The Progressive Corporation (PGR) Q3 2016 Earnings Report, Transcript and Summary

Good afternoon, everybody. Thank you for those out-of-towners who came in, and welcome to Progressive's 2016 Investor Relations Day. We're glad to have you. I think you're all familiar with the Safe Harbor statements, and if not, there's also a copy in your materials as a reference guide. So this particular investor relations meeting represents a really big change. For many, many years, Glenn Renwick was the host of these meetings. In fact, I think he started them. Last week, Glenn celebrated his 30th anniversary with Progressive, of which more than half of that he was our CEO and President. During his tenure as CEO, he tripled the company in size. So by most record books you would say that is phenomenal, and it is. But it's not just that. It's not just what Glenn did, it's how he did it: Always making sure to instill in us that desire to win, but doing it in the right way, always following our core values. He always would set the bar higher every year for all of us. And then he wouldn't walk away, he would help us get over, and that is really his legacy. So Glenn, on behalf of the 30,000 Progressive people, our shareholders and the customers we're privileged to serve, thank you for, well, being Progressive. Okay, enough about Glenn. I thought what I'd do to start off is just remind you of our segments. So we outlined these several years ago when we first started talking about our Destination Era strategy. So I'll just briefly go over those. The focus today will be on the Robinsons, and we're really excited to talk to you about what we have done since we last met you in May of 2015. So Sam. Sam is a segment we grew up with, and we call him inconsistently insured. He is who Progressive -- he put us on the map, and we call him nonstandard. He cares about rates, doesn't stay very long, but he could come back. We love Sam. We want to write all the Sams we can possibly write as long as it reaches our profitability goal. Diane. Diane is a big part of the Destination Era. We've talked in the past about her being a future Robinson. Diane is an auto customer, and she rents. So she may have renters insurance with us, and what we hope is that as her needs evolve, when she buys a home, gets married, that we evolve with her. It's the notion of Diane graduating to be a Robinson. She's our largest customer base, so the opportunity there is immense. The Wrights. The Wrights are unbundled auto and home. So they have Progressive home and someone else's auto. Again, we love the Wrights. We see them as future Robinsons because we hope that if they decide to shop their homeowners, they come with us and become a Robinson. The Robinsons. That's really what today is all about and what the Destination Era is all about, the auto home bundle. The Robinsons are 40% of the market. So the opportunity here is really great, and we are really proud of how, the momentum we've had in the last 18 months in particular, in obtaining Robinsons. Before we get started, what I thought I'd do as the new CEO is step back and just take stock of some objectives and some fundamentals that will stay the same during my tenure, but things that might evolve. And clearly, with Glenn's leadership, I was a part of the team that developed the strategy around the Destination Era. So to make a turn and do something differently wouldn't make sense. I believe in this model, the execution has been great and I'm excited about our future. But things do evolve as a team as we know more. Let me start with our first and most important objective, profit. We want to make at least $0.04 of underwriting profit of every $1 of premium. 96 is central to our culture. You're actually sitting in an auditorium that is called Studio 96. So you know how integrated it is into everything we do. It continues to be our most important objective, and that won't change. There is, however, an added dimension with the auto home bundle. We're looking at how 2 independent margins can be optimized for that package. So while the 96 will stay the same, how it will be derived will absolutely evolve. Growth. We want to grow as fast as we can with 2 constraints: one, the profit objective I just outlined; and two, our ability to service our customers. That second one is really important to us. So about 18 months ago, we had our 8.3 model on the streets, it was working, we were getting a lot more preferred customers. We also saw the opportunity, when some of our competition pulled back, to spend more on advertising, specifically, in the digital options. So we knew at that point we had a lot of opportunities to grow. We made a decision at that time to hire well in advance of need. So from the claims organization as well as our customer relationship management organizations, both on the auto side and the commercial side. And I'm so pleased that we did that, because we did get that growth. We got that preferred growth. And in addition, we had things that happened that we didn't necessarily plan for, in the form of catastrophes. So we've been able to service our customers, no matter if it was a CAT or a regular just increase in losses based on growth. It is sort of a rule here at Progressive that whether you're on the claims side or the product side, if you think we're growing too fast, you absolutely have to raise your hand and say, "No more growth. We can't do this." And we've had to do that in the past. I'm so delighted to say that this year we didn't have to do that because we prepared and hired in advance of need. So when I think about growth and profit and that balance, year-to-date, with August being the most recent time frame, we're at a 96.2 [Audio Gap] percent net written premium growth. If we ended the year today, this is a phenomenal year, especially with the business that we've been putting on the books. I'm extraordinarily happy. If we were at the same margin at 96.2 and had 4% growth with the same set of circumstances, I wouldn't be as happy. If we were at 88 with 0% growth, I wouldn't be happy. I'm very, very pleased with our growth. It really is sort of the perfect storm. So we had this new business, and we talked and talked at the conference calls about the sort of new business tax that you have, when our combined ratio is higher in that first term. And so we had that coming at us, and then we had CATs. This year, we have, on our CR, 3 points of our combined ratio are in the form of CATs, which is compared to this time last year, about 1.5%. In fact, last month, with the Louisiana storms, it was over 4 points of our CR. So you can't plan for CATs. Believe me, I'm watching Matthew closely, you can't plan for CATs. What you can plan for is how you react to those CATs. I am extraordinarily pleased with how we have reacted, specifically to the Louisiana [Audio Gap] So about 3 weeks ago, a group of us flew down and spent some time in Lafayette and Baton Rouge, Louisiana, to make sure our employees were okay, our customers were okay, and I was just amazed. I was amazed that our CAT team, who went down there and just made sure, they took it upon themselves to get our customers back on their feet. We literally had employees who were laying in bed, and they found out their home was flooded when they put their foot in water. Yet they came to work that day to help our customers. In fact, as of yesterday, we have closed 96% of those flood claims, which is really amazing because they're complex claims, mostly total losses. So again, I'm so amazed that what we were able to do and what we're able to do. And you know what, we are ready, with whatever happens with Matthew. As of today, we have a plan, and we'll figure it out. That's what we do. So for me, again, just to wrap up in that part, profit and growth, very happy, very proud, especially with how we've been increasing our preferred business, the Robinsons, and we're going to talk a lot about that today. I thought I'd take just a stab at some of the financial fundamentals. So it won't change. I think they work with our business plan. We want to continue to balance operating risk with the risk of investing in financing activities, in order to have sufficient capital to support our growth. So that's the overview. On the operating side, we want to maintain pricing and reserving discipline. So when we talk about our pricing, we talk about managing targets. We talk about targets at the lowest definition. So whether it's on the agency and direct side, product, state, we really segment our targets. And that's another reason why I'm so positive about this new business because we look at the new business coming in and say, "Is it at or below targets? Great, that's going to run off profitably." So we're excited about that. We want to have premium to surplus at or below state minimum requirement. So on auto, that's a 3:1 and about half of that for home. And on the loss reserving side, we want to make sure our loss reserves are adequate and developed with the minimal variance. Year-to-date, very proud of that. We are pretty close, but a little bit on the favorable side, which is the side you want to be on. On investing side, we want to maintain a liquid, diversified and high-quality portfolio. We manage our investments on a total return basis, which is very consistent with how we report our comprehensive income. And on the risk side, we want to manage interest rate, credit, prepayment, concentration and extension risk. Lastly, on the financing side, we want to maintain sufficient capital to support the insurance operations. That's measured by maintaining a debt below 30% of total capital of book value. So we're just slightly over 28% right now, and that's after issuing $500 million worth of debt in August. We want to neutralize dilution from equity-based compensation in the year of issuance, and we do that by repurchasing shares. And finally, we will invest in capital in expanding the business when it meets our objectives. A perfect example of this is our acquisition of ARX. We didn't necessarily feel like we had to have a homeowners company. What we felt is we had to have access to that addressable market. That is a $300 billion market with auto and home. We needed to have access, and it just so happened that we had been working with a company that we had a high amount of respect for, that had a very similar culture and so was a perfect relationship to start. So we're very pleased with our use of capital in acquiring ARX. Any use of underleveraged capital, we will buy back shares or issue dividends, whether special or variable. So let's get to the 96, since we are slightly over at this point, I want to walk you through kind of how we look at it. This is just a basic refresher. So everybody gets to a 96, all of our customers at some point get there, but sometimes in vastly different ways. On the direct side, you can see we run a much higher combined ratio in that first term. That's based on the fact that we have a little higher loss cost and some underwriting cost, but most of all, we front-load all of our acquisitions cost on the direct side in that first term. On the agent side, we have a little bit more losses, a little bit more underwriting, but it's a relatively fixed commission expense, so much lower in the first term. So let me take you through on the direct side, 2 ends of the spectrum. You have Sam and the Robinsons. So you can see the same dynamic. They both rendered a loss in the first term. But what is not evident in our monthly reporting or even our quarterly reporting is the fact that this very high-retaining business, the Robinsons, who stay 3x, actually, more than 3x longer than the Sams, are expanding our invisible balance sheet. So bringing on more of those Robinsons really build the intrinsic value of Progressive. Here's a basic income statement, and it's directionally accurate. So if we look at again, the Sam and Robinsons, you look at revenue. Clearly, the Robinsons are higher, based on the fact that they stay longer and they have higher average written premium. If you look at the operating expenses, which, of course, are losses, LAE and expense ratio, you quickly notice that neither of these equal a 96. Those numbers represent the nominal value of the cash flow, so they're under a 96. Our lifetime target, 96 combined ratio. For our direct business, we discount the cash flows to recover our acquisition expenses. So that will be different in different segments depending on how long they stay. The difference between those, of course, is our underwriting income. And then there's other income that we have, both on the investment side for both Sam and Robinsons, and the commissions side for just the Robinsons. So whether it's affiliated partner, like ASI, or other unaffiliated homeowners carriers, we get commissions when we have those bundles. So as you can see, the value of a Robinson, a pretax value, is over 4x that of a Sam. Again, we love Sams as long as we can make money on them, but the value of a Robinson is extraordinary, and the fact that we have grown Robinsons substantially in the last 18 months really does grow the value, the intrinsic value, of Progressive. So for years, we have shared with you the invisible balance sheet. We talked last year about unearned lifetime earned premium. So when we think about business coming on board, we look at new apps, for example. So in this year, for Wrights and Robinsons, we've grown new apps 30% and 40%, respectively. That's pretty great. We look at average written premium. Our average written premium has come up across all segments. And then we look at PLE. This is kind of going to be a spoiler alert for Andrew's section, but we finally have made a breakthrough on retention and PLE. So to us, those 3 things are really exciting. So we thought we'd share with you one other metric that we look at internally, that we call ULUP, unearned lifetime underwriting profit. So you'll see that the colors represent a segment, so green for the Robinsons, pink for the Wrights, et cetera. On the Y axis, you see ULUP in millions, and on the X axis, you'll see the time frame with which we are looking out. So right when we started to grow, right around March of last year, to our most recent data point. As you can see, the majority of our segments are growing substantially in ULUP, which, again, is building that intrinsic value. In fact, the Robinsons are growing over 60%. The chart underneath shows all of the segments' year-over-year growth in ULUP, and again, all are growing, but you can see the Robinsons are growing at a really quick clip. That's exciting for us. I'm going to shift gears now and talk about how important understanding the Destination Era is to the 30,000 Progressive people that work on getting our customers in the door and making sure we nurture them along the way. We look at it in terms of different frameworks. This is a framework we use called our strategic pillars. So let me talk about the first one, our culture. Our culture is so important, and it's not something you can really adequately describe unless you've spent time at Progressive. It's not something you're going to see on a spreadsheet. But it is really special, and it's really different, and we know it's a competitive advantage. It's really at the root of what I started to talk about, and that is doing the right thing. And we want to win, but you always want to do the right thing, and it's really rooted in our core values. I have the privilege to speak at every new hire class, actually, it was that one Monday, with a couple hundred new claims reps, and my sole topic is our core values. I talk about 15 or 20 minutes, use some examples and then let them ask questions. It's so important to me for them on their first day or their first week or their second week to know what an important part that is of Progressive, our values. Earlier this year, many of us went out to celebrate a milestone at Progressive, $20 billion in earned premium. And we got out face-to-face with over half of our employees, over 15,000 people, just to celebrate and say thank you for all that you do, and then of course talk about the next milestone and the next milestone, and it was really rewarding, and that again is a big part of our culture. Our brand. We want to be the brand people want. So whether it's awareness or preference or consideration, those things are important, but what we really want is to be able to, when people buy an insurance product from Progressive, that they feel confident that they've made a right decision and continue to feel confident as we service them and add more and more policies. We want to spend money on consumer marketing and customer marketing. We've talked often about how we want to have our customer marketing be as savvy as our consumer marketing. Andrew Quigg is going to talk a little bit about that today, but we continue to resource that, as it is an important part of nurturing our customers. Competitive prices. In a business with small margins, this is very important to us, and it's really all around segmentation. You're going to hear a lot about that today. So John Sauerland is going to talk about our continued evolution with UBI and our segmentation, and how important that variable is. Pat Callahan is going to talk about segmentation and the product models that we continue to elevate as well as our bundled competitiveness. Andrew Quigg will talk about segmentation on the customer part and using data to get out in front of our customers to know what they need, when they need them. So this is a really important part of our -- really central to things that we do, is segmentation. We'll continue to hone in on that. Another part about competitive prices is really having a competitive loss structure. So on the claim side, we continue to have -- try to have that near-perfect balance of cost and quality, so pushing down LAE, while pushing up accuracy, and we're doing a fantastic job. On the expense side, we look at what we call non-acquisition expense ratio. So we continue to try to push that down, be more efficient and then give that pricing advantage to our customers in the form of competitive prices. And the last pillar is meeting the broader needs of our customers: What products do they want, when do they want it, how do they want to shop. We want to make sure we can have personalized service and products for each of our customers. So if you want to buy through an agent, get service online or through a chat, we want to make sure if it fits your needs, we are there for you. We want to pick up the phone and say, "Yes, we can do that, absolutely." And whether or not we do that through our own products or on affiliated products, we want to make sure we meet the needs of our customers. We use this framework for how we're going to grow the business: Acquire, anchor, bundle and extend. We see growth in 3 ways: one, selling products to new customers; two, selling additional products to our current customers as the need arises; and three, keeping our customers longer. Let me start with acquire. Acquire is that initial relationship with us. So whether you come in through a special lines product or an auto product, we're just getting to know you. And in fact, we are pivoting the company and trying to be less about policies and more about customers and households. We're currently in 1 in 10 households in the United States, and we would like to double that in short order. And so we're doing that through acquiring more and more customers and understanding their needs and then evolving as their needs evolve. Anchor. Our anchor products are our auto and our home. If you trust us with those assets, we believe that our relationship is much deeper. I started today talking about the opportunity, the $300 billion, almost $300 billion opportunity in auto and home. Now that we have access to that and have more and more customers coming in either at home or auto, we know we can deepen that relationship. It's such an important piece of future Robinsons. Bundle, that's what we're going to talk about today. Our bundle are our Robinsons, our auto and home together, and just even a little bit of increased market share on the bundle part really has a big impact to the bottom line. Lastly, extend. We have been talking about retention for some time, and we have made a significant breakthrough. Last year, if you recall, I announced that I asked Andrew Quigg to lead a team of people to try to increase our retention by 50%. He's going to talk about that today, but we're extraordinarily happy with our results. For me, extend isn't about talking any longer about customers who stay 1 month or even 1 year. It's about having customers for decades. In fact, 1 out of 6 of our customers after the first renewal are decade customers. So we're pretty pleased with that. About 1 month or so ago, some of us went out to observe some focus groups of our customers. So we listened to Dianes with renters, Dianes without renters to try to get some insight. We went out to Minneapolis and Dallas, and we listened to some decade customers. And they came because they love the brand and they love the price, so that made sense. They had stayed because over that decade or more that they'd been with us, they'd had some sort of service, whether they'd had a claim or they need to call in to change a vehicle or deductible, and they were extraordinarily happy with the service. But what we heard from both of the decade customer groups was something I think really insightful. People said, "Why are you going to stay?" "Even if rates go up, will you stay?" Yes, because I'm proud to be a Progressive customer." And that really hit us. And we heard that over and over again, and so we're working on bottling that feeling and having it happen earlier in someone's tenure, have them feel that pride year 3, year 4, so we can continue to extend that and increase our decade customers. And in fact, we're looking at different metrics. At some point, we'll continue to have policy life expectancies, but why wouldn't we have customer life expectancies? So that's where we'll evolve as our customers evolve. Ultimately, our goal, our vision, is to become consumers' #1 choice and destination for auto and other insurance. The next 4 speakers are going to talk about how we're going to do that on both the auto side, and John Barbagallo will talk about how he's going to do that on the commercial side. So now I'd like to start and invite our Personal Lines President, Pat Callahan, up to talk about agency growth, our product and our bundled competitiveness.

