit about that. I won't share great details, but yes, we think that's a competitive advantage. So 96 Octane is a little bit different, that's more on the creative side. But our in-house media buy and understanding of making sure that we only spend to our allowable costs, I think has, is a competitive advantage.

[John Sauerland, executive]

To that I would add, the digital space is obviously highly measurable and we spend a lot in the digital space and have for quite some time. Dan Witalec, in the session that Tricia mentioned, also tried to convey the level of understanding we have now in advertising attribution in the mass media space. So increasingly, we think we are able to buy shows, dayparts and understand exactly what the cost per sale is for those purchases. That obviously allows us to buy very efficiently, but also to negotiate very effectively with the media channels.

[Operator]

Our next question comes from the line of Elyse Greenspan of Wells Fargo.

[Elyse Greenspan, analyst]

My first question, I just wanted to get a little bit more color, I guess, on the PLE disclosure that you guys have in your Q. The growth slowed in the agency channel and it's actually down in the direct channel for a 3-month basis this quarter. I was just wondering if you could talk to a little bit about what you're seeing there. And is that being driven by what's gone on with maybe some of your competitors taking less rate and pushing for growth?

[Susan Griffith, executive]

Elyse, so yes, in terms of growth, we still feel positive. And the measure we use internally is trailing 12. We've talked about trailing 3 being an indicator. So we obviously always look into that. We have seen from a rating perspective, less aggressive rate taking. So if you go back a couple of years, there was a lot of rate. It's smoothed out a little bit, for the most part. There's some competitors still taking a little bit of rate and some actually reducing some rate. So that will put some pressure on PLE. That said, we continue to think about the nurture part in the ways of making sure that we take care of our customers, whether it's on the CRM side or the claims side. In addition, and I'm the executive sponsor for our retention team, there are times when we might make a decision for, and knowing that it might affect PLE negatively, on a process change. I won't go into the details, but we might look at a process change where we know it might cause customers to not stay as long, but we want to balance PLE with lifetime underwriting profit. So occasionally, we'll make those tradeoffs, because it's the right thing to do for the business. But again, we are very dedicated to continue on our path of retention and I don't have a crystal ball, but we look at really everything from rate to service to nurturing our customers.

[Elyse Greenspan, analyst]

Okay, great. And then my second question is on the severity trends, which also did go up, based off of the disclosure in your Q. If you could just talk to what you're seeing there? And then as you think about pricing for frequency is still negative, but if severity continues to drift up, can you talk about how that might impact the prices that you plan to take?

[Susan Griffith, executive]

Yes, so as far as frequency, it's been -- we're pretty much in line now with the industry. We're a little bit -- it was a little bit lower in past quarters, it's fairly benign now. As far as severity, we do price to those trends. So the trend you saw this quarter was really, the increase was based mostly on collision and PIP. And PIP has a lot of volatility, so I'll talk more about collision. So our collision severity trend this quarter was just shy of 9%, about 8.7%. For the year, year-to-date, it's still like 6.7%. But we're seeing more frequency and severity on total loss vehicles, so newer vehicles becoming total losses. So if we continue to see that trend, of course, we'll price to it. And we don't price to overall. We'll price when we look at certain states and the channel, et cetera.

[Operator]

Our next question comes from the line of Amit Kumar of Buckingham Research Group.

[Amit Kumar, analyst]

Just going back, I guess, to the discussion on higher severity, and Allstate flagged the same thing this morning in terms of the newer vehicles showing the higher severity component. Are you surprised by this change in trend which we are seeing? And I would have intuitively assumed that with all of that ADAS, it would have actually helped the trend somewhat in terms of collision warnings and all that stuff. Maybe just talk about has anything changed over the last 3 quarters? Because if I look at the older collision trends, it was flat. And even if I go back to '14, '15, '16, it was up in the 4% range or so. So what has changed recently?

[Susan Griffith, executive]

Well, very specifically, like I just talked about with Elyse, it's the total loss, both frequency and severity. And remember, even though ADAS is more and more prevalent in vehicles, there's still a large fleet out there of cars that don't have that safety equipment and there are a lot of people that have that equipment and turn it off. So it's, that's really hard to measure. From the word surprised, we just look at data. So when we look at -- when we see the data, we react to it, we watch it very closely. I try to get away from being surprised at anything I see, since we've been in this industry a long time. We always follow the data.

