Lemonade, Inc. (LMND) Q2 2023 Earnings Call
Lemonade, Inc. (NYSE:LMND) Q2 2023 Earnings Conference Call August 3, 2023 8:00 AM ET
Company Participants
Yael Wissner-Levy - Vice President of Communications
Daniel Schreiber - Co-Founder & Co-Chief Executive Officer
Shai Wininger - Co-Founder & Co-Chief Executive Officer
Tim Bixby - Chief Financial Officer
Conference Call Participants
Yaron Kinar - Jefferies
Josh Shanker - Bank of America
Jason Helfstein - Oppenheimer
Mike Zaremski - BMO
Matt Smith - Halter Ferguson Financial
Tommy McJoynt - Stifel
Operator
Hello, and welcome to the Lemonade Q2 2023 Earnings Call. My name is Alex. I'll be coordinating the call today. [Operator Instructions] I'll now hand it over to your host, Yael Wissner-Levy from Lemonade. Please go ahead.
Yael Wissner-Levy
Good morning, and welcome to Lemonade's second quarter 2023 earnings call. My name is Yael Wissner-Levy, and I am the VP of Communications at Lemonade. Joining me today to discuss our results are Daniel Schreiber, Co-CEO and Co-Founder; Shai Wininger, Co-CEO and Co-Founder; and Tim Bixby, our Chief Financial Officer. A letter to shareholders covering the company's second quarter 2023 financial results is available on our Investor Relations website, investor.lemonade.com.
Before we begin, I would like to remind you that management's remarks on this call may contain forward-looking statements within the meaning of the Private Securities Litigation Reform Act of 1995. Actual results may differ materially from those indicated by these forward-looking statements as a result of various important factors, including those discussed in the Risk Factors section of our Form 10-K filed with the SEC on March 3, 2023, our Form 10-Q filed with the SEC on May 5, 2023, and our other filings with the SEC. Any forward-looking statements made on this call represent our views only as of today, and we undertake no obligation to update them.
We will be referring to certain non-GAAP financial measures on today's call, such as adjusted EBITDA and adjusted gross profit, which we believe may be important to investors to assess their operating performance. Reconciliations of these non-GAAP financial measures to the most directly comparable GAAP financial measures are included in our letter to shareholders. Our letter to shareholders also includes information about our key performance indicators, including customers in-force premium, premium per customer, annual dollar retention, gross earned premium, gross loss ratio, gross loss ratio ex cat and net loss ratio and a definition of each metric, why each is useful to investors and how we use each to monitor and manage our business.
With that, I'll turn the call over to Daniel for some opening remarks. Daniel?
Daniel Schreiber
Good morning, and thank you for joining us to discuss Lemonade's Q2 results and our updated outlook for the year. The second quarter bested our expectations on both top and bottom lines despite the outsized weather events, which dampened results across the entire industry. Tim will provide all the details shortly, though.
The headline is that our in-force premium grew 50% year-on-year, while our operating expense grew only 9% and our net loss decreased. Premium is growing more than five times faster than expenses, highlights the scalability of our business. As we continue to grow, we expect this dynamic to drive our progress towards profitability.
The importance of achieving scale was the driving force behind a major piece of news in Q2, the launch of our synthetic agents program with a long-time investor General Catalyst. We believe this program is something of a game changer and have written about it at length on our blog and we cover its mechanics in the back of the shareholder letter published yesterday. I do encourage you to study this program as it is not quite like anything we've seen before, and we believe its impact on our business will be material in 2024 and beyond.
Let me explain it briefly. To date, our direct-to-consumer business model has served us extremely well and with no plans to change it, but it does have one downside, customer acquisition costs, known by the acronym CAC are borne upfront, and it takes us about 24 months to recoup that initial outlay. To be clear, our expenditure on CAC is money well spent because over their lifetime with us, our customers typically repay their CAC three times over even accounting for the time value of money.
But because it takes time to recoup the initial outlay, rapid growth is typically cash flow negative. If we spend $100 million on CAC in year one, for example, and $200 million in year two and $300 million in year three, we could expect that $600 million of CAC investment to yield about $2 billion in gross profit over time, which is a very compelling ROI.
But before we saw that return, our bank account would see a dramatic dip in its balance, not a sustainable approach at higher growth rates. Without this synthetic agents program, long-term profitability comes at the expense of near-term cash reserves. This trade-off limits our pace of growth, particularly while the cost of capital is elevated, slowing growth preserves cash, which is good, but it also caps the amount of gross profit we can generate slowing our path to profitability and lowering our terminal value.
Now insurance companies that sell through independent agents don't have this issue. The agent finances the CAC and the insurance company can grow without depleting cash reserves. But while we do partner with agents to some extent, we prefer not to let this become our primary distribution model. For one, the agent mediated business replaces the magical Lemonade experience with the agent's own interface, commoditizing our brand and watering down the data we collect.
For another, the agent stipend [ph] offers matches half of the gross profit of the customer over the life of the customer, greatly reducing the lifetime value or LTV. So while agents do solve the cash flow gap, their costs in terms of gross profit, brand, data and customer relationship are significant, which is where synthetic agents come in.
Synthetic agents were designed to provide the cash flow benefits of independent agents without what we perceive to be their biggest downside. How? Synthetic agents finance our CAC or up to 80% of it to be precise, and they get the equivalent of a 16% commission from those cohorts they helped finance. They have no other recourse whatsoever, just a right to a portion of the premiums that wouldn't have existed if it wasn't for their funding. That is broadly similar to independent agents.