Thanks, Tricia. So last year, when we talked, I laid out 3 key strategic priorities for our Personal Lines business. And those are: Restoring growth in agency, increasing our bundled penetration and ultimately, and then advancing our product development cycle. I want to give you updates on all 3. And what we'll do is start with an overview of the market from an overall channel and market size perspective. As Tricia mentioned, U.S. personal auto and home is rapidly approaching a $300 billion market. And when you look at the independent agency share of that, it's a little more than 1/3, or about $100 billion of home and auto premium. And despite the fact that technology adoption has been rapidly increasing the growth rate for direct distribution of Personal Lines products, the independent agency channel continues to grow premium. And that's due in part to broad distribution of their footprint, but also a great value proposition from a breadth of products offered and depth of carrier choice available. When you look more specifically at the homeowners market in the U.S., it's about a $90 billion market, and that's about 90% sold through agents, more than 40% through independent agents. And as the #1 writer of auto insurance through the independent agency channel, we believe we're incredibly well-positioned to leverage our strength in the channel, combined with ASI's property offerings, to become the #1 writer of auto and home insurance through the agency channel. Last year, I told you that restoring growth in agency was a top priority for myself and lots of other folks within our Personal Lines organization. So today, I'm thrilled to show you some information that after our business shrunk slightly into 2014 and barely recovered into 2015, we've turned that agency business around from a size perspective. And at this point, our agency auto business, as well as ASI's homeowner business through independent agents, have both never before been larger. And if we zoom in on what's driving that auto improvement, we look primarily to start with at new business apps. New business growth is the lifeline or lifeblood of a growing organization like Progressive. And as you see on the screen, midway through 2015, we turned new business positive on a year-over-year basis within the agency channel. Our new business growth continues to be up in the double digits on a year-over-year basis ever since. New business growth originates from 2 primary metrics. The first is demand, so quoting. The second is conversion. And our improvements in the agency channel are driven just about by equal parts of both, which indicate to us: number one, we have a more competitive product in market; but number two, that more competitive product is inducing agents to come to us more frequently to meet the needs of their clients. An absolutely fantastic combination within the channel. But to understand how competitive we are overall in the channel, the best benchmark that we can use is how well we're converting on comparative raters. So last year, I shared with you that the majority of our agency quotes start on a comparative rating platform. And just to level-set, comparative rating software, desktop applications in agent's offices, that run a real-time, risk-by-risk auction amongst all the carriers that participate within that agent's office. And you'll notice that similar to what we saw from a new apps perspective, we're seeing increasing new business from a conversion perspective on comparative raters. Turned positive around the same time that we saw new business apps turn overall, and continues to be up solidly in the double digits. The exciting part about this for us, though, is competitive rates absolutely matter. But competitive rates are not sufficient to drive conversion in the agency channel. Because of the plethora of choices that agents have, you also have to have competitive service, competitive ease-of-use and ultimately, competitive compensation. So we believe comparative rater conversion is a fantastic benchmark for us to truly understand how competitive we are overall in the agency marketplace. Last year at this time, I also shared that we had just introduced 8.3, our latest auto product model. And while the early results were encouraging, we were in a couple of states and we were seeing significant growth overall from the product model. At this point, about 18 months later, I'm thrilled to share that, that growth continues to come in at expected levels, but it's highly concentrated on the more preferred end of the spectrum. So our growth rates with Wrights and with Robinsons are significantly higher than they are in our other segments. These segments are absolutely foundational to how we win in the Destination Era, and our product design and product development is helping us get there. But we all know that Progressive is known for new business acquisition. We do a great job. We probably do the best job in the industry. But the reality of it is, sustainable premium growth doesn't just come from acquiring new customers. We have to retain them. And I'm pleased to show that when we look at retention from the agency channel, we see after a dip in early 2015 through about the middle of the year, we've been increasing our agency policy life expectancy month-on-month ever since. And Andrew will talk a little bit more about more details around decomposing that, but the drivers of that are twofold: number one, it's mix, so our mix of new customers has improved. We refer to that as nature. But equal parts of improvement in our PLE are coming from experience improvements and relationships within our customer relationship management organization. Those investments are paying off, and Andrew will tell you a little bit more about that later. What's particularly exciting to me though, when I look at our retention improvements, is to overlay our rate activity during the same period of time. So your orange bars are annualized rate change within the agency channel for each month during that same period. What you'll notice is during the period of time that our retention was improving, we were raising rates, somewhere between 2%, 2.5% to 4%. And we do that to keep up with trend and ensure we continue to deliver our target margins. But the combination of being able to improve retention while also increasing rates indicates to us 2 things: number one, market conditions are fantastic for growth in the agency channel and we are really, really well-positioned; but number two, given the plethora of choices that agents have to move business to, when rates go up, the fact that our rate increases are sticking indicates to us we've got stronger relationships that are forming in the agency channel, and that's a critical element of our long-term destination success. Insurance is a trust-based business. And trust is a really difficult thing for us to measure, unlike many of our other metrics. But we look at actions as opposed to words to understand how we're performing from a trust perspective. Number one, are agents trusting us by placing their better customers with us? The answer is yes. Number two, are they trusting us to interact with those customers on a regular basis? The answer is yes. More agents are opting in to our joint marketing programs for additional protection, quote resolicitation, et cetera. And number three, in a recent survey of thousands of independent agents, they responded that increasingly, Progressive has all the products to meet the needs of their preferred customers and more of those agents responded that they're placing their preferred customers with us. The turnaround in the agency channel was driven by a combination of product and stronger relationship building. And we fully expect that the momentum we've established to date with this turnaround will carry into our second priority, which is bundled penetration. Growing bundled penetration is the key performance indicator of our Destination Era. Making progress, selling both home and auto to our combined customers is essential to our long-term success. And the things you need to drive bundled penetration are: number one, available and competitive property insurance; number two, available and competitive preferred auto rates; and number three, the systems, tools and experiences that enable customers and agents to find, quote, buy and ultimately own a Progressive policy or a Progressive bundle. We'll stay with the agency channel for now and talk about where we're headed from a bundled penetration perspective. Last year, I shared that we were in the process of deploying ASI's property product country-wide. Since we last talked, we've added 9 states, and by the end of this year, we'll be in 39 states, representing about 95% of U.S. homeowners, great coverage for where we are in the rollout. We go to market from an agency perspective with our property product through 2 tiers of agencies. Those 2 tiers of agencies are Progressive Advantage agencies, Progressive Home Advantage agencies and our Platinum agencies. The 2 tiers, Progressive Home Advantage a little more broadly distributed, they are our original property agencies. And Platinum is a much higher tier, more highly selective tier of agents and a much stronger partnership with us. I shared with you last year that Platinum is a brand-new program that we're introducing last summer that combined consumer benefits of annual policies, more competitive rates through multi-policy discounts and agent benefits, such as easier-to-use systems and higher compensation. We continue to deploy the Platinum program as well. Platinum, by the end of this year, will be in 34 states, which represent about 80% of U.S. homeowners. And because preferred auto and property availability are prereqs to the Platinum rollout, Platinum will always trail the rollout of our property product. But overall, we intend to continue pushing on a country-wide basis until we have full deployment of preferred auto, available property and Platinum on a national basis. But with Platinum, it's less about the states that we're in and much more about the agencies that we're in within those states. So over the last year, we've identified, selected and ultimately, onboarded more than 1,000 agencies into our Platinum program. Based on those agencies, access to preferred customers, their bundled quote and sale processes within the agency and their willingness to commit to placing a significant portion of their bundled volume with Progressive. But to be clear, Platinum rollout will be a metered rollout. Platinum has a scarcity and a scarcely distributed offering. So agents will be actively managed by our sales reps to ensure they're using the Platinum program and understand it and are using it for bundled business within their agency. So we will continue to grow Platinum, but we will grow Platinum as agents are using it effectively in market. I also shared last year that because ease-of-use for bundled quoting is a key element of our Platinum offering, we were starting to build a new Platinum quoting program. And earlier this year, we launched Platinum quoting which brings to the agent's desktop for the first time, bundled quoting of Progressive's auto and ASI's home, which enables the agent to ultimately save time, share data entry screens and external data fill and have more competitive messaging around the bundled economics available to the agent and making it easier for them to message those bundled savings to the customer. Now to be clear, this is a pilot. It's in 3 states today. We've learned a bunch to date, we'll continue to learn from the pilot, and ultimately, we'll build those enhancements into our quoting systems over time. But the billion-dollar question on all of our minds, and likely all of your minds is, is Platinum working? We've attacked the preferred market in agency a few times before, and we've learned from those attacks. We've learned from those opportunities that ultimately, for us to introduce a brand-new preferred offering into some of the best agencies in an incredibly competitive independent agency channel, we would have to change agent behavior, which we all know isn't simple, and frankly, it sometimes takes time. So to understand whether Platinum is working, we look at the 2015 cohort of Platinum agents, those that were appointed as Platinum agents during 2015. Think of it as the Class of 2015, if you will. And we look at 2 key performance metrics for that group of agencies relative to the rest of our property insurance agencies. What you'll see on your screen is 2 lines, the blue line is our Platinum agencies, the orange line is our PHA agencies. And the metric here is preferred auto production. What you'll notice between those lines is, number one, the blue line starts above the orange line. Not surprisingly, more productive agencies were selected to be Platinum in the first place. But you'll notice that after the vertical red line, there's an increasing separation from a production perspective between our Platinum agencies and our non-Platinum other property agencies. This continues into our second key performance metric, and that's new business home and auto bundles, where you'll see exactly the same pattern. Blue, Platinum agents, started above our PHA agencies, and the gap continues to widen. As our sales reps message the benefits of Platinum, our agents better understand the benefits of Platinum and for their customers, they start to experience some of the benefits that come from buying through a Platinum agency. The gap between that blue and orange line is what we refer to as the Platinum difference. And that Platinum difference is not just present with the agents we appointed during 2015. If you look at the entire network of more than 1,000 Platinum agencies compared to our non-Platinum property agencies, you'll see that our Platinum agency plant is responsible for creating more than 3x the number of bundles per agent, per week than our non-Platinum agencies, which absolutely tells us that Platinum is working. What you'll also notice on this graph, though, is the trend is upward, meaning we're improving over time. Numbers bounce around from month-to-month, but we are on an upward trend. And the guideline on the screen indicates where we expect the program to be ultimately from a success perspective. You'll notice we're not there. And that gap exists, but we know what we're going to do to close that gap. The primary driver of it is our bundled competitiveness. Well, not surprisingly, when we were launching our program, we started collecting data from the market as to where we were competitive and were not competitive. And the great thing is with 1,000 agencies, using their quoting systems to quote ASI property and Progressive auto, we are getting a ton of data on where we are and where we are not competitive. We use conversion as our primary metric, and when you look at states that represent about 1/3 of our premium, we are highly competitive on the monoline auto, the monoline home and on the bundle for an individual risk. In those agencies, our focus is to sell what the customer is looking for. If they're looking for home, sell them home; if they're looking for auto, sell them auto, or try to sell them the bundle, because we are highly competitive. However, in about 2/3 of the country right now, we are highly competitive on monoline home or monoline auto, but not yet as competitive on the bundle for individual risks. This shouldn't come as much surprise given we had a highly successful monoline auto business that we combined with a highly successful monoline home business. And the overlap between the 2 isn't quite where we need it to be. But the bottom line is we know exactly what needs to be done, and we're collecting a ton of data that enables us to map out how to integrate our products more closely to make sure that we're more competitive at the individual risk on a country-wide basis. So for us, the momentum that we saw as we were bringing our agency turnaround ultimately into bundled competitiveness, combined with the infrastructure investments we continue to make in the agency channel and early success of what we're seeing from a Platinum perspective, gives us a ton of confidence that we are absolutely on the right track from a bundled penetration perspective and very, very well-positioned to attack that $100 billion independent agency home and auto opportunity. With that said, we have another highly successful auto distribution channel and that's direct. So I want to update you on where we are from a bundled competitiveness perspective within the direct channel. To be a direct customer's destination for auto and other insurance, we not only have to have all the products they need throughout their lifetime, but they have to be aware of them and understand that we have them available before they have a life event that triggers demand for that product. So they come to us first as opposed to shopping in market. We generate awareness for our products with our direct customers 3 primary ways: Mass media advertising, it reaches not only our current customers but also reaches future customers; point-of-sale messaging, and this is both in our sale systems as well as our servicing systems, so both current customers and potential customers who visit to get a quote understand the breadth of products available and the economic benefits of buying through us; and finally, targeted marketing. We have a ton of monoline customers that we are now offering the bundled package and the complementary product through messaging the savings and ease-of-use benefits of bringing more of those products to Progressive over time. These awareness investments, coupled with a general shift in consumer buying trends towards direct and some internal messaging that gets our employees thinking holistically about protecting our customers' household needs, continues to drive up demand for direct property insurance, both online, on the phones, with current customers and with future customers. But great demand and increase in demand for our direct property products creates no value unless we can meet that demand with adequate supply. And frankly, in early 2015, our ability to meet the demand for direct property was less than we wanted it to be. Only about 4 in 5 customers coming to us for direct property could even get a quote. Part of that is just endemic with the property insurance business. Unlike Progressive, there's virtually no property writers who have similar broad acceptability. Property writers have a relatively niche focus and some may not accept 20%, 30%, 40% of available risks. That's an unacceptable outcome for our direct customers. So what we've done over time is worked, number one, with our carrier partners to expand acceptability and improve routing of traffic online and through our partner call centers to improve availability of direct quoting. But what's particularly exciting on this chart is the yellow line. The yellow line represents our Progressive Advantage Agency, which is our in-house agency that sells multiple products from a property perspective across multiple property partners. Over the last 1.5 years, we've made significant investments to expand the footprint of that agency by adding 14 states and adding more than 100 state and carrier combinations in order to drive up availability of this, of our property products. At this point, we can meet the needs of about 95% of current or future customers who contact our agency. And we believe that agency model is a superior shopping experience for anyone shopping direct-to-consumer property insurance, because it combines the technology enabled ease-of-use of a direct writer with the breadth and depth of an independent agency. We believe strongly in that model and we continue to invest to expand capabilities within our Progressive Advantage Agency. In the last 18 months, we more than doubled staffing and when calls come in, we route an increased percentage of those to the agency, to the point that at this point, we've more than tripled the sales over the last 18 months through the agency. We will continue to invest in that space and continue to use this superior business model to delight our direct customers from a bundled penetration perspective. But for us, bundling is not just about home and auto. It's also about home -- renters and auto. So in our world, we enjoy an advantaged position with younger, simpler needs, auto insurance customers, many of whom don't yet own a home. So having renters insurance available for those customers enables us, number one, to establish a property insurance relationship with customers and to be able to meet a broader need of those customers earlier in their insurance shopping journey. Renters insurance will never be a massive business for us. It's only about a $4 billion market. But for us, it is a strategic investment, because renters insurance is absolutely a gateway property product towards becoming a bundled home and auto customer with us. We have evidence that those who bundle renters and auto with us are significantly more likely to keep their auto with us and convert that renters to a homeowners policy if and when they eventually buy a home. So from a graduating Diane perspective, this is a strategic play for us and one we continue to invest in over time. We are absolutely thrilled with the acceleration of our penetration from a bundling perspective within our direct book of business. Today, more than 1 million direct customers trust us to provide property protection. And the key performance metrics for our direct property business, increasing quote starts, bundled new apps and bundled PIFs, are all up more than 20% on a year-over-year basis. And with the investments we continue to make to drive demand and then meet that demand with ample supply and great experiences, we have every confidence we will continue to accelerate bundled penetration within the direct book of business. The third priority that I shared with you last year was to continue to advance our science of risk segmentation. Customer segmentation, matching rate to risk and pricing accuracy are in our DNA and absolutely a key source of competitive advantage for Progressive. It's been that way for decades and one of the things that truly differentiates and distinguishes our offering. As we look to invest in rate accuracy and pricing segmentation, we recognize that the value it creates, through more competitive new business pricing, which lowers our customer acquisition costs on a fixed advertising base, it increases our retention even if customers shop, they can't replace similar coverage at a similar or lower price and it creates adverse selection within the marketplace, a fantastic trifecta of opportunity for Progressive. So when we think about the product development space, we focus on 3 areas, the first is rate accuracy, the second is coverages and the third is improving customer experience to drive policy life expectancy. We build bundles of these enhancements in the product model and deploy them like building blocks, stacked on top of all of the segmentation that existed in the prior model. Our pipeline from a road map perspective is absolutely robust from a product development standpoint. We're currently on 8.3 in the majority of our states. 8.3 was designed to be more competitive on preferred households and multiproduct customers. You've already heard how that's performing in market. We are thrilled with how it's working. 84 is deploying currently. Tomorrow morning, it'll be our seventh state elevating for 84 and 84 brings with it more stable rates for our renewal customers as well as advanced segmentation based on billing, claims experience and vehicle history. 845 is 84 with some additional protection for transportation network company drivers, Progressive customers who happen to work for an Uber or Lyft. And finally, 85, which is under development right now and will launch in early 2018, will bring with it much more competitive rates for households with usefuls as well as -- well, I'm not going to tell you what else is in 85, you're going to have to come back next year to learn more about that. But rest assured, we are not taking our foot off the accelerator from the product development perspective and intend to continue to lead from the segmentation perspective. But the best segmentation in the world creates no value if we can't get it to market quickly. And while we've always been known for our nimbleness and ability to deploy our products to market quickly, we recognize from a core value perspective that excellence is doing things better today than we did yesterday and we continue to invest to advance our speed to market. So in the last 24 months, there's been a joint research and development and IT team, reengineering our process of product development and product deployment. And since we last talked in May, they've been able to double the annualized rate with which we can deploy our product innovation into the marketplace. So what that means is when we talked in May, we had 3 states on 8.3. Today, we have 32 states. As of tomorrow, we'll have 7 states on 84. When we talk next year, we'll have 38 states on 84 and 85 will be spun up just about to hit the market. This lethal combination of product innovation and speed to market enables us to widen our lead over the competition. And when we look to assess how well we're doing relative to the competition, we start by looking at some industry data. The industry data you're seeing on the screen is loss in LAE for PCI, essentially about 2/3 of the market, 9 of the Top 10 carriers in the orange line, and then Progressive in the blue line. What you'll notice is that environmental activities, be it economic recovery, gas prices, severity on low price in the cars due to safety equipment, trend in the industry has been climbing since the middle of 2012. But you'll also notice that the blue line, the progressive loss in LAE, we've been competing in that exact same environment, and the reality of it is our results have been rock solid. The difference between these 2 lines is product segmentation. It's our ability to better understand what risks we're insuring and better sort out which ones are high cost from low cost. What that means in the marketplace is those in the orange line, the rest of the market, as their loss costs are climbing and they don't know which risks are driving those loss costs up, they raise base rates. And when they raise base rates, the best customers in their book of business shop and they can often find a lower price with companies like us that are better segmented. Run this cycle over and over enough times and they're left with residual risks that they're still underpriced on and priced dramatically higher than the market on any of the better risks that we insure. What it also means, though, for us is today, we enjoy about a 10-point gap in loss in LAE relative to the market and when you combine that with our technology-enabled expense gap, specifically in the agency channel, what it enables us to do is combine incredibly competitive pricing in market, which drives the growth of the business, with wider underwriting profit margins than the overall industry, delivering on our profit targets. A great combination for an insurance business. Now I also told you last year that emerging data sources were an opportunity as opposed to a threat for Progressive. And since making our majority investments in ASI or ARX, we've been building a massive data set of home and auto risk characteristics and home and auto loss characteristics. What this has enabled us to do is improve model accuracy, identify new segmentation and reduce our rating costs. An early example of how we're leveraging this data set was an auto rating variable that was highly predictive of home loss frequency, to the tune of about a 30% difference in pure premium. As we identified that, ASI quickly jumped on it and deployed it to market. So it's providing a segmentation list as a bundled household that they wouldn't normally have had absent the joining of this data. And to be frank, we are just getting started digging into this massive data set and fully expect that the joint R&D team between ASI and Progressive will leverage this treasure trove of information to identify new ways to assess and ultimately price and improve rate accuracy for our household bundled pricing. So last year, I shared with you some fairly aggressive goals for our overall business. With the turnaround in our agency business as well as great performance from our direct business, coupled with the investments we're making to increase bundled penetration across both our direct and our agency business, our robust product development pipeline of future segmentation enhancements and the faster speed to market on top of already leading the industry, I am absolutely confident that we are incredibly well-positioned to become consumers' #1 choice and destination for auto and other insurance. With that, I'll bring John Sauerland on to talk about Snapshot and data management.