[John Sauerland, executive]

Yes, and I think the important thing to note is that our average premium has kept up, or actually, in this case, outpaced the pure premium growth, the frequency times the severity or average loss cost. So I think by that fact, you could assume that we don't have an exact under-the-line coverage. We can't predict the future, but in aggregate, our pricing has been keeping up or actually outpacing the loss cost trends.

[Susan Griffith, executive]

And we often talk about that. And Pat hit on that when he showed the chart of the delta. It's really about staying ahead of that trend. And we've always talked about small bites of the apple. We don't want to rate shock our customers, and we do think that influences PLE as well, so we'll continue to do that.

[Amit Kumar, analyst]

The other question, the only other question I have -- and this 8.5 is actually very -- it's fascinating and very helpful. I don't know if this question is for you or Pat or someone else. When I look at the 8.5 deployment on slide, I think it's 14 or 15, I can't see the page number. Today versus the footprint at the end of the year, is there any method to that launch? I guess what I'm trying to understand is was there something which you noticed in those states which are being cashed up on initially and then it's being rolled out into other states? Or is that just as usual, how it sort of, how the process evolved?

[Susan Griffith, executive]

I'll -- if Pat wants to come up, he can, but I don't think there's a huge amount of rhyme or reason. Obviously, we want to get it out. When we believe it works, we want to get it out in high premium states, if that makes sense. Right now, we're in 15 states, 40% of the premium. There are also, there are states where we still have to get up to the 8.4 to get to the 8.5. So it's sort of a whole map and blueprint of how we're going to do things, and frankly, how it can go through our IT department as well, and so that's -- that would be my answer.

[John Sauerland, executive]

Yes, and how the states are currently performing on the 8.4 model, the regulatory environment can also affect the schedule. So we have a schedule that's fairly dynamic. But those are the considerations that go into that.

[Operator]

Our next question comes from the line of Josh Shanker of Deutsche Bank.

[Joshua Shanker, analyst]

I want to talk about the new business penalty. You have been growing very aggressively, but you've also talked about telematics and also about customers who've been customers before who are coming back. Over time, does the new business penalty diminish for you in terms of taking on new business?

[Susan Griffith, executive]

Well, I think, and you can add on a little bit on this, I think from the new penalty, a lot of that comes from new business, but also from the advertising. But also, that's balanced with retention. So as our retention has grown, that has equaled out a little bit. We're going to continue to spend a lot on advertising. So that will be expensive on the direct side, but again, we only do that if we think it's efficient. You talked about telematics, that's a huge variable for us and in part, in terms of not just understanding the ultimate loss costs of those customers, and now especially, as we evolve into more distracted driving, but our preferred customer base as well. And then you talked about customers coming back. That's just us to understand both how to underwrite and understanding more about customers that have come before. So a lot of times, if it's a new customer, we don't know anything about them. We learn along the way. We've actually already known this person. It's like if you dated somebody, you broke up and you start dating again, you already knew he didn't put his shoes away. So you've got to figure that out and we know a little bit more. Not a great analogy, but it's the best I can come up with.

[Joshua Shanker, analyst]

And along those lines, do you have a vast sort of collection of people who were customers or not -- maybe they weren't customers, they were 'try then buy' type people, who you never got around to dating? Or is that really negligible?

[Susan Griffith, executive]

So customers that quoted and didn't buy?

[Joshua Shanker, analyst]

Or initially, you were talking about -- you said people could try out Snapshot and see what their discount is, and if they like the price afterwards, then they can come and you get to see how their driving is before you've actually taken on the risk.

[Susan Griffith, executive]

No, we had tested something called Test Drive a while -- years ago, and it was a little bit -- it was like that. But no, we, when you're with -- now you're with Snapshot, you are a customer, and we have other variables as well.

[Operator]

Our next question comes from the line of Kai Pan of Morgan Stanley.