But unlike independent agents, payments to synthetic agents aren't for the lifetime of the customer, far from it. They stop after 2 to 3 years and Lemonade owns 100% of the LTV thereafter. That's a huge difference.
Secondly, synthetic agents are just financial partners, and therefore, they don't intermediate the relationship between us and our customers. Our model remains direct-to-consumer, and we own the customer relationship, the customer experience and the customer data, another huge difference. The upshot is that our synthetic agents program enables rapid growth without foregoing the customer relationship without forgoing much LTV without depleting our cash reserves and without selling equity to finance our growth. Synthetic agents that paved the way to a larger business and more profit sooner and with more cash in the bank. And hopefully, you see why we believe this program is something of a game changer.
Staying with the theme of making the most of our capital in addition to announcing our synthetic agents program, this quarter also saw the renewal of our reinsurance program, notwithstanding one of the hardest reinsurance markets in these many, many years. The [indiscernible] program is for the same 55% seed as we had previously to the same top-tier reinsurers yielding similar capital efficiency.
There are some changes to the renewed program, particularly around the treatment of cat events, named hurricanes are excluded, for example, and there's a $5 million per event cat. Yet these are risks we can comfortably bear in our newly formed captive structures while maintaining our target capital efficiency.
Taken together, the impact of our reinsurance program and synthetic agents program is significant. In that elemental state, the capital burdens of insurance, both regulatory capital and working capital would weigh up down, slowing growth, idling cash and delaying profitability. That's why our Q2 agreement are so material. Our reinsurance partners relieve our regulatory capital burden through our quota share program and General Catalyst relieves our working capital burden through our synthetic agents program.
At least from a capital perspective, therefore, the agreements that came into effect July 1 means that we're all set to grow and to go the distance, which brings us to the next hurdle we need to clear before picking up our growth rates, most notably rate approvals and loss ratio more broadly.
In my remarks last quarter, I said and I quote, we expect our current trajectory to broadly continue, albeit with occasional hiccups when outsized cats introduce a brief reversal. I stand by those comments. Q2 indeed saw a reversal due to outsized cat events, but the underlying trend line continues to be in line with our expectations.
Our rate filings have gained steam and approvals are also coming in faster now. Significantly, California approved a 30% rate increase for our homeowners product and 23% of our pet business. It will take some time for these rates to earn in, and we still need to take more rates, but we have reason to believe things are moving as they ought to.
Importantly, in parallel to our rate approvals picking up, inflation has been slowing down. This is really significant to us. And so long as these trends continue, as I said last quarter, we'll continue to expect the downward trajectory of our gross loss ratio to broadly continue, albeit with occasional hiccups when outsized cats introduce a brief reversal.
And with that, let me hand over the call to Tim, who can provide more details on our Q2 results and a view into the second half of 2023. Tim?
Tim Bixby
Great. Thanks, Daniel. I'll review highlights of our Q2 results and provide our expectations for the third quarter and the full year, and then we'll take some questions. It was a strong quarter across the board, with continued loss ratio progress despite cat headwinds, really nice marketing efficiencies and impressive expense control.
In-force premium or IFP, grew 50% in Q2 as compared to the prior year to $687 million. Absent the impact of the Metromile acquisition, organic annual growth was approximately 28%. Our customer count increased by 21% to $1.9 million as compared to the prior year. Premium per customer increased 24% versus the prior year to $360. This increase was driven primarily by both volume growth and mix shift, including the impact of the addition of Metromile's pay-per-mile customers and, to a lesser extent, increased price and coverage.
Annual dollar retention or ADR was flat as compared to the prior quarter and improved by 4 percentage points to 87% versus the prior year. We measure ADR on an annual cohort basis as a reminder, and this includes the impact of changes in policy value, additional policy purchases and churn.
It's worth noting that our ADR may decline somewhat in coming quarters as we recently passed the 1-year anniversary of the Metromile acquisition on July 28. This will add in a base of customers with slightly higher churn rates than the rest of our book. I expect this headwind to ADR driven by the change in product mix will dissipate over subsequent quarters.
Our gross earned premium in Q2 increased 53% as compared to the prior year to $164 million, roughly in line with the increase in in-force premium. Revenue in Q2 increased 109% from the prior year to $105 million. The growth in revenue was driven by the increase in gross earned premium as well as a reduction in the proportion of premium ceded to reinsurers to roughly 53% in the quarter as compared to approximately 71% in the prior year. Absent the change in the proportion ceded, revenue growth would be roughly in line with the growth in our gross earned premium.
Our gross loss ratio was 94% for Q2 as compared to 86% in Q2 2022 at 87% in Q1 2023. The impact of cats in Q2 was roughly 21 percentage points within the gross loss ratio. Absent the impact of all cats in Q1 and Q2, the underlying non-cat loss ratio showed solid improvement of roughly 8 percentage points from the prior year and roughly flat versus the prior quarter. Certainly, backing out cats is not something we can actually do as an insurance business, but we do think providing additional transparency and detail around our results can be analytically helpful.
Operating expenses, excluding loss and loss adjustment expense, increased 9% to $95 million in Q2 as compared to the prior year. This is primarily driven by increased personnel expense, stock-based compensation expense and legal and professional fees, in large part due to the Metromile acquisition, partially offset by lower sales and marketing expense.
