Lemonade: The Spotify of Insurance?
Q1 2025 ER
This is an update of my original Lemonade deep dive.
Lemonade has some odds of taking the whole insurance market over the long term. My view of the company has turned more bullish.
Since I wrote my Lemonade deep dive on in Q3 2022, Lemonade’s IFP has grown by 65% while operating expenses are essentialy flat. This demonstrates two things:
My initial impression of Lemonade’s culture is likely not as accurate as I thought. Fundamentally, it seems to be a better organisation than I initially estimated.
Lemonade’s platform is highly scalable and can therefore grow unreasonably going forward.
AI is thus certainty enabling Lemonade to do more with less, which speaks well of its engineers and of management’s ability to deploy capital efficiently. However, the main concern is whether AI enables them to price risk better than analogue competitors. I do not have visiblity on that yet. However, per their achievements over the past three years, I assign meaningful odds of management figuring it out.
At present Lemonade’s unit economics do not work, as evidenced by the negative free cash flow per share below - but it’s converging. The question is how much of this convergence can be attributed to operating leverage in terms of the mechanics behind the insurance business versus how accurately and quickly Lemonade is able to price risk. In my view, now that I see the value of Lemonade’s infrastructure, I believe the company has two paths forward to superior risk pricing:
Continue effectively subsidising customers until sufficient scale is achieved, in order to generate a dataset rich enough to obtain valuable signals.
Somehow acquire a dataset that gives them an in-depth, multidimensional comprehension of their user base, as would be Meta’s social graph.
The latter path is relatively unfeasable and Lemonade management has been guiding for EBITDA breakeven in 2026 for years. Which means they’re on path number one. As I will break down below, I think they’re doing a good job executing on this strategy. CEO Dan Schreiber made some insightful comments about Lemonade’s infrastructure in the Q1 2025 earnings call:
And when you are selling insurance using AI, that in real time can use all of the signals that we have. And Tim alluded to this in his answer to your question, right down to do I want this customer? How much am I projecting them to be worth over their lifetime? How much would I invest upon them?
To the best of my knowledge, there isn't another carrier in the nation, perhaps in the world that has that kind of capability.
Additionally, I believe that not having a dense social graph doesn’t mean that they can’t outcompete insurance incumbents, but it does expose them to the risk of a better pricing engine coming along, driven by data from dense social graphs. In other words, Meta and other social media platforms could in theory disrupt Lemonade down the line. In order to succeed long term, Lemonade needs both the infrastructure and the richest dataset in order to train the best risk pricing AI model on Earth.
However, since I last reviewed Lemonade the gross loss ratio has been declining speedily, which suggests that Lemonade is pricing risk better every quarter. Operating expenses do not factor into the loss ratio, which solely measures the relationship between an insurance company's claims paid and the premiums earned. The loss ratio is calculated by dividing claims paid by premiums earned and the decline. If this trend continues, together with the aforementioned operating leverage, Lemonade is likely to achieve sound unit economics down the line.
Notably, in Q1 2025 the gross loss ratio would’ve been 59% excluding the impacts from the California wildfires. Although the wildfires are part of the business, it points to a potential quantum leap in underwriting efficiency QoQ. We will know in the quarters to come, but overall progression remains appealing.
Networks tend to follow winner-takes-all dynamics. Now that Lemonade has exhibited clear operating leverage, it could evolve into a winner-takes-all should it ever unlock the ability to price risk better than in incumbents. Long term, if the name of the game is pricing risk via a vertically integrated AI infrastructure that leverages data, I’d be concerned as a shareholder about the ability of companies with rich social graphs to disrupt the operation. Indeed, all my picks either have or are working towards proprietary data moats in absolute terms - and not just relative terms.
While Lemonade has made notable progress over the past few years, cash from operations (blue line, left axis) continues to be negative. Although total cash (purple line, right axis) remains markedly higher than total debt (green line, left axis), the negative cash production adds a considerable component of financial risk to the thesis.
Over the past year Lemonade’s price to sales ratio has essentially doubled, which means the market has taken note of its good fundamental evolution. At a current price to sales ratio of 4, Lemonade would be an interesting investment thesis if I had clarity on its ability to price risk. Since I don’t, I continue to prefer other investment theses in which I have absolute clarity, like Palantir, Spotify, Hims and Duolingo. If the gross loss ratio in Q1 2025 excluding the wildfires is driven by a jump step underwriting efficiency gains, we’ll likely see the P/S ratio go much higher.
Until next time!
⚡ If you enjoyed the post, please feel free to share with friends, drop a like and leave me a comment.
You can also reach me at:
Twitter: @alc2022
LinkedIn: antoniolinaresc
I think one thing this misses is that each state has different laws around what data can be used in underwriting, as well as how it can inform pricing. Additionally, any insurer’s rates are publicly disclosed. This leads to a commoditization of how risk is priced.
That said, there is alpha in the fidelity of data collected and how it is analyzed. Telematics collection has been around for decades, but Root and Lemonade are better at it than Progressive due to their proprietary technology.
I’d highly recommend digging more into the loss adjustment expenses LMND. These are expenses spent to handle claims. The tech really shines here and provides significant operating efficiency in an industry that has tight margins.
Hey Antonio, thanks for sharing your thoughts. I wanted to drill into your comment about their culture - "Fundamentally, it seems to be a better organisation than I initially estimated." - would you mind sharing what made you think they look like a better organization? Is it based on their improving operating leverage?
I went through their Glassdoor reviews for a bit, and they are full of comments about mismanagement, high turnover, unrealistic application of metrics, and bureaucratic culture. None of these should happen for a successful tech company. Curious to hear what you think!