Thank you, Pat. Good afternoon. It's been a while since we gave you a deep dive on Snapshot, our usage-based rating program. We have a new model in the marketplace, I'll share results around that, but also share more holistically why Snapshot is so powerful for Progressive. Later this year, we will roll out a new delivery mechanism to experience for Snapshot and that is via mobile. I'll share the rollout plans there, we're very excited by that, but I'll also share what our expected benefits from that rollout are. But overarching all of this is something that I think is even more powerful than Snapshot, sort of an ancillary benefit from Snapshot, if you will, and that is the big data management and analysis capabilities that we've developed around Snapshot that are now paying dividends across the organization and I believe can be a source of competitive advantage going forward. If you're familiar with Progressive, you know that we have a history of innovation. I would say that Snapshot is as a result of that history of innovation. We started out many years ago installing this football-size device in the trunks of some cars in Texas, proved out the concept a little bit. We enjoyed the benefits of Moore's law, shrinking of technology. I mean, we're able to get the technology into a device that could plug into an OBD port on a vehicle. Unfortunately, back then, connectivity was an issue. We mailed this device to our customers, along with a cord that we asked them to go find. Six months after plugging the device in their car, plug that cord in the device, into their desktop, more than likely back then, by the way, and more than likely try and upload via their dial-up modem. Obviously, not a big hit. What was a big hit was when we were able to get the technology and especially the cell chip small enough and the cell transmission costs low enough so that we had virtually real-time transmission of the data. In today's world, we sort of call that the Internet of Things. Back then, it was just via cell towers. Soon, we're going to be moving to mobile. I will talk more about that in a bit, so I won't spend time here now. The ultimate model for Snapshot is not via device, not via your phone, it's directly from the vehicle. Really happy to say we're actually in market today with that model with General Motors. General Motors is the only car manufacturer, to my knowledge, with that capability today. As you all know, technology takes a long time to move into the fleet. This won't be a broad-based model soon. But nevertheless, we're excited to be in partnership with GM, in market, and I'll share a bit more about that in a moment. If Moore's law allows you to shrink technology, I believe, there's got to be a corollary to Moore's law that says that data coming out of that technology doubles every 2 years or even every 1 year. We now have 2.2 trillion records in our database from Snapshot, 2.2 trillion records. If you think about those records as rows on a spreadsheet, those rows are widening. The blue area on the chart here is GPS data. While we don't today rate on GPS, we are collecting it for research purposes and it widens the data field, if you will. When we introduce mobile, we'll be collecting even more information. Mobile will take that width of that record 20-fold from where it is today. The data explosion here is incredible. It's been a challenge, it's also a huge opportunity. We'll talk more about that in a moment. But let me take a step back first and make sure I have everyone grounded in Snapshot. We have 2 models in the marketplace today. We're not great with naming, it was the 2.0 model and the 3.0 model. The 3.0 model has rolled out with 8.3. So in the majority of the country today, as Pat Callahan just mentioned to you, in the 2.0 model, we didn't give you a discount when you opted into Snapshot. You came to Progressive or an agent, agreed to the Snapshot program, 30 days later, we give you a discount based on your 30 days of driving behavior. Drive for 5 months more, we lock you in for the rest of your tenure with Progressive. 3.0, we offer a participation discount. It varies by customer segment, we'll talk about that in a moment, then you drive for 6 months. Here again, we lock you in to your discount or a surcharge. So that's a big change with the 3.0 model. With 2.0, we're very concerned about establishing consumer acceptance for the concept because we have been the leader in UBI. We've gotten consumer acceptance, we're in a place now where consumers are willing to opt-in to the program, as you'll see, even if there's a chance for a surcharge. So in the 2.0 model, we had up to a 30% discount, 0% surcharge. In the 3.0 model, we have up to a 20% discount and up to a 10% surcharge. So similar range, very different model. In 2.0, about 60% of our customers who partook, got a discount. In 3.0, about 75% of our customers get a discount and in 3.0, more than half of our customers received more than a 10% discount, which is pretty material in the auto insurance space. So giving you perspective on the different models, keep that in mind as we talk about results. And I'll also give you a little bit more perspective on the distribution of our customers' driving behavior. The results. They're good. New apps are up, I can't attribute all of this to Snapshot, however. As I mentioned, the 3.0 model rolled out at the same time as 8.3, we made a lot of changes here, we can't tease it out completely. But new apps are up, and this is just Snapshot apps on the graph in front of you. One of the things I hope you will notice is that agency apps are up more. We've been challenged around agency adoption of Snapshot for many years now, frankly. The inclusion of an upfront discount, a participation discount, helps us a lot in the agency channel. As Pat told you, it's largely driven by comparative raters, somewhat of an auction environment, the more we can get a more competitive rate up there at new business, the better chance we have of getting customers into Snapshot. I mentioned those discounts, the participation discounts vary by customer set. Ranges from about 2 percentage for Sams, up to about 6-ish percent for Robinsons. We define that participation discount not based solely on the customer marketing target there, but that's the average of the discount for each respective segment. Not surprisingly, conversion is up more on the preferred end of the spectrum. Even with 2.0, Snapshot customers converted at a higher rate than non-Snapshot customers. So being interested in Snapshot improves your conversion rate, all else equal. And in 3.0, we obviously see that continuing, but we also see a greater disparity, not surprisingly, on the more preferred end of the spectrum, where the participation discounts are higher. When we look at new apps by customer segment, here again, the mix skews a bit towards the preferred end. But what I hope you also notice is that the slope, if you will, of that relationship in the agency business is higher, it's steeper. We're getting more Robinsons into this business, into the Snapshot program. Pat and Tricia told you, the Robinson segment is the largest segment available to us in the Personal Lines space and it's our lowest market share. So to the extent we can get more customers in here, that's fantastic. Something not necessarily specific to 3.0, but I think representative of where we're at today when it comes to usage-based rating, is the fact that it is broadly acceptable. And I'm just sharing you one example here and this is age. So when you look by customer segment, the percentage of folks 18- to 22-year-old who opt-in to Snapshot is about the same as the percentage of folks who are 55 and above. Not 100% acceptable in the marketplace for sure, but broadly acceptable. So hopefully, I've convinced you that Snapshot is helping us on preferred, it's helping us grow. I often get the question, and even got it in the lobby before this, is that can you tell me how is it helping? Put numbers to that for me. Some facts. The most powerful rating variable we have ever seen and with 3.0, even more powerful than 2.0. So our analysts are working hard on improving those algorithms. 3.0 even more predictive of losses than 2.0, and 2.0 was way more predictive than any other rating variable we've seen. Another fact. We provided more than a $0.5 billion of discounts to customers with Snapshot inception to-date. More than $0.5 billion worth of discounts. Those folks think, well, that seems to be a benefit for a customer, how is that a benefit for Progressive? To help answer that, I'm going to take you back to a chart that we showed you actually in 2012. This chart shows the loss ratio and the retention rate across discount buckets, if you will, for Snapshot customers. So we have 0% to 30% discount along the horizontal there, and to the left of that 0, we have non-Snapshot customers. So think of those as your base case. So the majority of our customers don't opt-in to Snapshot, so that's a robust base case. And what do we see? This is again 2.0. When a customer got a 0% discount from us, they will actually do a surcharge. But again, we didn't do that for consumer acceptance reasons. The highest loss ratio is for customers with 0 discount. As you see, the loss ratio improves markedly as you go to the right. It's important to look at the scale on this graph. That loss ratio is much better for the best drivers who are receiving the biggest discounts. When you look at the retention line, the consumers who expected to receive a discount and did not, the 0 percentage, lowest retention. Those customers leave way more frequently and when they leave, where do they go? They go to our competitors. By and large, our competitors are not priced for this. So those customers are going for more than likely a lower rate than we're offering them, so we believe they materially miss price at the competitor. You go out on the right, not surprisingly, retention improves a lot as you give people bigger discounts, but the big 2 points I want you to take away here is the least profitable customers, and frequently, the unprofitable customers, leave most frequently. The most profitable customers are way more inclined to stay. In auto insurance, I would tell you, in any business model, if you have a program where your best, most profitable customers are most inclined to stay, and your least profitable or even unprofitable customers find their way to your competitor, that's a great program. Pat Callahan showed you the industry loss ratio graph, Progressive was here, right, the industry on a pretty decent slope up. This isn't the only reason, but this is the kind of segmentation that we believe drives that disparity in the marketplace. So let's look at 3.0. Here, you can see we are not surcharging those customers who are on the worse end. So our loss ratio for those customers goes down. Customers out on the -- more -- better driver end of the spectrum, higher discounts. Again, better loss ratio, not quite as low as with 2.0, but still exceptional. You look at that scale. And when you look at retention, because we're surcharging customers, they leave even more frequently than they did in the previous model, and not surprisingly, consumers we're offering the discount to continue to stay at a much higher rate. So Snapshot is getting us more preferred customers. It is helping rating segmentation, adverse selection a lot. Hope you got those points. Our next step for Snapshot is mobile. We've been very deliberate in coming to market with mobile. If you think about the challenges in measuring driving behavior with a phone, there are obviously many, right? Some of you take the train, some of you take the bus, you commute with a friend, you leave your phone at home, what have you, a lot of challenges and we wanted to make sure we got this right. We've done a ton of testing. And just to give you an appreciation of those challenges, what you're looking at here is an example of a real test case, so an individual with the app and the device in the car. This is about 2 months of driving, each row represents a day, 24 hours across the horizontal axis of a day, not surprising, and what do we see? So if the blocks are purple, that means the device and your phone are showing a trip, they're matching up. If the block is blue, that means the device is just showing a trip, not the phone. Red the opposite. Red is the phone showing the trip, not the device. So if you stare at this, you say, okay, it looks like we have a commute pattern, right? Five days a week, we drive to work and we drive home. And we take our phones to work and they're on. We don't see that on a commute, right? What else do you see? Well, there's some blue sections, those blue sections look to me like they're more on the weekend, maybe a spouse is driving the car, maybe a kid is driving the car. The red dots in the middle are very intriguing to me. My conclusion, I don't know this test case, is that this person has a regular scheduled Monday lunch date and the buddy drives, to lunch then back, interesting things. But the point is how do we account for this and still have really accurate driving data to rate on. We worked with a ton of app developers, the technology is amazing. The app is able to learn the driver's fingerprint, if you will. And it learns it pretty quickly, I have an app, and I can tell when my child is driving the car, when my spouse is driving the car, it excludes those trips. Thankfully, for my own personal rate. You get the point. How about this, we can look after a trip starts, back to see if the driver got in the left side of the car or the right side of the car. Technology in phones is amazing now. We worked really hard to make sure this is right. Let me give you a picture of the rating reason behind why we've worked so hard to make sure this is right. We worked with a number of different companies developing apps. Hard brakes are our most important rating data field for UBI. So you want your phone to reflect a brake about the same time as you want your device to reflect a brake, right? The graph on the right shows pretty tight relationship, graph on the left not so much. I'm really happy to tell you that we chose the app on the right. And I'm almost as happy to tell you that we're pretty confident a competitor was using the app on the left. Really important that we got it right. We think we have it right, we'll have 2 states out before the end of the year. We expect to have the rest of the country rolled out in the first half of next year. We think it's a great rating tool. We also know it's going to be a better consumer experience. Today, we mail a lot of devices to customers, not all of them come back, but the customer has to take it, unbox it, put it in their car, extract it from the vehicle, mail it back to us. It's a logistical challenge and it's costly. We've previously told you the technology expenses with our current model is about a point on the P&L. Mobile will completely eliminate technology expenses for sure, but it will materially reduce those costs. As I mentioned to you, the long-term model, we believe, is data right from the vehicle. Again, this won't be a broad-based model anytime soon because of the pace of technology penetrating the fleet. But it's a fantastic model, it's working really well, it's small so far, admittedly. General Motors customers' model year 2015 and greater have this capability, all voluntary. They can opt-in to sharing 90 days of driving behavior with Progressive. If it looks like we'll give that customer a discount, we'll offer them a quote. If it looks like they wouldn't be due a discount, we won't offer them a quote. Think back on that loss ratio curve. The customers we are selling to here have a very high average discount. What was true about the most -- the highest average discount customers in our programs, they're the most profitable, right? So we've got a model here where we can just sell to the most profitable drivers with our current algorithms. It will evolve, we'll probably tailor to the various methods of measurement, but today, it's a great model, we're really excited to be in market with General Motors and we're excited to talk to you more about that in the future. I told you that the data explosion has been rapid. I told you that it's been a challenge for us, frankly. This graph shifts to terabytes as opposed to trillions of records. Back in 2013, our analysts were struggling because it was taking them almost a month to process the Snapshot data set, almost a month. We literally had analysts trying to create mini server farms in the closets near their workspaces. We got smart, we brought in the right technology, we developed the talent internally, our processing time dropped dramatically, going up a little bit as the data set grows further, but the better news here is that we are now using this technology across other parts of the company. When we were together last year, Pat Callahan told you we were starting to look at claims fraud using big data techniques. We now have that fully rolled out. Every night, we are looking at hundreds of data elements through our claims file, through our policy files and looking for the potential for fraud. We don't identify fraud with this model, we refer it to an individual who does the further investigation and does the right thing handling the claim. We're using traditional data, but we're also using what's called unstructured data, which in this case means text. So claims adjusters are typing in notes into our systems, this mines the text and joins that with other data to look for potential fraud. What you're looking at on the screen here is a network of claims that we had previously identified as fraudulent, run it through the system overnight, the claim in blue pops out, suspect claim. Our adjuster does more investigation: 3 vehicle accident, 6 injuries claimed in this accident. Further investigation revealed the damage on the vehicles did not match the facts as reported by the injured parties, we are able to deny the claim. Previously, our model here has been if it looks suspect, the adjuster refers to an expert. I can't tell you we would have missed this one, but I can tell you, I bet we would have missed this one. And I can tell you, we're looking at a lot more claims much more with a zero in target [ph] sort of approach today than we previously did. One more example in that arena. We previously had identified a suspect claim, run through the system overnight and we find 1 data element common with 166 other policies in our database. What we had identified here upon further investigation is what we call a street broker. In urban areas, where auto insurance is very expensive, there are enterprising individuals who use our systems to take your information and create an ID card and sell you that ID card for a couple of hundred bucks cash. They generally are using a stolen credit card or another means to buy the policy or immediately cancel the policy. So what we had here was 166 people who were probably not in the right state with their auto insurance, we were able to reach out to those customers and do the right thing and get them into the right place. Another quick example, you probably know that big data has been in play in the e-marketing, online marketing space for a long time and we've actually enjoyed the benefits of vendor or third-party solutions, optimizing advertising networks online with banner ads, that kind of thing. We are now using this processing power to design the experience. In our direct business, we pay a lot to get a prospect in the door. We do a lot to make sure we streamline that funnel, if you will, get them the quote, get them into the buy process. Historically, we've done really well with that, but we've done it somewhat painstakingly, frankly, because we do A/B tests. We want -- try the orange button versus the blue button, the button over here versus the button over here. But we've done it really, really well. And historically, we did something in this case that we call the coverage packaging tool. We used 7 variables to optimize the experience with the customer and we were able to improve the funnel, the yield, by about 4%. And if 4% doesn't feel like a lot to you, we have a lot of direct prospects. That's about $0.5 billion of lifetime premium. Today, we're using what we call real-time decisioning. We're looking at more than 3x as many variables as we used to. We've been able to improve the yield another 4% and growing and we've been able to do it in about 1/5 of the time. Tons of opportunities, I could go on with examples. I hope you get the picture. The data space is booming. Even with Snapshot, due to GPS, due to mobile, this curve is going to get steeper. Layer on top of that, that unstructured data we just talked about, the text. In our call centers, we've been translating voice to text for many years now. We're able to mine that as well. Andrew Quigg will talk about that in a moment. We're now starting to ingest images, photos and video. That will soon be, not today yet for us, that will soon be data we mesh with all the other data in our systems as well. We have been a leader in using data and analysis for segmentation, for experience design, to our competitive advantage. I believe the opportunity to do that and differentiate ourselves versus the competition is only going to grow and we expect to lead. Now I'm going to hand the floor over to Andrew Quigg to talk customer relationship management. And here again, I think you will see similar themes through Andrew's talk.