Other insurance expense grew 55% in Q2 versus the prior year, roughly in line with the growth of earned premium. While total sales and marketing expense declined by $12 million or about 33% primarily due to lower growth acquisition spending to acquire new customers. Notably, our growth spend efficiency improved significantly in Q2, in large part due to this lower spend level. Each dollar spent on growth generated roughly 75% more IFP this quarter versus the prior year. We spent approximately $12.5 million for growth advertising in the quarter or roughly 50% of our total sales and marketing spend.
Technology development expense increased 35%, primarily due to the Metromile acquisition, while G&A expense increased 37% as compared to the prior year but notably decreased 6% as compared to the prior quarter.
Personnel expense and headcount continued to be quite stable despite continued growth in customers and premium. In fact, our headcount actually decreased 2% as compared to year-end 2022 to 1,333 and has been essentially flat for three quarters running. Headcount increased 17% as compared to the prior year, primarily due to the acquisition in Q3 last year.
Net loss was $67 million in Q2 or a loss of $0.97 per share as compared to the $68 million loss we reported in the second quarter of 2022 or a loss of $1.10 [ph] per share. While our adjusted EBITDA loss was $53 million in Q2 as compared to the $50 million adjusted EBITDA loss in the second quarter of 2022.
Our total cash, cash equivalents and investments ended the quarter at approximately $942 million, reflecting primarily a use of cash for operations of $97 million since year-end 2022. With these goals and metrics in mind, I'll outline our specific financial expectations for the third quarter and for the full year 2023.
So for the third quarter, we expect in-force premium at September 30 of between $703 million and $706 million, gross earned premium between $166 million and $168 million, revenue between $102 million and $104 million and an adjusted EBITDA loss of between $51 million and $49 million. We also expect stock-based compensation expense of approximately $16 million, capital expenditures of approximately $3 million and a weighted average share count of approximately 70 million shares.
And for the full year 2023, we expect in-force premium at December 31 of between $710 million and $715 million, gross earned premium between $654 million and $658 million, revenue between $402 million and $408 million and adjusted EBITDA loss between $199 million and $196 million and stock-based compensation expense of approximately $62 million. We also expect capital expenditures of approximately $12 million and a weighted average share count of approximately 70 million shares.
And with that, I would like to hand things over to Shai.
Shai Wininger
Thanks, Tim. We'll now turn to our shareholders' questions submitted through the safe platform. In the first question, paper bag asked about our plan to reach profitability between the years 2025 and 2027 as we laid out in last year's Investor Day. He asked for a time line update for when we think profitability would most likely occur?
Well, paper bag, based on what we know today, we see little bit changes in our multiyear breakeven timing. When we shared long-term financial scenarios in November 22, it was before our synthetic agents funding and before a notable improvement in our EBITDA. So we plan to work that into our long-term planning and give a more updated view shortly.
In the second question, Patrick K. wanted to know about our Giveback program, citing that the Lemonade Twitter feed has demonstrated the left leaning bias for the company and noting that I twitted back that this is unintentional. He asks how the company will show political neutrality going forward?
Well, Patrick, this is a topic that has always been top of mind for us since we started the company. Lemonade Zeno [ph] was founded as a public benefit corporation and integrated social impact into the core of its DNA. That means that we may be vocal about topics like gun control and climate change, which can be considered political, but we stay above the fray when it comes to party politics.
Beyond doing the right thing, we believe that taking a stand is important for our business and brand even when it comes at the cost of not being everyone's cup of tea. As I once role [ph] as part of our branding strategy, we'd rather be loved by some than ignored by all. We believe that being bold and having an opinion helps our brand rather than hurts it. But please take the time to read our post on this issue. And I think you'll find that while they are values driven, they're also sensible, moderate and without any intentional party or political affiliation. This is where I believe our Giveback and the Lemonade foundation coming to perfect alignment with our team and with our investors. As for our Giveback causes, I think you will find that we offer quite a wide range of options for which most of our customers found a cause that aligns with our values.
Patrick, to sum up this answer, we welcome yours and other thoughts on charities you believe can be part of our Giveback and have no intention of making this a politically charged program. In the next question, Darrin asked how many generative AI prototypes have made it into production? And what is the estimated impact on '23 and beyond?
So just to give some context, last quarter, we spoke about the potential positive impact of generative AI in our business and how this is the perfect moment in time for it to be added into our already highly advanced AI and machine learning platforms. Using generative AI, we plan to take our automation even further and alongside other major tech developments now going into our platform, I trust we will start to see efficiency gains in a year or so. I believe that many insurance companies will find it extremely challenging and quite risky to implement generative AI in their systems.
For one, generative AI models are highly unpredictable by design, and it's often impossible to create a consistent and perfectly compliant result. Secondly, insurance companies who rely mostly on agents, we'll see much less efficiency gains using this tech because more often than not, they don't communicate directly with their customers. Since we handle and control 100% of our customer communications, being able to automate a large portion of those can translate to reductions in our expense ratio, which leaves us with a compliance challenge.
As we started experimenting with generative AI, it was clear that we needed a way to attain these models. Regulators require predictability and auditability, things that aren't a strong size of large language models. This has been an area of focus of ours and one which I'm proud to say we've now solved. With our new generative AI compliance platform, we're able to combine the generative capabilities of large language models with our predictable, consistent and compliant Chat platform. We're now able to deploy fully compliant generative AI capabilities at scale and in a very short period of time. We're still running in lab mode, but I'm happy to report that our new generative CX technology already handles hundreds of customer e-mails and is capable of performing complex tasks with zero intervention from our team.
