Making Lemons Out of Lemonade

So the Lemonade IPO has doubled and I’m not surprised. In fact, I called it. Wait, you say, when did I write that? In a half written post that didn’t get out in time because I didn’t find out until too late that the deal was coming yesterday instead of next week. 🙁

Obviously, I can’t take full credit in my prediction track record on this one, though I do have some text chains as proof for those calling BS. Why did I think it would double? We’ll get to that below. And to be clear, I’m not saying it should have doubled. I’m saying I thought it would. Big difference.

The One Paragraph To Read

If you want to know all the reasons not to own Lemonade, they are in one paragraph on p.11 of the S-1. You don’t have to read the whole document of even my Cliff Notes to understand the bear case. There are some more items in the risk factors I’ll mention but the core issues are all contained here.

We will encounter challenges while pursuing our growth opportunities and implementing our business model, and our continued success will depend on our ability to overcome such challenges. Our business model is premised on the expectation that a significant number of our customers who are renters will continue to retain coverage with us as they transition from being renters to homeowners, but we cannot provide assurances that we will succeed in retaining existing customers as they become homeowners. Additionally, our Giveback may not function as intended and may not align our interests with those of our customers to the extent anticipated and our commitment to charitable giving through our Giveback may not resonate with our existing customers or may fail to attract new customers. Further, the inadequacy of current reinsurance to protect against catastrophic losses and the unavailability of acceptable reinsurance protection in the future would have an adverse impact on our business model, which depends on reinsurance companies to absorb any unfavorable variance from the level of losses anticipated at underwriting. Future legislation may affect our ability to use artificial intelligence in our business and operations, and artificial intelligence is integral to our business model.

Lemonade S-1 p.11

OK, let’s break that down some. There are four issues highlighted: homeowners conversions, the Giveback, reinsurance dependency, and use of AI.

Home Conversions

“Our business model is premised on the expectation that a significant number of our customers who are renters will continue to retain coverage with us as they transition from being renters to homeowners.”

This is the smoking gun really. This is a confession that renters alone is not a sustainable business model. However, homeowners customers are very different from rental customers.

The (Missing) Evidence To Date

So how is Lemonade doing at growing home? Unless I missed it, they do not even tell you their home GPW or PIF. They tell you they have 12,000 condo policies, of which 10% were converted rental customers. However, condos are a lot smaller market than general homeowners.

So there is, to date, zero evidence they can be successful in homeowners, whether through new business or “graduating” rental customers.

To their credit, Lemonade does acknowledge this in the risk factors:

(O)ur user base is made up primarily of renters and we have very few instances of those renters becoming homeowners, a key element of our business model. It is also difficult for us to track that data. We cannot provide any assurance that the data that we collect will provide useful measures for evaluating our business model. Our inability to adequately assess our performance and growth could have a material adverse effect

Lemonade S-1, p.24

Not only have they not been successful at converting renters when they buy homes, they don’t even know how to track it!!! They are wishing that these customers will convert.

The limited success in condos suggests they have not figured out how to spread geographically beyond major city centers into the suburbs where actual homes are. This is not surprising because their main advantages (usability, claims AI) don’t currently extend to the traditional home market.

Renters Claims: Home Claims as Building a Car: Building an Airplane

Home claims are much more complicated than renter’s claims (and condos) and AI is not as likely to solve them. The great majority of home claims don’t even apply to renters.

The other issue is home customers care much more about claims handling than renters do because the costs and personal disruption are more significant. Lemonade has had mixed performance with claims based on review sites.

Home claims are also less likely to be solved by quickly cutting a check. That works if my TV gets stolen. It doesn’t work if my roof is damaged and – Giveback or no Giveback – you can’t trust a customer to get a reputable roofer.

No Rating = No Growth?

Recall Lemonade has no plans to get an AM Best rating. That might suffice if you are a Florida only homeowner’s carrier where there are few alternatives. It is less likely to work in major markets with plenty of options.

Customer Acquisition & Retention

And I haven’t even gotten to the big issue yet…there is little evidence that Lemonade is good at acquiring and retaining rental customers. As I mentioned in the earlier post, two year retention is <50%. Forget about converting people when they buy a home, Lemonade’s first priority ought to be keeping what they have and figuring out why they leave so soon!

But arguably, the cost of acquiring customers is a bigger problem than the retention sieve. Lemonade says they acquire $2 of premium for each $1 they spend on marketing. That is still a pretty poor outcome (it needs to be 10:1) especially when half the customers leave, so you are paying $1 to acquire $1…of premium, which obviously has associated losses and expenses. Yikes!

To clarify the point, Lemonade spent $89M on marketing last year to earn $64M of premium. In 2018, they spent $42M for $21M of EP…which is down to $10M after two years, so they’re paying $4 for $1 of EP. Double Yikes!