Thank you, John. I have the pleasure of speaking to you today on 2 topics. The first half of the section will be on retention, and I'm going to tell you about the improving momentum we see with our retention. And then in the second half of the section, I'll provide an overview of our customer marketing initiative and I'll describe the advanced analytics we're applying to this problem, as John Sauerland alluded to. So first off, I want to just reiterate our preferred retention measure, which is Policy Life Expectancy or PLE. PLE is the average length of time we expect a new policy to stay with Progressive. We calculate PLE by first calculating our retention decay curve, and then we take the area under this curve to get our PLE statistic. That retention decay curve can be calculated multiple ways depending on the trade-off in responsiveness versus stability that we see. Our standard measure is to use 12 months for each point along that curve. That is our trailing 12 PLE that we've always reported out. We also calculate a trailing 3 PLE starting in 2014. That trailing 3 PLE uses 3, the most recent 3 months along that retention decay curve. So it's more responsive to more recent observations. Just some notes on the PLE calculation. The first 60 months represent actual data that we've recently observed and between months 60 and 300, we use an extrapolation. And we've seen all policies end after 300 months, which is a conservative assumption. We think the PLE statistic is accurate for a number of reasons and I'll highlight one of those today. We think it removes any tenure bias. So if we use simple retention rates, as our book of business ages, all else being equal, we think we'd see higher retention rates and that would be although that retention curve didn't shift. So using the PLE calculation removes any tenure bias from our statistics. So with that background, I wanted to give you an update on the current progress with trailing 12 PLE. I'm happy to report that we've seen very positive PLE growth in 2016. We are up 6% year-over-year. 6% year-over-year, there we go. This is the largest increase since 2009 and as Tricia said on our last quarterly conference call, for every one month of PLE, we see $1.5 billion in additional lifetime earned premium for the company. We're also pleased that we're seeing similar percentage increases within our agency channel and in our direct channel and direct is at an all-time high. While we're really pleased with the overall momentum we see, we're even more pleased with what we see with our Robinsons in the Destination Era. For our Robinsons, PLE is up 17% over the past 3 years. And the past 12 months, we're up 9%. So just great momentum that we're seeing with the Robinsons. As Tricia alluded to earlier, the Robinsons are now 3x greater and yearly 3.5x greater than our Sams for PLE. We'll also be adjusting the PLE calculation in the coming months. We haven't updated that extrapolation for a while, and we're seeing better retention in that tail of the PLE curve than we've seen historically. And so when we update the PLE calculation, overall PLEs will go up by more than -- by about 2%. And for our Robinsons, it will be even higher. Our Robinsons are disproportionately represented in the tail of the PLE curve. This means that we'll have even more opportunity with our Robinsons. We'll be able to acquire them more aggressively, and we'll be able to bundle our current customers more aggressively. So far, I've spoken inclusively about the trailing 12 PLE, but I want to go back to the trailing 3 that I talked about in the opening. If we think about the difference between the trailing 3 and the trailing 12 PLE is the unrealized momentum in our PLE data. And we're very pleased with what we see here. For total auto, the trading 3 PLE is up 8% year-over-year. For the trailing 12 PLE, it's up 6%. So we think we have about 2 points of unrealized momentum within our PLE. As you can see across all of our segments, the trailing 3 PLE is higher year-over-year than the trailing 12, so we feel like we have a lot of momentum. But the highest difference is for our Robinsons. For Robinsons, we think we have about 4 points of momentum that we haven't yet realized. I want to stop here and just talk about the goal that Tricia outlined for us in 2015. Tricia has challenged us to increase retention by 50%. We think we can achieve this in 3 ways. First, we believe we can change our mix. In the Destination Era, we think we can change our mix to more preferred customers and new business. Pat outlined some of the strategies we have there. In addition to changing mix, we think we can improve our price competitiveness relative to the competition. Pat showed the slide where we have increasing agency PLE while overall rates are being increased. And third, we believe we can improve the customer experience. We think we can nurture our customers to stay with Progressive longer, and I'll spend some time on that in the next section. We think this path to a 50% increase runs through our more preferred segments. We see more opportunities with our more preferred segments to increase retention. And then as the mix of our business changes, we believe we'll hit that overall goal of increasing retention by 50%. I want to switch gears here and start talking about the nurture side of things. But I want to pause and just talk about data for a minute and specifically, how we're using data for customer experience. We see a virtuous cycle in our pricing. I started off in product management, this was a big part of our roles. In pricing, we see that when we acquire customers, we are paying their data. And with that data, we can derive segmentation and insights and improve our pricing, which allows us to acquire more customers, which gives us more data, which gives us better segmentation, a virtuous cycle exists within our pricing. We think the same virtuous cycle exists with our customers in customer experience, specifically with our customers, we gain customers and then we interact with them. Those interactions generate data and that data provides insights and ideas in segmentation. That allows us to increase retention, which allows us to have more customers. And more customers provides more interactions and more data and we get a virtuous cycle here within customer experience as well. And in the Destination Era, as we gain more products, we think that cycle spins even faster, more products means more interactions, which means more data. So as we invest in the Destination Era, we think we'll have even more predictive data to use with our customers. We give you that background because leveraging data is part of a larger initiative that we're working on here at Progressive. In the 2015 investor relations meeting, Glenn said that we need to be as good at customer marketing as we are at consumer marketing. And so we've put a joint team together from our customer relationship management organization and corporate marketing, and that team is working on this customer marketing initiative. There are 4 pillars that we talk about with the customer marketing initiatives. The first is that we want to improve our customer knowledge. We want to assemble as much predictive data as we can for our customers, so that we know the experiences that they're going through. After that, we want to improve our customer engagement. There's 2 parts to this. We want to improve the processes that they go through, and we want to segment those customers better. The third pillar is interfaces. As we've come up with segmentation and processes, we need flexible interfaces to deliver those to our customers. And then finally, we want to brand those interfaces for our customers. Jeff Charney talked about this last year and so I won't spend time on it this year, but we want to have the branding that resonates with our customers as they interact with us. So with that said, let me walk through some of these pillars in particular, and I'll start with knowledge. As an initial example, I'll talk about small data that we are -- have been working on for our customers. Over the past 18 months, we've looked for internal and external data elements that are highly predictive for our customers. I have a smattering of those data elements on the screen. On the X axis here is the volume of data. You can think of this as the number of customers described by the data. And then on the Y axis, I have a general lift associated with that data. You can think of that lift as a lift in retention or a lift in graduation to become a Robinson or a lift in endorsements. No matter which of those 3 we're talking about with the data, it's highly profitable for Progressive to know. If we know that there will be a change with that customer, we can act on it in a profitable way. So let me give you some examples. Example #1 is an internal data element. It came from our digital servicing environment, it's a digital servicing action that we see our customers take. It wasn't within the structured data warehouse, but we had it in-house. And when we dug into it, we saw there's a 14-point difference in retention if we saw this data element. To give you another example that I can get a little bit more transparent about, example 2 is an external data element from a data vendor, we call it the newlywed variable. It comes from wedding invitations, bridal registries, courthouse records, things associated with getting married. And when we see that data element on one of our customers. In the next 6 months, there's more than a 200% lift in whether they change their marital status. And when a customer changes their marital status, it changes the retention rates, it changes the number of vehicles and drivers in the household and sometimes it adds additional products like a property policy. Now beyond small data, we've also started to think more about the big data available to us in customer relationship management. A leading technology company said that 80% of the world's data is unstructured, and we think that's true within CRM as well. I'm going to describe one area today, although there are others that we are working on as well. The area I'm going to describe today is the action that we're taking to put together an unstructured database of our customer words, the words that we hear from our customers. So let me give you some examples of some sources of this data. The first area is from our survey responses. Annually, we receive more than 80,000 survey responses from our customers. In addition, we have more than 1 million chat transcripts with our customers. When they go to our digital servicing environment, they can chat with a Progressive consultant. But as John Sauerland alluded to, the largest and most significant source of words is from our calls with our customers. We received 36 million Personal Lines calls per year, and we've been transcribing those calls, and we have -- using advanced analytics, and we have all those calls in text form that we can put into a data warehouse. In total, Progressive has more than 10 billion words annually from our customers. So what can do we do with this? Well, I'll give you a few small examples today. One thing we've done is just search all this text for 3-word combinations, we call these trigrams. And there are some trigrams that are very predictive of, say, defection. I have 3 of those on the screen. I can tell you, for example, that trigram #3 is, 'I never received.' I never received my ID card. I never received that document that I requested. But I never received. And when a customer says that to Progressive, they are 1.4x more likely to defect from Progressive in the next 6 months. Another thing we've looked at is metadata. You can think of metadata as something other than the words that are present on those phone calls. For example, metadata #2 here is the amount of silence time on a phone call. As that silence time increases, the future defection chance increases as well, so there's about a 1.1x relativity on the amount of silence time on a phone call. There's even some significant words, some single words that if spoken or written or shared are a good predictor of future defection from Progressive. I won't share those today, but we've parsed through the text to look for those single words. So I'd like to move beyond knowledge now and talk about our engagement. And as a reminder here, I'm going to talk about the segmentation side of the engagements. In Progressive, we've certainly built traditional models. We use generalized linear models in a number of situations. These are tried and true models that are used throughout insurance and throughout business. And here, I want to tell you about the probability of Robinson graduation. This was a traditional modeling exercise that we went through in CRM. So we gave a team, I'll call them team #1, some of that small data that I talked about in the knowledge section. And we asked them to predict which customers may become a Robinson in the future, and they did a good job with this. They built a model. And the model in the top decile had a 4x greater chance of becoming a Robinson than the least responsive decile. But then we took that same data and we gave it to a separate team, one that did not work with the first team. And the separate -- second team did an even better job. In the second team, we found that the team came up with a model that was 7x greater. But the best result was if we combine models. And when we combined models, we found that the best tranche had a 9x greater likelihood. So even in the situation where we're building traditional models, we try to put the models together and combine them in ways to create the best outcome for Progressive. Now the world has moved beyond just generalized linear models, given the vast amount of data and the computing power that's available. It's moved to things like random forest or gradient boosting machines or neural networks, some of these advanced predictive models that are now available to us. We've invested in an advanced analytics team within CRM to take advantage of these models. The first assignment for this advanced analytics team was to build a defection model, to try to predict who might leave Progressive. And the best models were the gradient boosting machines and neural networks are being built. Because of the power of these models, I want to spend a little time just walking through the first generation neural network with you today. On the screen, I have the basic structure of a neural network, and I want to explain it to you. The first layer of a neural network is known as the feature layer. Features are like the independent variables that you would've seen in our regression historically. They are the attributes to the policy. Here on the screen, we have 2 features. F1 is a categorical variable of whether or not the policy is on EFT payments or automatic payments with Progressive. F2 is the percentage change in premium that, that policy will see in its upcoming renewal. We're going to have 2 predictors in this small snippet of a model. I can tell you in our first generation neural network, we had more than 900 predictors, and that was before we had access to all those billions of words that I described previously. From the feature layer, data passes along synapses to an output layer. Synapses are weights between negative 1 and 1 that are multiplied by the data. And then the data reaches this hidden layer. At the hidden layer, there's functions that are set there, common functions that are used in neural networks that transform the data. From the hidden layer, it travels along another set of synapses to the output layer. Again, the synapses multiply it by a number between negative 1 and 1 and then in the output layer, there's another function that transforms the data. And out of that output layer comes a predictor of defection or retention with Progressive between 0% and 100%. With that set up, let me walk you through a training of this neural network. On the screen are 8 policies from a training data set. These are 8 policies where we both know the predictors, the features of those policies, and what happened with them. So we have some policies that renewed with Progressive and some policies that defected from Progressive. What we do is we run these policies through the neural network. This is called forward propagation. Each of these policies passes through the network and is calculated the retention of it. So the features enter that feature layer, they're multiplied by that number between negative 1 and 1 in the synapse layer, they hit the hidden layer where they're transformed by a function. Another weight hits it and finally, it's transformed by a final function and a predictor of retention comes out. You can see here in this first iteration that the neural network is almost random. All those policies are clumped around the middle of the scale for predicting retention and the overall model accuracy as well. But what happens next is 1 of the 2 reasons why a neural network works. What happens next is a self-correcting mechanism exists within a neural network. It's called backward propagation. During backward propagation, each of those synapse weights are recalculated. It's calculated based on an iteration method called stochastic gradient descent. Stochastic gradient descent helps the model move towards the maximum accuracy. So as those weights change, this model is changing itself to become more accurate. After this backward propagation, we train the model again. We take these 8 policies and we pass them again through the model. Each of the policies enters the feature layer, hits the hidden layer and goes through the output layer. And you can see this time, the policies that renew with Progressive are starting to go higher on the renewal probability scale and those that defect are starting to go lower and the model accuracy improves, and then we backward propagate again, again adjusting the model. In actuality, we do this tens of thousands of times. Here in the room, I'm going to do it 10x quickly, just so you can see what happens. And this is the second reason why a neural network works. It's because of the brute force computing power that's available to us in this day and age. At the end of this, the model is trained. You can see after 10 iterations that the model accuracy has improved, and we have a model that's now ready for production. I do want to note here when we do this in actual practice, we keep a hold-out data set to make sure that we don't over fit the model. So what happens now? Well, we want to use this model in production. So I brought in 2 policies that we don't know the outcome for. They may be coming up for renewal with Progressive, Policy 9 and Policy 10. What we do is we take the features of these policies, and we pass them through the neural network. And here we can see that Policy 9 will retain with Progressive likely, and Policy 10 might defect from Progressive. We started to discriminate between policies that will stay and those that might leave. Now I hope you take 2 things away from this. First takeaway is that this is an incredibly intensive calculation process. For this first generation neural network, it took us 7 full days to run. We were using our local infrastructure but still it took an incredibly long time. And that was before we had tens of billions of words. That was before we tried adding on more hidden layers to move towards cognitive computing. It's just a very large and complex process. And it's because we have an impressive analytical infrastructure that we built through the Snapshot program that we can start tackling problems like this. The second thing I hope you take away is that this is an incredibly accurate model as well. Now we showed overall model accuracy on the screen. Truth be told, that's not what we're interested in. We're actually interested in the tail of the model, if we can predict those policies that are very unlikely to renew. We want to have good tail accuracy. And from the data that we used to build this neural network, we also built a traditional model, a GLM model. And when we did that, the neural network was twice as predictive as the generalized linear model. We're using the same data, but using more predictive advanced analytics techniques, we can be much more accurate in what we're doing, which has larger benefits to Progressive. Let me walk through interfaces to tell you a little bit more about how we take these models and use them for business benefit. I want to start by talking about proactive outbound communication. This is us reaching out to our customers. On the screen, I have a couple of examples of digital ads. These are digital ads we put in front of our current customers when we want them to graduate to become Robinsons. We target these digital ads based on that traditional model that I talked about previously. When we combine those models, we're focusing the most responsive deciles, we show them these digital ads. And when we've done that, along with some other activities like this, we've increased the number of Robinsons we have by tens of thousands, adding millions of dollars in lifetime underwriting profit to Progressive. We're able to do this because of all that acumen we built in advertising to consumers. We use those same patterns in advertising to our customers. Just another example is Policy 10 from that neural network example. We know that Policy 10 may not stay with Progressive, and so we started campaigns to reach out to folks like Policy 10 and invite them to call in to talk about their policy. And when we've done things like that, and we've got a few in practice now, the run rate is that annually, we'll save tens of thousands of customers, worth tens of millions of dollars in lifetime underwriting profit to Progressive. Beyond proactive outbound communication, there's also proactive inbound communication. We have built a tool that our phone consultants can use when someone calls Progressive. We call this tool the customer relationship assistant. It's a user interface for our phone consultants. It took us about 2 months to build, and the beauty of this tool is that it houses all that data and all those models I spoke about previously. So when Policy 10 calls, the consultant on the phone knows to talk to that customer about their policy and about staying with Progressive. It's really personalized, right? We can tailor all these interactions and make sure we have the right treatments that fire when that customer calls in. It's very flexible. If we want to add new treatments, we can program them in about 2 hours. And earlier results from this are very impressive. We've seen an increase in products added, a reduction in future calls and a lift in retention. Let me recap what I talked about today. In 2015, in the investor relations meeting, Glenn said that retention was our Holy Grail. And Tricia challenged us to increase retention by 50%. I'll underscore the progress that you heard today. PLE is up 6% with momentum higher. Direct is at an all-time high, and we're making the most progress with our Robinsons as our mix in that business shifts towards the Robinsons. In customer marketing, as a direct carrier and a servicing independent agent carrier, we have access to billions and billions of pieces of customer data. We have a historical strength in mining that data, driving insights to segmentation and quickly moving to market. We believe we can do that with our current customers. Feed that virtuous cycle of data and extend policy life expectancies. And with that, I'll welcome up John Barbagallo to talk about Commercial Lines.