Our new system cancels policies, handles non-renewals, adds interested parties in secondary insureds and more with new functionality coming online almost every day. In parallel, we're also in advanced stages of development of features that utilize the vision capabilities of generative AI, allowing you to review documents such as VAT records, look at damage photos and more. Mix with our existing models, we're seeing extremely positive results.
As I mentioned before, I believe that generative AI, when combined with Lemonades-tech will help reduce our operating costs from building and maintaining software to how we service our customers, and I promised to share more about it soon.
And with that, let me hand the call over to the operator so we can take some of the questions from our friends on the street.
Question-and-Answer Session
Operator
Thank you. [Operator Instructions] Our first question for today comes from Yaron Kinar from Jefferies. Your line is now open. Please go ahead.
Yaron Kinar
Thank you. Good morning. My first question or a couple of questions are on the reinsurance program. Is there a loss corridor in the new reinsurance structure? And also, maybe you can touch on the fact that you're now retaining the hurricane risk through the affiliated entity at Bermuda. How should we think about, let's say, losses - cat [ph] losses this quarter - net losses this quarter and that track cat losses in 3Q when you had in, if they were applied with the new reinsurance program or gone up, they have been the same.
Tim Bixby
Sure. So just – thank, Yaron. This is Tim. A couple of comments on the new reinsurance structure. So in terms of a loss quarter, no, there's not a traditional - traditionally defined loss quarter. There is a sliding scale commission. So it's somewhat more nuanced in our prior structure, which was a fixed static commission rate with some potential upside. So it's somewhat different. But overall, I would sort of point out the quota share ceding proportion is the same. The players are the same, so substantially unchanged.
With the acceptance that you noted, we are retaining more of the cat risk. So hurricane, for example, main hurricanes is fully excluded. If you roll back historically, our losses have been not zero but quite low for named hurricanes and that's more a result of how we underwrite and where we're present. So no real homeowners presence in Florida and then fairly conservative underwriting in other areas where we are active with homeowners.
In terms of other storms, obviously, in the last 2 years, we've had fairly significant and quite unique storms that were not main hurricanes and so our existing - or our previous reinsurance did exactly as designed, which was protect us against the most unpredictable events and those kind of perform as they should have mitigated a significant amount of those losses. That's why we're able to hit - achieve an EBITDA result, for example, in this quarter despite a fairly elevated gross loss ratio.
We're not designing for the things that you know have happened in the past, really thinking more broadly. And so we are taking on more risk. We'll use our new captive structures that we've put in place. Those are recent ads. That enables us to continue to be sort of capital light in our approach, but we will take on a bit more volatility risk than we would have had previously. But given the hurricane history and that exclusion, we're quite comfortable with that.
Yaron Kinar
Got it. Thank you. And then my second question, on the synthetic agent, given that it now lowers your upfront cash burn or would potentially do so. Do you see yourself updating your growth targets near term and longer term? And maybe taking them up?
Tim Bixby
Hey, Yaron. Yes, I think we will. We now have the flexibility certainly from a financing point of view, from a capital structure point of view to do a lot more degrees of freedom have significantly expanded. When we gave our last kind of in-depth analysis back in November during our Investor Day, we spoke about a 20% to 25% growth rate on a multiyear kind of CAGR basis as being optimal. You grow much slower and we don't get to scale, grow much faster and the capital that's required along the way would be too excessive.
So there was a path to profitability with the money in the bag. That path remains available to us. But now actually, we've got quite a wide corridor on one end of that, to the right of that, we can now expand significantly without meaningfully impacting our cash reserves.
So at least from a financial point of view, from a capital requirements point of view, there are new degrees of freedom. As we mentioned earlier, we will be constrained by other things. We still only want to grow profitably. So getting rates approved and being able to grow in places where we see the kind of LTV to cat that we want is a precondition to accelerating our growth rates.
But as those rates come online, as our products continue their downward march in terms of the underlying loss ratio, we do hope to be able to pick up our growth rate.
The one astrix [ph] I will underline is our guidance for this year remains as we - well, we upgraded a little bit, but it remains largely as we've spoken about in the past because we don't anticipate all those conditions coming through in the next two quarters. So we do think that the next two quarters will still be quarters where we slowed down our growth and focus on implementing those rates, earning into them. And we are hopeful that at some point in 2024, if things go to plan than earlier in the year and if they take a little bit longer, they'll take a little bit longer, that during 2024, we will be able to reaccelerate growth.
Yaron Kinar
Thank you, understood.
Tim Bixby
Yaron, if I might. Sorry, Yaron, if I might. I was just checking my notes here. I skipped over one of your questions, which I think is worth clarifying, which was around the reinsurance. If you look at Q2, which was notable for its combination of cats, a very large quantity of relatively small events that aggregated to a significant number for that kind of event, we would expect to continue to be covered under the new structure. So not exactly the same, but substantially unchanged given what we saw in Q2.
Yaron Kinar
Got it. But the commit [ph] sliding commission structure wouldn't have impacted the net results?
Tim Bixby
It would have in isolation, but again, on a sort of a macro view over the course of the year would not have a significant impact.
Yaron Kinar
Got it. And then a quick numbers question on PYD or prior period development. Did you have any in the quarter?