[Kai Pan, analyst]

So I just want to be more specific on the pricing versus loss cost trend. If you look at frequency up 6%, frequency down 3%, so that loss cost trend like up 3%. Your average premium per policy is up 4%. So you still have a little bit, like 1%, room to improve your margin. But my question is really, is that going forward, how quickly are you going to react to the potential rising cost trends on the pricing side? Will you try to catch up with it to maintain the margin? Or you could let it drift a little bit higher, because now you're in the 90%, away from the 96% target?

[Susan Griffith, executive]

From the reaction perspective, we are a rate machine. So we can react, I think, quicker than anybody in the industry. So that to me is something that's a really big strength of ours. We will react if we continue to see trends, whether they go down or up. And it really depends on, we can look at our conversion and we might do things differently if our conversion changes. Right now, we feel really great about where we are rate-wise. And we're not going to react, to again, one line coverage code in one quarter of severity.

[John Sauerland, executive]

And I'd also highlight what Sanjay was talking about, which is our product management structure that's very local and down at a pretty low level in terms of product analysis and product action. So if the product manager believes there is benefit to using that 1 point for a certain segment where they can get more competitive, perhaps it's Robinsons, and grow more, again, subject to that 96 combined ratio, they'll make that decision. So this really comes down to very smart, very motivated local folks who are figuring out how best to deploy that margin. It could be advertising. It needn't necessarily be rate level. We can say we want to advertise more in that marketplace to, again, grow as fast as we can, subject to that 96 ceiling.

[Susan Griffith, executive]

Yes, it's so important to have that balance of having a really healthy margin and growing. And if you make that decision because you think you can grow, we really want to make it about the -- you can grow and not just throw away that margin.

[Kai Pan, analyst]

My follow-up question is that since you're so fixed on the 96 number, so I imagine you probably cannot wait for your segmentation model from 8.5 going up to 9.6. My question is that on the way from 8.5 to 9.6, is some -- is there a phenomenon, like, call it, diminishing marginal returns, that you get less as you move further and further? What I really am trying to get at is of your other opportunities, for example in the Robinsons, which you probably just started this progress, you could put more effort in that, making a bigger marginal impact, and with that helping accelerate the growth in Robinsons segment as well as improve the margin of that target market.

[Susan Griffith, executive]

Sure, and we've shared over the years, sort of those Gini curves, where you get incrementally better, and obviously, there are big segmentation variables that kind of give you that leapfrog. Think of credit and especially for us, usage-based insurance. Where I think our biggest opportunity is, and John, feel free to weigh in, is the fact that now we own a homeowners company. And to be able to share the data of losses that happen in both home and auto and have that deep segmentation on the home front and the auto front and combine those. So to me, there's so much to do there, and that's what's so exciting about our ownership, our eventual full ownership, of Progressive Home ASI.

[John Sauerland, executive]

Yes, and we did start at product version 1.0. So there's been a lot of iterations. And they've all been far incrementally better than the prior version. We think there are a lot of other opportunities. You both hit on, in talking about the Robinsons and pricing the household more effectively over time, even beyond auto and home is an opportunity for us as well. But we absolutely, as Pat showed, have 8.6 in the R&D lab, and we'll be out soon with that. We even have 8.7. So we look at the opportunities that are in front of us, figure out which we want to deploy and which product model. And invariably, we have many more opportunities on our list than we actually roll to market and they just roll to the next product version. So we are not yet seeing any diminishing returns in terms of product model upgrades.

[Operator]

Our next question comes from the line of Paul Newsome of Sandler O'Neill.

[Paul Newsome, analyst]

You talked about the importance of, in prior companies, prior insurance, has the type of company that you get your customers for, I'm assuming it has changed. Maybe you could talk about how it has changed over time in the last year or so, in terms of where you're -- from whom, what type of competitor are you getting your typical customer today, versus before?

[Susan Griffith, executive]

Yes, I mean, I think there's obviously a lot of big players. So we focus on the bigger players, as I'm sure they focus on us. A lot of it has to do with rate. So if you get behind in rate and you have to take a lot, those customers are going to shop and they will likely come to another company that has a well-known brand, like Progressive. And so I won't necessarily name competitors, but we've seen a lot from some of the major competitors, and I think that had to do with just rate. And it's easy to shop auto insurance in both the agency and the direct channel and we see that when people raise their rates.