Thanks, Andrew. Hello. Good afternoon. I'm going to share a little bit about what we're seeing in the Commercial Auto insurance marketplace. I'm going to touch on our most recent results and share a little bit about our near-term action plans. And then I'm going to shift to talk about 2 disruptive trends that are coming to you, small business insurance and non-fleet Commercial Auto insurance. And these are trends that we have not only been monitoring, but we've actually been making investments around them over the last several years, and we think we're very favorably positioned to take advantage of them. But first, let's start talking about the marketplace. I'm going to start with a little history. This is a chart that I've shared with this audience in the past. It's now been updated through 2015. And what it shows is Progressive's relative performance relative to the industry on 2 measures: combined ratio and growth in direct written premium. And 2 things you can take from this graph. One is that over this time period, Progressive, by being intensely focused on Commercial Auto as a specific line of business, has been able to consistently outperform the industry on combined ratio, usually by a margin of 10 to 15 points. And for the last few calendar years, it's actually been by 20 points. Now some of that is from our expense ratio. We have a very cost competitive expense ratio. We believe we are the low cost provider in Commercial Auto, but a lot of that is loss ratio. And by being focused in the way we are, we believe we see things sooner and react faster than most of the rest in the industry and that contributes to this advantage. The other thing you can take from this chart is perhaps a little more subtle, is that we will and have consistently slowed our growth in the business when we were at risk of exceeding our underwriting target. And we will keep that commitment to those underwriting margin targets. Even with that approach, over time, we have been able to grow our market share consistently to the point where 2015, Progressive became the #1 Commercial Auto writer in the United States. All right, let's jump into 2016. This is a view, again, same 2 measures, combined ratio and direct written premium, Progressive versus the industry, we've added a few of our key competitors, as of the first quarter of 2016. You can see Progressive up in the upper right, cruising along with really good growth and a combined ratio in the high 80s at that point. And the industry had about 6% growth and a combined ratio of 102, which actually isn't too bad when you consider how it ended in 2015, that would appear to be improvement. But if you update through the second quarter, you see general degradation of combined ratios across the board, including our own. In fact, for the industry, the change in loss ratio, if you take the loss ratio for the 3 months comprising the second quarter, it is fully 8 points higher than the first quarter loss ratio. Now there's some seasonality at play there, but that is a huge, huge change. Now I'm sure you've noticed in our last 2 or 3 monthly earnings releases that our combined ratio has been elevated, certainly above historic norms. There's a little bit of prior year development in there. For the 3 months period, it's about 1 point, 1.25 points. But the real driver of that is actually frequency. What we have on the screen now is our bodily injury reported frequency for the last couple of years. Now the blue line is the monthly frequency. As you can see, that's pretty volatile, it jumps around a lot. We don't tend to pay too much attention to that. Even though those kind of last 2, last couple high watermarks there, they're high, but certainly not beyond anything we've seen in our history. If you look at the orange line, that is the trailing 6 BI frequency, and that's a much more stable measure, but it's still -- it is relatively reactive to changes in the environment. And there you're seeing what appears to be an inflection point starting in May, maybe as early as April. We believe that is real, at least it's real enough to me that we are reacting to that. We also believe it is largely environmental. Pat talked earlier about general a rise in frequency for auto, there's theories around that, distracted driving, low fuel prices. But in Commercial Auto specifically, particularly the verticals where we tend to compete, we are seeing positive movement in a number of the macroeconomic factors that we track that generally correlate with frequency. Things like home sales, construction spending, trade tonnage, aggregate deliveries, all moving in a positive direction, which suggests to us higher levels of economic activity, also contributing to higher frequency. So we believe this is real and we are reacting to it. Primarily, we're reacting with rate. Now I've talked to this audience before about our rate revision infrastructure at Progressive, our ability to change rates across the broad portfolio relatively quickly, and we have kicked that into gear. What this graph shows is our anticipated rate change for 2016, which will come in at about 9% year-over-year change in rates, most of that in the second half of the year. We were actually holding rates flat through the first 4 months of the year, as we were operating in the high 80s. So we've responded decisively and we've responded quickly, and those rate changes will certainly help the combined ratio. But it will take time. About 90% of our policies are written on an annual basis, so we have to get through a renewal cycle to get the full effect of it. It will certainly help with new business as soon as we can get those rates in place. But in the interim, and I said, we will slow down growth to ensure we make our margins. In the interim, we're also restricting some high-frequency segments of our business for new business. We're going to stop taking it until we can get those rates in place. Now those segments have always been high frequency, they will always be high frequency, and we're perfectly comfortable writing that business at the right rate, we're just not there yet. So that's what we see going on. Again, by being focused, we think we are seeing this before much of our competition, and we're reacting before they do. And when that's happened historically, that has typically set up very favorable market conditions for us down the road. So we'll see what happens. All right. Now I want to shift to those disruptive trends. The first of which is what I believe will be an inevitable shift to small business owners accessing direct channels for the purchasing and servicing of their small business insurance policies. Consumers across a range of industries have demonstrated not only acceptance of direct channels, but in many cases, preference for those channels. And we don't expect small business owners to be any different, especially as digital continues to close the information gap between the buyer and the seller. And we're starting to see investment. Investment by people other than Progressive. And these are companies that are well-known. They are well-funded, and they have experienced management. And they are putting the assets in place. Underwriting companies, product filings, technology platforms, all to go after the direct channel for small business owners. And they're starting to spend some money. Not a lot on the marketing front, and actually, it's happening slower than we would've anticipated. We still are the largest spender in the category, but they're starting to spend some money as well. So we're seeing investments. We're also starting to see some growth. This is a view provided by the Boston Consulting Group of the U.S. small business insurance market. Now this is all lines of business, it's all lines of business for small business insurance, and they peg that at about a $65 billion to $70 billion marketplace. And as of right about now, they're saying direct channels have 8% share. Probably more interesting on this chart is what they project to be the growth rate of the channels. So a pretty aggressive growth rate for digital channels in small business insurance. In fact if you buy into this view, it suggests that traditional channels will begin to decline in absolute dollars as early as 2019. We'll see. When it comes to Commercial Auto, where we have some pretty good data, we're seeing a somewhat similar pattern. Now again, the direct channel for Commercial Auto is still small, less than 2%, but the growth rate relative to the industry growth rate is 2x to 3x. Now full disclosure, on this particular graph, that orange line is largely driven by Progressive. We're just about half of the direct commercial auto market right now, and we actually can affect the slope of that line simply in terms of how much we choose to spend. And that's reflected in 2014, we pulled back in our spending, and you see that little dip there. But again, you have a channel where there's lots of investment lining up, and you're starting to see signs of growth in the channel. So the question then becomes who is going to be able to most cost effectively drive demand and control the top of that sales funnel. And this is where we start to feel really good. Because while the others are just getting started, we have actually been at this since 2008. Now we've been selling direct commercial for more than a decade, but it was really in 2008 that we began investing in developing a business insurance brand. In 2008, we started -- we embarked, really, on a 5-year plan, and it was a 5-year plan to prove to ourselves that we could make the economics of direct marketing commercial insurance work. And it was a very auto-centric approach, targeting specific verticals where we tend to do well. We concluded that 5-year plan successfully in 4 years. That gave us the confidence in 2012 to begin broadening our marketing. And by broadening, I mean, both broadening the media we would access and broadening our message to be much less auto-centric and really built around insuring and protecting the business. And that has now evolved to our current campaign, which we just launched, which could really celebrate the small business owner. So we've been investing here for a long time. Has that investment been worth it? We would say it definitely has been. One of the key measures of brand health is awareness, and as you can see on these 2 charts, in terms of total awareness and top-of-mind awareness, Progressive is the leading small business insurance brand. Again, this gets back to who is going to control the top of that sales funnel as this channel begins to grow. How fast might this all happen? I don't really know, but maybe we can look to private passenger auto as a guide, because this has happened. The chart on the left shows the shift in channel for private passenger auto over a 20-year period. Now keep in mind, that orange line for the direct channel developed concurrently with the development and acceptance of digital channels for consumers. So the Internet, gradually migrating over to mobile. We believe those consumer preferences are already in place. And I'll offer an opinion that small business owners, at least a percentage of them, their expectations currently exceed the industry's ability to deliver really good digital experiences. The chart on the right is the effect of that channel shift on Progressive's business. So over roughly the same time period, Progressive's direct auto business grew from being about 12 -- 11%, 12% of our total business to where it is today, half our business. Now it's not that our agency business didn't grow during this time. It did grow. You heard earlier from Pat that it is at an all-time high. It's just that the direct business grew that much faster, creating lots of value for Progressive and for our shareholders. We've modeled this out. We've done a number of different scenarios around how quickly the direct channel may grow for small business insurance, based on different spend levels by us, spend levels by other. And when we surround what we think are the most likely outcomes, it all leads to significant growth for Progressive Commercial Lines. So we continue to invest in our capabilities. We are in the midst of a major systems upgrade for commercial. One of the benefits of that will be the ability to deliver better digital experiences, and perhaps more importantly, deliver them to market faster and enhance them more quickly. And we continue to invest in and build our in-house business insurance agency. The map on the left shows our footprint for being able to offer non-affiliated partner company BOP and general liability policies to prospects and our customers. We also sell workers' comp and professional liability, but the primary coverages they're seeking are BOP and GL. And today, we have -- in virtually every state, we can offer between 5 and 8 products, covering a range of underwriting profiles. And as you can see on the graph on the right, we are growing the agency. And there again, the shape of that growth curve can largely be determined by how much we choose to spend at a given time. And we are growing the agency, and we are having success bundling these small business customers with Progressive Auto, be it Commercial Auto or Personal Auto, and that's an important point. We are creating a lot of bundles with a BOP or GL policy and a Progressive Personal Auto policy, which is just fine. First of all, it helps the agency's economics because those customers are much stickier. Second, it is entirely consistent with the strategy Tricia laid out earlier in terms of acquire, anchor, bundle, extend. This may be how we acquire customers. Certainly, could be how we bundle customers. And also, we're finding that the need for commercial auto in a small business oftentimes emerges late or later in the life cycle of that business, and business insurance is very situational, unlike personal lines. Usually the first purchase of a personal lines product, typically auto, is driven by a state law, state requirement to have insurance or a bank. In small business insurance, that is not usually the case. So to try to give some context to this slide, I'm going to tell you a story. And it's a story about a real-life, thriving business right here in Northeast Ohio. So a work colleague of mine has a friend, and she's starting a business in her home, in her kitchen actually, cooking bone broth. And she sold this bone broth out of her kitchen to friends and neighbors. Now she didn't have insurance, so we told her she needed insurance. You can argue she probably should have had insurance, but she was flying below the radar, which a lot of small businesses do, particularly in their early stages. The sales went well. She got a lot of positive feedback, and she was encouraged to take her product to one of the local farmers' markets. So she decided to give that a try. She applied. She was accepted, but the farmers' market told her, if you want to come on-premise, you have to have a vendor's liability policy and you have to have a general liability policy. So where previously, she was in that first circle of not really caring about insurance, she now found herself in that second circle of, well, at least temporarily, I have to care about insurance if I want to go to this farmers' market. Not sure if I have a real business yet, but I'll buy the insurance. Farmers' market went well. Again, a lot of positive feedback, again, getting inquiries from local restaurants who wanted her to supply them with bone broth. She even got a contract with a local grocery store chain, and the business is starting to take off. In fact, taking off to the point that she outgrew her kitchen, and she had to lease commercial space, and she had to hire 2 employees to set up more of a regular production cycle for the bone broth. And now she needs a BOP policy for her premises, and she needs workers' compensation for her employees. I would say now she's in that third circle of, I definitely really have to care about insurance. She still hasn't had a commercial auto policy, but that's about to change. She had been hiring a third party to make her deliveries. She now buys her first delivery vehicle, she needs a commercial auto policy. I understand from my colleague that she is now thinking of expanding her business regionally, maybe nationally. She realizes she has a real asset here, an asset she wants to preserve, protect, maybe pass on to the next generation. I would argue she's moving into that fourth circle of, I genuinely care about buying insurance. The point here is for small business owners, it is a journey of one, and they are all different, and they are not necessarily linear like the real-life example I just gave you. People move back and forth. Some small business owners never move out of that second circle. They're kind of the Sams of small business ownership. And to deal with that, we have developed a series of customer journey maps, by business type and by business stage. And do what these journey maps allow us to do? It maps our creative and our messaging and our media buying with those key trigger events to deliver the right message at the right time to that business. So this a trend that is clearly coming. You've been reading a lot about it. It's been a little slow to gain traction. We think it's about to take hold, and we feel we are really, really well-positioned to capitalize on that. The other trend I want to talk to you about is the application of telematics to non-fleet commercial auto insurance. I'm not sure if this is a trend yet, but I do believe it's coming. So telematics have been broadly applied in fleet situations. But even there, there has never been a really strong insurance tie-in. That's been more about asset management and logistics than insurance. But we think that's about to change, particularly in the trucking sector. Now John Sauerland has already talked to you about Progressive's leadership position in terms of usage-based insurance pricing science. He's talked to you, and Andrew's reinforced, the big data capabilities Progressive can bring to bear. Those are absolutely assets we can leverage in Commercial Lines. But we have taken a slightly different path in Commercial than they have in Personal with respect to UBI, and we have focused on the transportation industry, the over-the-road truckers. Two reasons for that. One, we feel a version of the Snapshot product will work perfectly well for a lot of our business auto contractor-type customers. And now as the agency adoption rate of Snapshot is starting to go up, it may be the time for us to introduce that to commercial. The other reason that we focused on transportation is we were anticipating a federal mandate, which is now law, that will require federal motor carriers to have an electronic logging device in their vehicle and operational by December 2017. And these electronic logging devices will be fully telematics-enabled, if you will. They'll have all the data you could possibly want. So we have, for the last few years, been working directly with the vendors, the manufacturers of these ELDs to put in place data sharing agreement and cooperative marketing agreements so we can go to market together with our products as the demand increases, leading up to the compliance date in December '17 and we've also been collecting our own UBI data for the last several years. Today, we probably have about 600 of our insured trucks on the road being monitored. We've collected over 41 million miles of data, and those data are beginning to reveal some really, really interesting things. Perhaps the most interesting is that we find with commercial operators the degree of variance in their behavior is far, far greater than Personal Lines customers in terms of how, how much and where they drive. To the extent those behaviors correlate with losses, and they do, that's very exciting, because that allows for greater rate segmentation. And John's already talked to you about the power of rate segmentation. He's told you that in Personal Lines, Snapshot is the most predictive, most powerful rating variable we have. We believe, for trucking risks, it will be even more powerful than that. So we're very excited. Let me give you just a little sampling of some of the things we're seeing. All these examples relate to the same 3 insured vehicles. One of the things you rate on in Commercial Auto is something called radius of operation, which is basically how far do you go from your principal garaging address, how far do you go. It's a rating variable we use. ISO uses it. A lot of our competitors use it. A problem with radius of operations as a rating variable, it could be hard to verify and it changes over time. Truckers go where the work -- I think that's true of all commercial vehicles, they go where the work is. But we all use it. Well now, with electronic logging devices, we actually know where they go. And in this sample, we've got 3 insured vehicles that were all rated based on the same radius of operation, 500 miles, which was pretty good for Vehicle C. We're probably overcharging Vehicle A and Vehicle B. Well, more important than how far you drive is where you drive. Now that we have geolocation data, we can actually score different routes and different roads in terms of relative safety or risk. From there, we can overlay time of day, so you incorporate things like rush hour and potentially things like seasonal weather pattern. You begin to see where this can go. But even with a relatively simplistic model for scoring, which is what we have here, you see pretty significant differences in distribution, which we believe are actionable and ratable. One more, this is how much you drive, and this is really the whole impetus behind the federal mandate. Better regulation of hours of operation, hours of service. Driver fatigue is a major, major safety issue for over-the-road truckers. The graphs on the bottom reflects hours per day of driving. So each bar on the graph is a day. And the white line going across is the federal limit of 11 hours per day. So you can see pretty significant differences in these distributions. Vehicle B -- and remember, this is the guy that goes the furthest from home by far, seems to be doing a better job of managing his hours per day. Vehicle C actually had some 20 and 22-hour days in there. Now what that probably means is we have team driving, 2 drivers taking shifts. And still, that's really important and valuable information and represents a different risk. So this is just a few samples of the types of data we are beginning to develop, that we think will be very, very actionable. I'm going to conclude with one more chart, just because I love the data, and this is another comparison between Personal Auto, and in this case, some long-haul trucking customers. So the bottom graph was taken from a sample of Snapshot customers, Personal Auto Snapshot customers. And all it does is it shows kind of where their cars are during different parts of the day. And what you see by and large is most of the time, those cars are at home, in the driveway, in the garage. And during the daytime hours, that's where you see the driving. Not so much driving in the wee morning hours. Not surprising. The graph on the top is a sample of Progressive-insured long-haul trucks. And what you see there is lots of driving, not surprising, not unusual, 100,000, 120,000 miles a year for a trucker. But when they're not driving, they're typically away from home, sleeping in their truck, showering at a truck stop, using a, I don't know, Cracker Barrel. It could be a pretty hard life for a trucker, and this is where I really start to get excited about these data, because this is where I think we can go beyond just straight rate segmentation and start to develop value-added services by using risk management services to hopefully make these truckers a little bit healthier and a little bit safer while they're on the road. So again, another trend, we think, is coming to small business insurance, in this case, non-fleet auto insurance, that we feel we're very, very well-positioned to capitalize on. With that, I'm going to bring up Tricia Griffith for some closing remarks.