Tim Bixby
Yes, but fairly modest. The vast majority of the impact was in period.
Yaron Kinar
Okay. Do you have the number by any chance?
Tim Bixby
Let me double check that, and I'll add that in a moment. We'll go on to the next question. Thanks.
Operator
Thank you. Our next question comes from Josh Shanker of Bank of America. Your line is now open. Please go ahead.
Josh Shanker
Yeah. Good morning. Following up a little bit about the conversation with Yaron on growth. You've spoken in the past by conserving cash until the capital markets are more willing to embrace limited ambitious plans. But you materially exceeded the growth guidance in 2Q '23, and this is before the capital light [ph] agents program was put in place. Did Lemonade grow more quickly than desired? And how much control do you have for raining in the growth in the back half of the year before the rates that you're really desiring pushes through?
Tim Bixby
Yes, this has been sort of a continuing theme for a couple of quarters now when we're by choice, choosing lower growth rates, lower spend rates to conserve capital. And what you see when you're in large and established growth channels is that when you dial back spend by definition, reducing your less efficient or less productive spend than what you're left with is the more efficient. And so sometimes it's difficult to predict how much more efficient you will become.
And so some of the upside you've seen versus our own guidance, particularly in Q1 and Q2 has really been as a result of that. So Q1 - or sorry, Q2, for example, if we just compare Q2 to the prior year, you saw something like a 75% increase in the efficiency dollar for dollar versus a year ago. Now we can't take credit for all of that. We're spending fewer dollars if we were to spend the exact same number of dollars as a year ago, I would expect that efficiency difference would not be so significant. It would likely be favorable, we consistently get better over time in increments, but it would probably not be so dramatic as you saw.
So going forward into the rest of the year, we have a guidance that lays out a sort of a mid-teens growth rate for the year in terms of ISP because we're updating guidance quarter by quarter, we do try and capture some of that overperformance or underperformance, which we haven't had, but the variance in what our guidance is. So I wouldn't expect - we don't expect to sort of see that dramatic an overperformance versus our guidance, but you could see some portion of that repeat.
Coming back to Yaron's question real quickly on the prior period development, just over 1% or so, 1.3 or so percent was prior period development and the remainder was otherwise.
Josh Shanker
I - 1.3% [ph] favorable or unfavourable?
Tim Bixby
Unfavorable.
Josh Shanker
Okay. And then so coming back to the growth question. I mean if I subtract the 4Q guidance from the 3Q guidance, the guidance basically implies almost no growth in 4Q. You have the 30% rate coming through in homeowners and 23% in California, if I layer that - I mean it does suggest that there's almost no policy count growth. Your anticipation is that you can shut it down, I guess. Is that - am I reading the numbers correctly when I think that way?
Daniel Schreiber
Josh, Daniel here. Yes, yes, we'll - what growth we're going to do in the next 6 months will be largely skewed on Q3. It is the moving season, it's when every dollar goes further. So we are taking our dollars spent and we're going to skew them towards Q3 in general. We've spoken about this in prior years. Q4 gets busy for a couple of reasons. One of them is, as I said, the moving season tapers off, but also just the shopping season, the holiday season means that ad words become more costly. So we'll definitely skew this towards a Q3 spend. Do we expect policy count growth in Q4, we do, but I think the broad strength of your analysis holds.
Josh Shanker
Thank you very much.
Operator
Thank you. Our next question comes from Bob Wang [ph] of Morgan Stanley. Your line is now open. Please go ahead.
Q – Unidentified Analyst
Hi, good morning. One quick question regarding just your commentary around AI, right? As maybe not generative AI, but AI broadly, I think in the past, you talked about machine learning, which is a much lower form of AI, so to speak, if at all, would get you to about a sub-75% loss ratio.
Just as we see the continued development of AI and the continued more efficient data analytics, especially on the cloud, which is much more scalable. Can you maybe help us think about what would be the path to achieve that sub 75% loss ratio for you just given the technological implementation going forward? And how do you plan the next 2 years in the 5 years in terms of data infrastructure as well as your AI impetration?
Daniel Schreiber
Hey, Bob. So I think our AIs are pretty much where we need them to be. Our analysis was shared what LTV-6 did back in November where 6 graduated to LTV-8, it's worth just delineating Shai's comments were about generative AI, which is - we spoke about it in our Investor Day, but it certainly exploded over the course of the last few months.
Our machine learning AIs are mature and they're much more focused in on risk assessment. So every customer that comes into Lemonade, we have about 50 different machine learning models, making predictions about likelihood to claim, severity of a claim, likelihood to channel, likelihood to upsell, et cetera, et cetera. So we have a pretty robust infrastructure now making fairly specific and detailed predictions. And as we audit them, we are finding them to be holding true. So we our confidence in relying on these models is growing with every timing of the cycle.
The big hurdle today for us in terms of loss ratio does not lie in the domain of machine learning or AI. It's about getting regulatory approvals, and once those have been received, implementing that. So particularly in an inflation heavy environment, let's start with - before you even get to regulatory hurdles. The fact that you price a policy today and you don't get to amend it for another year other than in car where you get one opportunity midyear means that if inflation has been significant 10% as a year ago, we had let alone the 15% or 20% that you saw in the field of car and home repairs, it means that you price today and then you're fielding a claim 6 months from now, which may be 10% higher than the price when you set it, and you don't get another bite at that apple until renewal a year later.