[Paul Newsome, analyst]

What about cross distribution? I mean, obviously, direct is taking share in general. But could you talk a little bit about any changes in how people are -- I mean, are people going straight from agency to direct? Or are they going from agency to independent to direct? How has that process evolved in recent years?

[Susan Griffith, executive]

It's hard to say. So I think for years, people thought everyone would go direct. We have a very healthy robust Agency business. We have 30,000, 40,000 independent agents and they are growing. We're really happy about that. So I think it really is an individual thing. And for us, we want to be where, when and how customers want to shop. So we've got the direct channel, we continue to grow in that across the board, but more specifically on the homeowner side. So I think years ago, people said, "I wouldn't shop for auto online, that's crazy." A lot of people do it now -- actually, obviously, a lot of people. And from a homeowner's perspective, the same thing. And we're seeing a lot of people shopping through our HomeQuote Explorer. So really for us, it's about the individual comfort level of how you shop. So it's hard for me to look at data across there, but I will tell you, we are growing in both channels, obviously as you see, substantially, but very happy with that. And we'll watch trends as they change, but so far, we really do focus on what the customers want to do.

[John Sauerland, executive]

And while we've seen, in aggregate, a trend towards more direct, what we see within channels is generally, people are shopping within the channel. So if you looked at the distribution of our customers, new customers, it would pretty fairly reflect the companies and their market share within those respective channels. The direct channel, actually, in aggregate over the past few years, has been actually taking more share from captive agent companies than from independent agent companies. Share for independent agents has actually stayed fairly constant. And we've seen captive share drop a bit and direct, increase. When we talk about prior insurance in our rating, we're talking about a lot more than just from where you came most recently. We're looking at a longer history of how many different carriers you were with, how long were you with them. It's a combination of a lot of similar attributes like that, that we're looking at that segments what was previously a fairly binary rating variable of do you have prior insurance or don't you? We've now segmented that far more finely.

[Susan Griffith, executive]

Thanks for that, I wasn't catching that. And I think on the captive side, it makes sense, because you only have one choice. And so for us, that's why we've really loved working with independent agents, because they do have choices, and I think people will demand that.

[Operator]

Our next question comes from the line of Marcos Holanda of Raymond James.

[Marcos Holanda, analyst]

So in the context of the achieved milestone of $30 billion in premium, I was hoping you guys could discuss the balance between capital demands and growth, and how that could potentially affect your annual variable dividend, and if you can discuss the formula for that dividend, that would be great.

[Susan Griffith, executive]

Yes, so I mean, obviously, we want to grow as fast as we can. And this is a fairly capital-intensive business. We're not buying things, but we have to have regular -- regulatory capital and then contingent capital, so it has taken a demand. We obviously just went to the debt market for another $550 million. And we believe that will allow us to continue to grow throughout this year and next. The variable dividend is based on a percentage of after-tax underwriting profit. It's a formula that the board approves every year, so -- and that's changed over the years. It used to be 20%, then it was 25%. This year, it's 33.5%. We multiply that times the gain share factor within the company. So right now, our gain share factor is 1.91. It varies from a 0 to a maximum of a 2 point. So if you do that math across that multiplication, that's how we come up with the variable dividend. The only caveat to that with the variable dividend is if our after-tax comprehensive underwriting, I mean, comprehensive earnings are less than our after-tax underwriting profit, then we don't pay anything. We pay 0. That's happened once, and that was in 2008 during the financial crisis.

[John Sauerland, executive]

And that is, as we say in the Qs and the Ks, it's subject to board discretion. So the board could choose, even if that criteria that Tricia just mentioned, of comprehensive income having to be greater than after-tax underwriting income, is not met, the board could choose to pay a dividend nonetheless. And if the board chose to pay less than the formula, it's under their discretion as well. We do have a lot of capital needs, given our growth. We're on track for almost an additional $6 billion in premium this year. We think of our surplus needs as a 3 to 1 for auto and around half that for home, so a blend that's in the 2.8, 2.9-ish kind of range right now. And obviously, we want to continue to grow. And that's why we did access capital markets recently, again, to ensure that we have the capital we need to support the operating business. If you look at our financial policies, I think you'll find it is all about supporting that operating entity, because that is how we consistently drive shareholder value. So we think we're in a pretty good position going into the next year. Depending upon how much growth we have next year, we may find ourselves in need of incremental capital. That will remain to be seen. Margins obviously right now are very good. Investment returns could be better, is the way I would characterize it. So we'll see how that plays out going to the next year, but right now, we feel great about our position.