Thank you. We hope it's evident how bullish we feel about the future. Clearly, last year, Pat stood up here and said his primary goal, our primary goal, is to restore business in the agency channel. We didn't just restore it, it's at the highest level ever. We continue with our segmentation of more and more product models and having deeper segmentation at the quickest amount of time ever. So we really think that's a competitive advantage. And our bundled PIFs have grown 20% to 30% across channels. John Sauerland talked about our most important variable, our UBI or usage-based insurance, so Snapshot, and how we're continuing on that evolution and how with that work we've actually had even more data capabilities across the enterprise that we believe will be a winning strategy for Progressive. Andrew Quigg talked about PLE. We finally made that breakthrough in PLE, especially in the Robinsons segment. Andrew talked about the massive amount of customer data that we have and how we're going to use that data to understand the moves of our customers and be out in front of that, give them a reason to stay. John Barbagallo talked about the fact that we're outperforming our peers relative to profit and growth, so that's extraordinary. Bottom line is though, we have to respond quickly to our frequency trend, and we're doing that with rate. John has his own era coming up in terms of the direct business with small business owners as well as technology and data-driven technology in the commercial space. I hope you're thrilled as our team is about what we've executed upon and what we're going to be doing in the future. And I want to thank you for your interest in Progressive. So with that, I will ask John Sauerland to join me for some Q&A.