And that is if regulators approve every bit of your filing, more likely since regulatory processes take more time and sometimes have limitations on how much rate they'll approve and at what frequency, it may take longer than a year. And then once it's approved, you then have an earning in period because everybody who was priced at the older policy you have to wait until their renewals come up, and it will take you pretty much a full year for those new rates to take effect, which is why if inflation remains high, you're in a constant race to adapt rates.
You don't have a knowledge gap. Our machines tell us exactly what rates we need to have for each risk. What we have is a time lag between when that knowledge comes in and when it actually hits our books and that dissipates in two ways. One is inflation comes down as indeed it is. It's come down significantly, and we're feeling that. And the second one is we pick up our rate of filing and approvals, and we're seeing that as well. It will remain - certainly, the inflation remains elevated, it will remain a cat and mouse game. I don't want to pretend that, that goes away, and we've seen this across the industry.
But in the event, as we're seeing these two trend lines coming one on top of the other, which is decreasing inflation and increasing rate, the two combined to give us optimism that we've broken the back of this thing and on a path towards getting to where we need to go.
Q – Unidentified Analyst
Okay. That's very helpful. But sorry, if I can just stay on that topic as my follow-up. Then in that case, is it kind of fair to imply from what you said so far that essentially, the technological development so far is somewhat of a secondary to essentially the regulatory and the macro environment, whereas the underwriting efficiency from the tech-driven underwriting really is more of a secondary and then the macro environment, such as catastrophe losses or the current regulatory environment, would you essentially nullify a lot of the tech advantage you're having. Is that the wrong way to think of that? Or how should I think about it?
Daniel Schreiber
It's a fair question. I'd draw your attention to a couple of things. One is have a look at our - the loss ratios by product and the high cat impact of this past quarter perhaps masks some of the dramatic improvements that we shared in our last - just draw your attention, look what happened to homeowners ex cat dropping from like 110 to the 60s over the course of the last few months when you neutralize cat. As Tim said, we have to pay for cats. This isn't an effort to sidestep our responsibility for paying for cats, but it does demonstrate a fundamental improvement.
And indeed, if you look at some of the best players in the industry and the loss ratios are some of the best known names, I won't name names, but everybody in this call knows who they are. And you look at what their loss ratios were for this last quarter, you'll see that we came in significantly better than some of the best-known names in the industry suggesting a competitive advantage.
So no doubt, while we are in a high inflation environment, you will see the whole industry as well as us suffer the brunt of that. That is true. But you can also see when you look kind of beneath the headline is that there is a competitive advantage emerging. And as the inflation received, I think the competitive advantage remains, it is already an evidence if you know where to look. And as times normalize, it will become more and more pronounced.
And in the long term, that is how this industry has won by being superior at selecting risks and pricing them. And I think that the technology and infrastructure that we're building affords us that advantage agreed that when there's a storm howling outside, it's hard to see that. But as the storm passed us by, I think, will become increasingly obvious.
Q – Unidentified Analyst
Thank you. That's very helpful.
Operator
Thank you. Our next question comes from Jason Helfstein of Oppenheimer. Your line is now open. Please go ahead.
Jason Helfstein
Thank you. I want to go back when you originally kind of came up with the long-term plan, whether it's - when you guys came public, et cetera. I mean look, I think regardless of your views about climate change, it does seem that storms are just seem like it's more frequent, right, whether depending on how you categorize the weather, et cetera, et cetera.
Do you feel that as a result of that, like if we're like we're starting again, you'd say we need to have a bigger business to kind of absorb the risk because it's all about kind of spreading it out. And then just how that - again, that may have been some of the catalysts in some of the recent announcements, just how you think about that now if you reflect back 5 or 7 years? Thank you.
Daniel Schreiber
I'll take a back at that and then Tim come in with anything that I've omitted. Jason, that's a great question. I think that rather than suggesting a different course of action, it reconfirms us in a multi-product, multi-geography strategy. So, yes, we saw some pretty severe outcomes this quarter. Of course, we're not just a homeowners business. That's a sizable minority of our business. It's a fraction of our business, about a quarter of our business. And the rest of our business is performing very, very well. And we shared, again, our per product loss ratios, you see what's happening in our pet business, which is now almost as large as our homeowners business. You see what's happening in our renters business, which is larger than our homeowners business.
So our multi-product and multi-geography is already mitigating the worst of those risks. Indeed, the fact that we are able to report the EBITDA that we reported a beat on the bottom line and a beat on the top line, notwithstanding a 94% gross loss ratio, I think, speaks volumes to the structures that we put in place, including reinsurance.
It's - I don't want to oversell this, but if we didn't tell you our loss ratio, and we just sold you our financials, our P&L, you wouldn't know that this was a particularly severe loss quarter. So there are structures in place that allow us to buffer ourselves from the last worst of these storms and to be able to deliver a strong EBITDA and strong top line notwithstanding.
So coming full cycle, I agree with your underlying premise, which is that major catastrophes are becoming more frequent, certainly, that has been our experience. Ultimately, insurance gets a handle on that through pricing. Once you understand risks, you can price for them, there is a time lag in doing that. We discuss the regulatory and other time lags that will allow it to course correct.
In the meantime, I will tell you, part of our slowing down, and we laid this out in our earlier comments and in the letter, is to focus on the areas that are less cash exposed. So Tim already mentioned that we are - have been for these many years, very cautious about wildfire exposure and hurricane exposure. We understood those risks, and we're pretty conservative. We have been able to write around those. As these other risks become more palpable to us, we are sidestepping them as well. We have written stuff prior that we are now paying for.