[Julia Hornack, executive]

I'm actually going to take a question from the webcast now: Please discuss your ambitions. And I think that this is because multiple times in the past, we've talked a lot about the different sizes of the property and casualty market, the Personal Lines segment versus Commercial Lines segment. Please discuss your ambitions in the Commercial Lines insurance business. Do you envision in 10 years that Progressive will be a significant underwriter of such lines as worker's compensation, commercial property, commercial liability, et cetera?

[Susan Griffith, executive]

So we talked a little bit, I talked a little bit about this in my letter. And we obviously just rolled out what we call BQX, BusinessQuote Explorer. So that is insurance for small business. Think of GL and business owner policy. That is something that we believe we have a lot of opportunity, just getting started there. I'm not going to venture out and talk about 10 years from now. What I'll say is this, that we're the #1 Commercial Auto writer. We have a brand and we have the availability to, I think, really -- we have a lot of runway, I should say, to really grow in that space, because of our brand, because of our knowledge of that area. And whether we write it all on our paper, or use partners, which we will be using, assume it will do -- it will be like what we did with home, with our Advantage Agency on the Personal Lines side. We see this whole destination for commercial and that's what we have been building. We started building it last year. We just rolled out, had a soft launch this quarter. So all I can say is, I'm not going to talk 10 years out, lots of opportunity, we're very excited about it. And we believe it'll be something really significant.

[John Sauerland, executive]

Yes, we've noted before, the Commercial Lines space in property and casualty is in excess of $300 billion. And we play in a slice of that Commercial Auto historically and today, really that is probably 10% of that. And we are aspiring to move into a broader offering on our own paper as well. So the BusinessQuote Explorer offering today is predominantly third-party carriers for coverages outside of Commercial Auto. We have filed with the first state a general liability and business owners policy, and we hope to be out in that state early next year.

[Operator]

Our next question comes from the line of Meyer Shields of KBW.

[Meyer Shields, analyst]

I'm trying to understand whether the increasing sophistication of the pricing that you're talking about, does that make your near-term results more or less subject to [ just ] variables like gas prices?

[Susan Griffith, executive]

I'm not sure if I understood the question. I missed -- I -- you mumbled a little bit in the front or I didn't really hear the first part of it.

[Meyer Shields, analyst]

I'm sorry I mumbled. I'm just trying to understand, given the reliance on data, and obviously, it's working out really well, I don't mean to sound critical. But does this make your results, your underwriting results, more or less dependent on external realities? So I'm using gas prices as an example, maybe that will affect overall mileage.

[Susan Griffith, executive]

Right. I mean, I think from that perspective, when you look at macroeconomic data like gas prices and unemployment, those normally end up affecting frequency. And we sort of, we don't price for frequency. We really price for severity. So I would say it's always important, so we always look at it. We follow it very closely to determine trends. I would say -- I wouldn't say it necessarily changes. Would you differ?

[John Sauerland, executive]

No, I really don't think having greater sophistication in your pricing puts you at any -- in any different position relative to loss cost trends that are driven by gas prices, other economic changes. I really don't. So based -- those other economic changes might affect segments of customers more or less than others. And to that degree, we might see changes in trends more so. Say if unemployment increases over the past, we have seen changes in claiming behavior and coverages such as Personal Injury Protection. So to the extent your segmentation drives your mix of customers to one end of the spectrum or not, you might be at greater risk in terms of economic changes. But in aggregate, I would say it's really not an issue.

[Meyer Shields, analyst]

And then the second question. I'm trying to understand the elasticity of pricing within homeowners compared to auto. My superficial assumption would be that there's less elasticity, because typically, a smaller dollar policy, but I don't know if that's accurate.

[Susan Griffith, executive]

I think that would be accurate. I'm not -- I -- do you...

[John Sauerland, executive]

Yes, so shopping for auto insurance is a lot easier than shopping for homeowners insurance. So I think there is some greater stickiness. The fact is that for a lot of homeowners, that homeowners premium is bundled within their mortgage as well. So it's not quite as obvious when the rate changes. So I think if your hypothesis is elasticity is greater in auto than home, I think that'd be a fair assumption.