Elyse Greenspan with Wells Fargo. So my first question is in terms of the 96% margin target. The way you spoke about it today, seems like a company-wide goal. From the last conference call, you kind of spoke as the homeowners focus not necessarily included in that 96% target for now. So if you could just kind of clarify that. When you look at the 96% goal versus the 96.2% year-to-date, is that company-wide goal? Or I thought for the time being, you were excluding the homeowners portion of the business.

Yes, Ian asked that question during the conference call. I corrected myself. I was thinking from a gain share perspective. It is a 96% across all products, including homeowners.

Okay. And then so, I guess, combined with that target as well is that we have this storm approaching Florida. So as we, I guess, double-part question, think about both your retention and exposure to the loss, and we do see it kind of contained to your retention, how are you thinking about your exposure, just in terms of the hours clauses in your contracts, given that the storm might end up impacting Florida twice? And then how do you just think about exposure there in terms of still hitting that 96% target for the full year?

Well, when you mean exposure, reinsurance? In terms of that, I'll let you answer that part of that. But obviously, we can't predict the amount of the storm and the loss cost that will come from that. We have a CAT load that we put in each year. This seems to be something extraordinary. So right now we're at a 96.2%. Depending on what happens, clearly, you can kind of get to the outcome of that. We feel good about where we're at with reinsurance, and I'll let John talk a little bit about that.

And my bet was that was going to be the first question, so thank you for it, because I'm sure it was on all of your minds. ASI has maintained the reinsurance philosophy they had before we acquired a majority stake in the company. We haven't disclosed the complete reinsurance program, but I'll give you some quick highlights. So $50 million of retention for a main storm. The first storm has a tower that is up to what -- the average model, I'll say, equates to about a 1 in 300 years storm, so that's pretty high up relative to what we believe other carriers have. We have a second event that also has a $50 million retention on it. But we're -- I think ASI has done a great thing in that it's only typical in the marketplace, in that they prepay reinstatement for about 70% of that tower. That tower is not quite as tall as the first tower, but it's about a 1 in 100-year event, and about 70% of that has reinstatement premiums prepaid. And Trevor Hillier and John Auer -- Trevor Hillier, CFO of ARX, and John Auer, the CEO, are both here. So if I misstate or I need corrections, please feel free to chime in. So $50 million on an ASI book, it's material for sure. $50 million on the Progressive book, you can do the math. We can absolutely sustain that and still hit our calendar year 96% targets.

That's homeowner only?

That is property only, yes. So we also have boats sit on the coast of Florida. Thank you for reminding us of that. That exposure, we're pretty confident we have a pretty good feel for it as well, and our expected, our modeled exposure for Matthew is lower than that $50 million. It's probably like half. Now will that be -- it's a model. We all get that reinsurance models are not accurate. So the ultimate losses won't be known until they're known. But we think again, with even -- and auto on top of that, we'll have flood losses for sure, but we would still expect to hit a calendar year 96%, even with the storm.

Amit Kumar from Macquarie. Two questions, both broader. First of all, just going back to the target of 50% increase in PLE. Last year, obviously, when we were here, you sort of sounded a bit disappointed in terms of trying to get to that number. Based on what we have seen today, do you think that 50% is a time-bound target or is it an aspiration target? And I guess linked to that is based on the numbers which we saw in the slide, is the improvement from here more linear or is it more exponential? That's my first question.

Okay, great question. I -- from the PLE, I believe that it is, it will continue. It's not -- we haven't time-bound it. And we look differently at our PLEs across segments. So when we look at Sam, Sam stays about the same amount of time. So we often look at our PLE in, excluding Sam, and then all else. So Dianes, Wrights and the Robinsons. I believe we have just gotten started. So when we think about PLE, we think of nature and nurture. So as we continue with our new product models for those preferred customers, and we understand what Pat talked about, having 2 competitive products in the -- both home and auto available, I can only see that continuing to improve. And then what Andrew talked about in the nurture side is we're learning so much about our customers, so I think those 2 together, I believe, the possibilities are extraordinary.

Got you. That's helpful. And the second question is, maybe switching gears, going back to Snapshot mobile. Doesn't that product skew the customer base away from the preferred customer? Wouldn't Snapshot model, I guess, attract a younger driver or is this much more, I guess, non-standard-ish? Or am I thinking of that in the wrong manner? In terms of the customer base, who will use that product?

Well, I can start, and John went over -- no, I think though the chart he showed actually talked about consumer acceptance at any age.

Yes. And so we haven't marketed the app yet for Snapshot. Smartphone adoption is pretty broad now. I don't know the exact numbers across ages, but it's a very high penetration. Might adoption be younger? Perhaps. We wouldn't expect it to skew towards Sam, so to speak, for any reason. Snapshot, generally speaking, as you saw, is more relatable and preferred by preferred customers. So we'll see how the app distribution goes, but we're not going into it expecting it to skew any differently than our current Snapshot market.

But does that mean -- following up on the loss cost trends, based on all the data you have initially, the loss cost trends are similar, is what you're saying?