But if you were to look at ourselves these many quarters, you'll see that our new sales in exposed areas are really de minimis. So we are taking course corrective actions. We are filing for the price rate that we need. We are diversifying our portfolio geographically and by product. And I think all of that translates into a healthier business as time passes up by. Tim, anything I omitted there?
Tim Bixby
I would just add one thought, which is, again, if you sort of look in broad brush at several years since going public, probably one of the larger surprises is in this period of more what feels like in the short term, more frequent intensity of more volatile storms, we've weathered that test nicely. We've seen elevated results, but they have not been too far out of line from much larger and much larger incumbents.
I think if we've seen - it would have been more reasonable, I think, in this period with newer products and a much smaller base of premium to actually see more of a loss ratio challenge than we've seen. So I think that's good news. And then on the expense ratio side, I think if you just look at the consistent improvement in consistently declining losses relative to our premium line, really solid improvement there, too. So I think that's been a - we didn't set out for the last 3 years to be a test of this rigor, but I think we've weathered that test very well.
Jason Helfstein
And then just like a technology question. So now that large language models and machine learning is becoming more accessible to other companies without some of the hard work that companies like yourself did as early kind of, call it, pioneers, do you think that as a competitive threat because those technologies technically were not available and are going to become more available if your competitors choose to use them for the next kind of upstart competitor, et cetera? Thanks.
Tim Bixby
We're too troubled by - we're not too troubled by large competitive access to technology. It's been something that's been true forever. Certainly, there are things that are about large language models in the transition of the past year, so that are new for all of us. But having built our platform from day one in anticipation of just such data advantages only, I think, amplifies the advantages that we believe we have in place.
Daniel Schreiber
Yes, just, hi, Jason. Well - okay, I start to finish. I was just going to say, if you look back at something I commented on that at the time, but went back to Hathway held our AGM a few months ago. Ajit, the Vice Chair of Berkshire Hathaway spoke about GEICO and he said that, we have some 500 systems and then he corrected himself. He said it's actually over 600 systems that don't talk to each other. Those kinds of legacy challenges are - I don't want to say insurmountable, but they certainly make it very, very difficult to overcome the kind of challenges that Shai referenced in his comments, which is these are novel and powerful technologies, but applying them seamlessly and integrating into the operations of the company is an entirely nontrivial matter.
So just reinforcing what Tim says. I think at a headline it sounds like everybody would be able to deploy these technologies, having now in can put a lot of effort into these models and trained our own models. We rest easy that Tim's comments are exactly right. This is not a major threat to us.
Jason Helfstein
Thank you.
Operator
Thank you. Our next question comes from Mike Zaremski from BMO. Your line is now open. Please go ahead.
Mike Zaremski
Hey, good morning. Thanks for taking my question. First question is just a numbers question on the catastrophe losses on a gross basis, we're calculating it was around 21 points. Is that similar on a net basis?
Tim Bixby
Yes, it is.
Mike Zaremski
Okay. Okay. And my follow-up, you - Tim mentioned that in the prepared remarks that Metromile's churn rate is slightly higher than the rest of the portfolio. I was looking - I think Metromile's annual customer retention rate it said it was around 60% annually back when I disclosed in 2022. Is that kind of - is Metromile's annual retention rate kind of still around that level?
Tim Bixby
Yes. So we don't disclose that specifically, and some - a couple of things have changed, obviously, since we took over the book are - how we deal with customers, renewals and marketing of course is more in the realm of eliminate approach than the Metromile approach. The retention rate is somewhat better under the period of time when they've been part of eliminated since previously.
And what's been interesting is the actual premium run rate has continued even though the customer base has declined, we've not been proactively increasing that customer base. And therefore, the churn has outpaced the growth and therefore, the customer count has declined. The premium level has been fairly steady. It has declined, but at a much more modest rate than we had originally assumed when we acquired the business.
Mike Zaremski
That's helpful. Thank you.
Operator
Thank you. Our next question comes from Matt Smith of Halter Ferguson Financial. Your line is now open. Please go ahead.
Matt Smith
Hi, thanks. I wanted to stick on the Metromile theme a little bit. One of the notes you made in the letter was that the auto loss ratio, we haven't really made a lot of progress on. And it would just strike me that you probably have more kind of textured and personalized data for those customers. So I'm wondering what's the plan to kind of get that loss ratio in mind going forward?
Daniel Schreiber
Matt, good to talk to you. Thanks for the question. Yes, we have a good amount of clarity there. The bulk of a welding majority of our car business is the paper mile Metromile business and the overwhelming majority of their premiums are in California. So we've got a book here that is very geographically concentrated and there's a waiting rate approvals there, which I hope will be coming in the not-too-distant future. That would be a big unlock for that loss ratio. But with such a concentration, we're very dependent on a single approval cycle in order to get the rates back in line with the risks.
Matt Smith
So are you seeing - I mean it struck me that you're not trying to grow that piece of the business until you get the loss ratios kind of in line? Are you just trying to prove that out in the California market first and then kind of expand and try to have more bundling and other geographies?
Daniel Schreiber
We do have it in a number of states. So we are selling car in close to a dozen states. But in terms of just fixing the existing loss ratio, which is where the bulk of the results come from, you were referring earlier to the comment that said we - from our letter that said that car hasn't improved dramatically. I was just explaining why our historical book hasn't improved dramatically. In terms of new sales, we are making those absolutely in the areas where we feel our rates are adequate. That just happens to account for a small part of the book.