[Susan Griffith, executive]

Yes.

[Operator]

Our next question comes from the line of Gary Ransom of Dowling & Partners.

[Gary Ransom, analyst]

I wanted to ask about UBI and its use in segmentation. I assume it's still an add-on at the end of the various other variables. And you can correct me if I'm wrong on that. But do you envision a time when that actually might begin to be more infused in the rating plan and become more of a primary variable, one that replaces other variables? I just would like a view of how you're looking about the UBI portion of the sophisticated segmentation.

[Susan Griffith, executive]

Yes, I mean, I think for us, the UBI is obviously so powerful that we would love to have it in every single auto quote. I think the issue is, obviously, it's a voluntary -- it's at a voluntary perspective. At some point, the car could be -- the car could be the place that actually tells you how you're driving, if we have that technology and the infrastructure to be able to understand that driving behavior. So right now, it's one of many variables. We think it's -- we know it's really powerful. We want to be able to give good drivers great prices. And so we'll continue to advertise on that. I wouldn't say that -- I actually wouldn't say it's necessarily an add-on, but it's not something that everyone has. So it's not like everyone has -- we're going to have everyone's proof of where their garaging address is. So I don't know if I'd call that an add-on or not. It's a variable that we think is important. And for people that want to get great rates, they're going to actually, obviously, do UBI. And if not -- or they go and they don't get great rates, they will leave and go to another competitor, which is okay with us.

[John Sauerland, executive]

So Gary, I'm guessing you're thinking from your actuarial point of view, and in terms of solving for the efficacy in terms of increasing the accuracy of the rate, it is actually solved last. Because not all of our customers take UBI, we want the greater program to be accurate for all of those customers. Then we solve incrementally, and it has a lot of incremental power, we solve that last. Would we prefer that everyone take it? As Tricia said, absolutely yes. Will there be a day when that is the model? It very well could be. Certainly, as data comes from vehicles directly going forward, and even today, data coming from handhelds increasingly. So I think there could be a day, I wasn't -- in the very near future, where it is all solved simultaneously. And that would get us the biggest benefit from the rating information. But today, we solve that -- for that last. Does that answer the question?

[Gary Ransom, analyst]

Yes, that does. And maybe a follow-up to slip into homeowners, is, are you using data from any of the Internet of Things type technologies and collecting that? Is that a useful means of segmentation in homeowners?

[Susan Griffith, executive]

Yes, we're working on that. We have a couple of people, both at Progressive Home and here at Progressive, working on connected home and ultimately how that -- how we can understand what people do and what people buy and how that affects loss cost. Probably too early to tell right now. Those would be my answer, but yes, we're definitely looking at that.

[John Sauerland, executive]

We have pricing segmentation based on the different perils and the different technologies that can be deployed to try and limit losses on those perils, so water, fire, theft. The take rate now is very, very low. So the deployment of such technology is really just beginning. And we have discounts for those devices. I will admit that the data behind the discounts for those devices is somewhat limited, based on the fact that we simply haven't had much history. But that is absolutely where our product is designed to go. And as Tricia said, we have a lot of people who are actively making sure we are a leader when it comes to deploying those types of things in our rating for homeowners.

[Susan Griffith, executive]

And I think ultimately, what you want to get to from a home perspective, because the severity is so much higher than say in auto, is if you have a device and it alerts you, you have to stop the loss or at least lessen the loss from happening. That's really the key, is if you have something and you are on vacation and you can't stop the water leakage, that doesn't really help much. It's really about understanding and lowering loss cost.

[Gary Ransom, analyst]

I wonder if I can -- can I just fit it one little one, too? How important, on the different levels of segmentation, that you're going from 8.4, 8.5, 8.6, 8.7, the speed of that. How important is that replacement, that constant replacement, to your ability to develop the adverse selection that you talked about earlier?

[Susan Griffith, executive]

I think it's critical.

[John Sauerland, executive]

It's very important. So as you know, most of our rating algorithms are filed and in the public domain. The minute we send them to the Departments of Insurance, as we mentioned in the presentations, we are also looking at our competitors' filings. So we are copied very quickly. This is, frankly, about running faster. And then we talked about the iterations of 8.7, 8.9. 9.6, I think, was where we were headed. It's critical we continue to get them out fast.