So we have -- Snapshot app is in pilot stage today. We don't have it broadly, so we don't have enough data to tell you that it's exactly like Snapshot, but we are confident that our algorithm will react the same as the device or similar to the device, so we're confident that the rates that we apply will be accurate.

We'll know a lot more next year. We're rolling out in 2 states later this year and the rest in the first half, so probably later part of next year we'll have more data.

Greg Peters with Raymond James. I just wanted to follow up real quick on the Snapshot product. Some of your peers, I think, have observed that their product stays on continuously even past that -- your pricing period. And I'm curious if you have a perspective on that as it relates to what you're doing. And then I have a different question, sort of in the same area.

Yes, I think that we believe that we obtained enough data to properly rate that, that person with other variables. And we feel like we did not need to pass that time frame.

So in the data collection process, do you feel like you're missing out on data because you're not continuously tracking while they are customers, especially as you move towards the Robinsons, who are going to be with you longer, don't you think that data would be more valuable if you continue to collect it?

We felt -- we thought that we had enough data to accurately price for those exposures.

We tested multiple models years ago, actually, with Snapshot. So we actually tested a continuous model, to use your words, meaning the device remained in the car past 6 months. And we concluded we were good with 6 months.

Okay. And then I just want to get back -- I was fascinated with Andrew's presentation. Thank you for offering us the opportunity to listen to that. And I'm just curious, from your perspective, can you tell us a little bit about the amount of money that he's invested in all of these different models and initiatives in marketing and maybe how it's changed this year versus last year? Because maybe you're spending less in advertising, more on modeling, quantitative analytics, et cetera. Just give us some color behind that, that would be very helpful.

Yes, I mean, Andrew was the perfect person to put in the new role that we put him in last year, because we really wanted to understand deeply about our customer and how to get out in front of those things. And we just didn't have a lot of modeling done. We've been hiring across the company for data analysts. But I wouldn't say it's had any effect on what we pay for advertising. So that, those are 2 different things. We pay for advertising based on trying to get -- use an efficient amount of money to get customers in the door. So I would say this is -- we're spending the right amount -- I don't know the exact amount, I don't know that Andrew does either -- with just understanding data and then getting out in front of understanding how to proactively approach our customers when they think they need us and make sure they nurture right. This, to me, is worth it because how it should play out is our customers should stay longer and it should pay for itself by far.

I would agree. It will pay for itself many times over. I think the biggest investment is people. And we are moving some of our best data modelers in the company into the CRM group, which we previously hadn't done. And we think it will pay big dividends.

Kai Pan, Morgan Stanley. I have 2 questions. Number one is on the combined ratio. It's increased about 3.5 points compared with last year. You mentioned that CATs contributed to 1.5 points of that. Is the rest of the 2 points mostly because the new business have higher loss ratio? And do you expect that you were slowing down the new business to improve the 96% combined ratio currently?

Actually, the CATs for 2016 have contributed 3 points.

Yes, but relative to last year, it was 1.5. So it increased 1.5.

Yes, so the increase is 1.5. We don't intend to slow down our new growth. We want as much new growth as we can. We anticipated the new growth tax that came into play. I think as the next 4 months play out, we'll clearly take action, but rational action, to get closer to that 92. So yes, the new business has been a huge part of that process, especially on the direct side, where we front-load all those acquisition costs. So yes, we prepare to have that new business tax for that growth.

I might point out one difference, if you go on year-on-year, last year, we had about 2 points of favorable development at this point in the year. Right now, we have 0.1 point of favorable development. Again, I want to be as accurate as possible, but if you're doing the year-over-year, some of last year's calendar year result was prior year favorable development.

And just to add on to that, we started out this year with some unfavorable development from December losses from 2015. So there's a couple of different things in play there that account for it.

Okay. And the second question is sort of larger picture. If you look at the adoption technology in the car -- safer car as well as driving behavior change like using shared mobility, using Uber, how do you think this would impact your business over the long term as well as in the next 3 to 5 years?

Well, we've showed several times on how frequency over the last 50 years has continued downward. And of course, it's been offset by a lot of severity trends. We know that continuous safety models, and ultimately, at some point, we think that will be in the far future, autonomous cars will happen. So that's obviously a big threat to the insurance industry. And we're thinking about that. We have a group we call a runaway team that every day thinks about what could we do to monetize different things, what could we do to have new innovations? And so normally, we've been on the cadence of discussing the technology around vehicles about every other year. As our -- we did it last year, we'll update you this year -- the next year, I should say, on what we're doing with that. But we don't think -- we don't believe frequency trends are going to reverse, and we have to prepare for that.

Meyer Shields, KBW. John, I have to ask, just because I couldn't hear, were you saying that the model retention was 5-0 or 1-5 on the catastrophe?

What's the question?

On the catastrophe model, catastrophe loss, I just couldn't tell if it was...

5-0 retention.

Okay. Real question, I'm trying to understand why, when we had sort of an industry-wide uptick in frequency, the Personal Auto performance was actually pretty robust and held up through that. In other words, you didn't see margins deteriorate rapidly or beyond expectations. On the commercial side, it seems to have been very sudden, and I', wondering if that's an internal issue or the nature of the external variables.

Well, John, you can add more to that. We -- it did creep up pretty aggressively. We added a lot of new business in, and we saw that -- the first few months when we saw the frequency trend go up, we weren't too shocked by it, because relative to those months, it was a little bit higher, but -- to the months compared to last year, it was a little bit higher, but nothing too detrimental. And then when we saw that third month, we dug in deeply and reacted quickly.

I guess I need to get on camera.

Okay, fine.

I think and I tried to add this color, in Commercial Lines, we're also seeing increased levels of economic activity. And we know when we see that, frequency follows. So that, when you put it all together, I think there's a general rise in frequency, which we see across all auto. But when you look at the verticals where we participate, a lot of construction, a lot of aggregate haulers, over-the-road trucking, so freight transport, definitely positive movement in all those indicators. So higher levels of economic activity is driving that as well, we believe.

You want to stay up here? Because Jay Cohen from Bank of America has a question. Like homeowners, will you feel the need to manufacture your own commercial policies ex auto, such as BOP and GL?

Well, my comments today were mostly focused on the direct channel. And there, we feel very comfortable that we can develop a very robust platform for direct small business customers using non-affiliated partner products. We've already done that, and we think we will continue to enhance that capability. I think a solution for the agency channel is more challenging. We found that with homeowners. And the bundling dynamic for commercial is also different. I tried to share that today, so when does Commercial Auto even get introduced into the equation. It tends to be these other coverages that drive the relationship. And I would just say we are exploring a number of different alternatives for the agency channel, but we have not necessarily settled on a particular strategy.

Thanks, John.

Yes. Brian Meredith, UBS. A couple questions here. I'm just curious, how are the agents reacting to 3.0? Given the surcharge, I would think that might be something that they may not like, having to go back to a customer and say, "Listen, you're getting a higher rate," and that could potentially hurt your relationship.

Actually, the agents have reacted favorably to 3.0 because of the discount. So it's easy for them to talk about what is involved in the UBI approach. So it's actually improved -- the take rate has improved in the agency channel.

Okay, great. Second question, I'm wondering with the -- on the Commercial Auto side, has any change in mix maybe caused any more seasonality to loss ratios and those types of things in the business?

You're digging into it more than them. I think you've seen some seasonality. And I think because we've had a lot of new business growth as well, we're learning some things.

Yes, we've always had seasonality in the business. I mean, a lot of it is tied to construction-related trades. We have seen a shift in our mix of business a little bit toward more of the transportation, which arguably is less seasonal. But in terms of a direct response to your question, I don't know that seasonality has become any more or less of a factor in what we're seeing.

And I think it's fair to say that we've seen that BI frequency uptick across segments in the Commercial Lines space.

Right. And then I guess my last question, and kind of to what Kai was asking, with respect to this new business tax that you all talk about, is there anything that you can help us out with in trying to figure out exactly what that new business tax is right now on your underwriting results? How should we think about it? And as policies in-force growth starts to moderate, is the new business tax higher on direct versus agency? Can you give us some color? Because that's something a lot of us are asking ourselves right now.

I tried to do that with showing both the agency and direct. So clearly, it's higher on the direct model because of the acquisition cost. So those acquisition costs are all front-loaded versus a relatively fixed commission. And I tried to talk about that in terms of the ULUP as well as how we think that's going to earn in. I mean, I can't predict exactly how it will earn in, because there are a lot of inputs. But we feel positive about it because of the influx of the type of new business we're bringing in, the average written premium and how we believe they'll run off in terms of ULUP. I'm not sure if I'm answering your question. I mean, we, I...

What about the loss ratios? You generally have the new business tax on loss ratios.

Higher [ph].

Yes. Exactly. How to think about that?

Well, I think it's a shorter-term thing. And as those customers stay longer, especially the Robinsons and the Wrights, with that PLE, you're well below that 96. So that's why we feel good. So you've got that new loss -- and I figure, are you saying is it coming in, the new -- the next new tranche of new business.

Yes.

Yes, I mean, I think....

Each new tranche that comes in, it's got a higher loss ratio. As it runs down, obviously, the loss ratio improves. But how do we think about the math and how that kind of comes down?

I think we think about the math in terms of targets. And so we know in each segment, new and renewal, agency, direct, Robinsons, Sam, at the very lowest level, we have specific targets that we believe will run off profitably. And we know when we're writing at or below targets, we feel that business is good. And so that's how we think of it.

I might add, just for -- and we won't give you the complete math. But know that we are hitting -- the calendar year 96 for the aggregate of the business remains. So regardless of which channel is growing, which segment is growing or not, we are committed to hitting that calendar year 96 regardless of there's a storm. We also work to price segments to lifetime cohort targets, as we call it. So we would take a direct auto segment and say over the life of that policy, we want to hit that 96. And that's why Tricia was showing the economics very differently in a calendar year period than a lifetime period. So we're trying to manage through the lifetime targets as well as the calendar target. And we can switch a lot of levers to ensure that we hit the calendar year target, which may or may not have implications on the longer-term lifetime targets. But more recently, if anything, it has had the effect of having the lifetime profitability, the ULUP, if you will, of that business, even better.

James Naklicki with Citi. Two questions. So first, you talked about 1 point on the P&L from technology spending from 2.0, that I guess it goes down under 3.0 because of the mobile device. So can you talk about would that flow through to bottom line or will it be factored into pricing? And my second question was, it appeared to me that loss trends were elevated in the second quarter relative to what they usually are historically. So -- and I think that was Personal Auto. What was driving that?

Do you want to take the first one, and I'll take the second one?

Sure. So as we were just saying, whenever we find opportunity to price stuff differently or if we improve our cost structure, we plow that back into the pricing. So we will continue to price the Snapshot business to those similar lifetime targets of 96. So if we can take 1 point out of the expense ratio, we'll plow it back into the rates, be it through the non-acquisition expense ratio efforts that Tricia mentioned earlier, technology expense, what have you, so that goes back into the rates.

Yes. Loss trend, we talked about that for new business, as well as frequency was relatively flat, severity was up about 4 points.

And second quarter was a little higher than first quarter, but I wouldn't interpret that as a trend. And 0% frequency and 4% severity is very consistent with historical trend that we've seen.

On Commercial Auto, how does the new business penalty work through and stopping new business applications? I mean, how many points does that bring the combined ratio down, put another way?

I'm not sure how much. I mean, it has an impact on it. Again, most of our commercial business is written through the agency channel. So you're not going to see that same new business tax because we're not using the direct, so it's commission-based. I think it's also slightly higher on both underwriting and loss cost, so...

And the new business restrictions will affect about less than 10% of our normal new business mix. So that will significantly affect the short-term combined ratio.

With the underwriting that you're putting in place.

Yes.

Yes.

So slowing down growth does not impact the run rate in the combined ratio by itself? Is that your answer?

No, it does. Not in the short term. So those actions are really designed to make sure 2017 is where we want it to be. And those restrictions will stay in place until the new rates are in place.

I guess, I didn't follow. What is the 10% that you're -- your new business that you're not writing in Commercial Auto?

Sure. It's some high-frequency segments. Logging operations, new venture trucking operations, things that always have a high frequency. And until we get our rates where we want them, we're just going to not take that business for a while.

So it's the most volatile segment in the business, in theory?

The highest frequency, and yes, typically high limit segments.

So I'll take it to a little -- maybe a little higher level. We can get more profitable by raising rates or we can underwrite out segments at new business or at renewal that we think are the cause of the problem until we have rates in that segment high enough where we are then comfortable letting new customers back in. So in Commercial Lines, they've identified segments of new customers that they believe will be the least profitable. And they're working upfront to underwrite that. In Commercial Lines, they have a lot of ability as well to underwrite a renewal with credits, debits. And they've done a lot of that over time and have targeted that, of course, as well. In Personal Lines, we're actually employing a lot of underwriting these days as well. It hasn't been what we've been -- it hasn't been our forte historically. But probably 4 or 5 years ago, we started getting a little more conservative around ensuring that we were bringing in business that was intent it was to insure and not otherwise. And I think that's helped us a lot. But higher level, you can improve profitability through raising rates or underwriting segments. And Commercial Lines has been very fast to react in turning on the underwriting where necessary, and they're now reacting very quickly in getting rates up and expect to have those rates up about 9% by the end of the year.

Great. If that's all there is, there's none from here. Again, we want to thank you for your interest in Progressive. We hope that this is informative. Normally -- we'll have the Q out probably in the next couple of weeks. So normally, we would have a conference call in November. We're not going to because of the recency of this meeting. So we will talk to you soon after that. Thank you.