Matt Smith
Okay. And then if I could just switch over to the synthetic agents. You mentioned, again, the LTV to CAC ratio over 3 in your opening comment. I'm just curious, given kind of what you're modeling versus what your realized results have been. Is that - what confidence do you have in that ratio kind of holding over time given the increase in cat events that we've seen recently?
Daniel Schreiber
It's been surprisingly perhaps constant. So we've been through as a company, as an industry, as an economy some multi years [ph] but we've seen that overall being fairly stable, fairly constant, slightly improving over time. So I hesitate to say too much about the future, but the optimism we suggest that this is an area that we can continue to improve upon as our retention rates continue to improve as our cross-sell rates continue to improve as our new rates come online, I think there is room for optimism on that regard. We - I'll put it in the invar side, I see no headwinds that we're aware of.
Matt Smith
Okay. Thanks so much.
Daniel Schreiber
Sure. Thank you.
Operator
Thank you. Our next question comes from Yaron Kinar from Jefferies. Your line is now open. Please go ahead.
Yaron Kinar
Thanks for taking my follow up. On the [Technical Difficulty] appreciate you provided the - loss ratios by line. But for homeowners, you offered the total number with CAC [ph] for the loss ratio?
Tim Bixby
Yes. We chose that carefully. We haven't disclosed the home rate at somewhat more elevated and you can probably back into it by the share of business, but we've not disclosed every line item. I just wanted to show where making good progress. And so we'll hopefully share more over time. And Yaron, I'm glad you came back in, so I can correct my earlier misstatement to you and to Josh on the prior period development. So I said it was unfavorable is actually favorable by 1 percentage point. So I just wanted to correct that for the record and for the transcript.
Yaron Kinar
Got it. Thank you. And for both....
Tim Bixby
Yaron, unfortunately, your audio broke up a little bit for us. Can you just repeat that, please?
Yaron Kinar
Yes. Can you maybe tell us where the capability came from?
Tim Bixby
Nothing notable. It was only - only one point. So I wouldn't highlight any specific category. It was fairly material.
Yaron Kinar
Got it. Okay. And then maybe going back to the synthetic agent. Just want to make sure I'm so about the accounting correctly. Does the call associated with the agency, do they go above the line into marketing? Or do they go below the line and to the - on our debt payment/
Tim Bixby
So just to sort of think about the mechanics a little bit, the expense that we spend to actually acquire customers will be unchanged and that will continue to flow through the sales and marketing expense line as in the past as current and that will not change. The incremental expense, which is the return earned by General Catalyst the provider, the 16% IRR will show up as an interest expense, so that will be excluded from EBITDA, but in the interest expense line on the P&L.
Yaron Kinar
Okay. Thank you.
Operator
Thank you. Our next question comes from Tommy McJoynt of Stifel. Your line is now open. Please go ahead.
Tommy McJoynt
Hey, guys. Thanks for taking my questions. I wanted to go back a little bit to the new captives that you guys are introducing. I guess, just do you see them as strictly a form of capital efficiency? Or is there an opportunity for true risk transfer, where there's third-party capital behind it?
Tim Bixby
I think we have optionality with the Cayman captive, that's really wholly integrated, and we will basically retain all that risk on a consolidated basis. So that's really a capital-driven structure. With the Bermuda transformer, there are opportunities and structures that exist there that are not available to us otherwise, where over time, there could be interaction with third parties. That's not something that we've instituted yet, but there are opportunities there where there could be third-party involvement. And so stay tuned as we roll those out over the coming quarters, and we'll provide more clarity when that becomes more operational.
Tommy McJoynt
Okay. Got it. And then other question, just on the synthetic agent, understanding that the cadence of the deploying customer acquisition spend might be lumpy and potentially pushed out a couple of quarters. Is it fair to assume that the full maximum of the $150 million financing liability would be on the balance sheet by the end of next year, just I guess, given your trajectory of marketing spend?
Tim Bixby
So I think it's certainly possible, but I won't - because we're not giving guidance beyond the current year, I won't say that, that is our expectation, but it's certainly within the realm of reason and again, coming back to our earlier comments, the driver of our decision there is primarily rates coming online, loss ratio improvement, the underlying sort of LTV to CAC of each of the product lines.
And so if we see that improvement continue or perhaps accelerate, and obviously, we can move to growth up higher and that would make it more likely that we have that full amount by year-end. So it's certainly possible, but not certain.
Tommy McJoynt
Okay. Got it. And then just my last question. This is obviously your first kind of form of leverage that you're putting on the balance sheet. Do you expect the rating agencies to treat this financing any differently than, I guess, what would be traditional debt on the balance sheet?
Tim Bixby
Well, they - I can't speak for rating agencies as a regulator. But yes, I would expect to take into account the true terms of the structure. There's essentially no recourse other than the cash flows that result from the acquired cohorts, which is significantly different is distinct from traditional debt. That said, it is a unique structure. We're not - we're one of a fairly relatively small number of companies that are employing something like this that has these unique aspects. We're hopeful, though that, that - those distinctions will be recognized.
Tommy McJoynt
Got it. Makes sense. Thank you.
Operator
Thank you. We currently have no further questions for today. So that concludes today's conference call. Thank you all for joining. You may now disconnect your lines.+