[Susan Griffith, executive]

And it's more fun. Actually, that's the way, what Sanjay said, that's the fun part of being a product manager, is you -- some of that segmentation comes from grassroots efforts, and then we look at them countrywide. And there's a lot on the table, a lot of exciting things. So it is about running faster. It's in our DNA and we'll continue to do it.

[Julia Hornack, executive]

So we are going to sneak in one last caller, even though we are past the 3:00 hour.

[Operator]

Our next question comes from the line of Yaron Kinar of Goldman Sachs.

[Yaron Kinar, analyst]

I have 2 questions. One, as you continue to improve on your segmentation and pricing, I guess one thing that's quite notable is that your combined ratio is actually improving along the way as well. And I would have thought that with the target of growing as fast as you can, 96, it would kind of be the opposite way around and maybe achieve even greater growth while keeping to the 90 -- or getting as close to the 96 as possible. So I guess my question is, is the fact that 96 is -- or that you're well below 96, is that just because you're still keeping some margins of error and you're being cautious in how you're pricing? Or frequency is below what you had expected it to be? Or is there some other deliberate driver there that would keep the combined ratio as low as it is?

[Susan Griffith, executive]

Remember, it's 96 or better. So 4 is sort of the, we have to have, and then it's better. Frequency is one input. Again, we talked about today when I wrote in my letter, mix is a big difference, too, having more preferred customers. We still have a small percentage of the Robinsons. We have a lot of room to grow on that. And we also highlight and put a lot of emphasis on our cost structure, both on the LAE side and what we call non-acquisition expense ratio. So it's a bunch of different things in play that goes into our widened margins.

[Yaron Kinar, analyst]

I understand all that, but I would have thought that, that could be priced for it to begin with. And not that I'm complaining about 20% net premiums growth in any way, shape or form, but why wouldn't it be 30% with keeping margins a little bit lower?

[Susan Griffith, executive]

Well, again, we said we look at state by state, channel by channel, segment by segment. And if we believe that -- one, the growth has been tremendous. But if we believe that there is an area where we think we could grow more handily, or we think that there's an opportunity to beat out the competition, we would absolutely consider whether limiting margins or reducing margins there to get that growth should happen. We go through that exercise all the time. So there's a handful of states right now where we have reduced new business rates in order to grow. We look at a lot of different data to understand that, conversion being one of them. Again, you could easily say, yes, let's -- if we can grow 40%, if you said to me right now, could we grow 10 more points and only lose 0.5 point of combined ratio, I'd say yes. But you just don't want to throw away that margin and not get the growth. So it's really surgical when you make that tradeoff.

[Yaron Kinar, analyst]

Okay. And then my second question, I think you've often described yourself more as an -- a technology company than an insurer. And it seems like now there is another technology company entering this space, or dipping its toe in the water, Amazon. And if I understand Amazon's mission correctly, it's basically to take margin out of businesses and out of industries. So I'm just curious to hear your thoughts as to how homeowners, or maybe even auto eventually, gets impacted as Amazon tries to build an -- a presence in insurance.

[Susan Griffith, executive]

Yes, I mean, I think we'll cross that bridge when we see it in terms of that, if we see it. Insurance is a complex product, having 51 jurisdictions. It's highly regulated, not an easy place to get into and do it well and make money. I think Pat said, "It's easy to make money, it's easy to grow, not easy to do both." So we have 80-plus years of experience. We've had ups and downs, way more ups than downs. That said, we've seen a lot of competition come along over our 80-plus years. And we believe we're well-positioned, with all the things that we've said, in terms of great brand, incredible segmentation, great people and culture, service like no other. So we'll take all of our advantages, our competitive advantages, and work to continue to grow and grow profitably.

[Julia Hornack, executive]

So thank you everybody for your time and interest today. We've exhausted our scheduled time and then some. So Chanel, I'm going to hand over the call back to you for the closing scripts.

[Operator]

That concludes the Progressive Corporation's third quarter investor event. Information about a replay of the event will be available on the Investor Relations section of Progressive's website for the next year. You may now disconnect.