Summary

OK, that was a lot of ground covered to show how Lemonade is failing to transition to home. Here’s a simple way to think about it. Lemonade in homeowners today is about where Tesla was in 2006 when it was developing its first model (the Roadster launched in 2008). Maybe Lemonade will follow Tesla’s path and hit on all its ambitions, but the odds on it today are still very long.

The Giveback

“Additionally, our Giveback may not function as intended and may not align our interests with those of our customers to the extent anticipated”

Manipulating Behavior

The Giveback is a flawed concept. The idea that customers will pick one insurer over another due to a token charitable donation (~1% of EP) really assumes people are easily manipulated.

Lemonade discloses this as a risk factor.

Our business model is predicated on behavioral economics…our commitment to charitable giving through our Giveback program may not align our interests with those of our customers to the extent anticipated.

Lemonade S-1, p.22-23

In other words, we are using behavioral economics to try to trick people into signing up for our service. Behavioral economics is supposed to be used to help people overcome irrational biases to make wise choices.

Lemonade has reversed that by incenting people to make irrational choices because they don’t understand the small charitable donation is less efficient when made through Lemonade than if they did it themselves and paid the same for insurance elsewhere.

It is also worth noting from p. 23 “If even one regulator were to disallow the Giveback, it could force us to abandon the Giveback in part or entirely, either of which could undermine the behavioral economics foundation on which our business model is based“.

Flawed Fraud Prevention

Now, there is one other argument Lemonade makes for the Giveback. They view it as a way for customers to signal that they are honest people who care about strangers and wouldn’t want to hurt the pool of customers by filing a false claim and lowering the potential charitable pool.

That might work in certain circumstances, but the incentive isn’t strong enough (which behavioral economics should have taught them).

If I can get a claim paid that is say 10X my annual premium vs. a 1% donation to charity, I am not going to say don’t pay it or only pay me 90% of it because I want to help the charity. There is a 1000X difference between my claim damage and the donation! The donation isn’t important enough to change behavior.

Furthermore, if I am a bad actor who wants to get a fake claim through, I probably don’t care about your charity anyway, so Lemonade isn’t dissuading bad actors from signing up and trying to trick the AI.

Sharing The Risk

Now, let’s address the most important point. Lemonade gets their 25% whether the charity gets paid or not!!! So, if results are poor, the customers donate less to charity. If they are good, the customers donate more. Either way, Lemonade takes it’s 25% cut!

If they genuinely cared about the charities, they would make their 25% take subject to a Giveback as well so that when underwriting results are poor, they take <25% to keep the charities whole. This would give them a behavioral incentive to care about the loss ratio, which they currently don’t have.

In fact, currently they have the opposite motive. If losses are low, rather than try to maintain them there and have a big Giveback payout, Lemonade’s motive is to crank up growth at a somewhat higher loss ratio because that gets them more 25% fees. With the reinsurer’s taking 75% of the losses, the growth in fees are way more important than their share of the higher loss ratio.

Summary

Lemonade is trying to confuse customers with the Giveback to potentially act against their own self interest. The merit of the Giveback would be more convincing if Lemonade were willing to subject some of their compensation to it. The fraud signaling component is not strong enough to matter all that much.

Reinsurance Dependency

“Further, the inadequacy of current reinsurance to protect against catastrophic losses and the unavailability of acceptable reinsurance protection in the future would have an adverse impact on our business model, which depends on reinsurance companies to absorb any unfavorable variance from the level of losses anticipated at underwriting.”

Lemonade has smartly structured their model so the reinsurers take nearly all the loss ratio risk while Lemonade gets its fixed 25% fee. This is very astute and a great thing if they can maintain it. If they can maintain it.

Reinsurer Appetite

While today reinsurers are willing to accept losses from Lemonade to bet on the long term upside, it is unclear if this is a stable situation. Lemonade did secure three year cover which helps but three years doesn’t equal forever.

What would make reinsurers changer their mind? The obvious thing would be poor underwriting performance. However, I’m going to assume that won’t be the trouble.

The bigger concern is if Lemonade does successfully transition to homeowners. Home policies bring much more cat risk than renters. Lemonade is concentrated in big coastal cities. How will reinsurers feel about taking on this cat risk as Lemonade succeeds at implementing their growth plan?

Impact on Gross Margin

If Lemonade is executing well and growing, reinsurers will be happy to grow with them. But as suggested, if the cat risk gets too big, they will charge more which means Lemonade might need to reduce the 25% ceding commission.

The other option is for Lemonade to retain more if they don’t like reinsurance pricing. The problem with that is they only have a Demotech rating and “We do not plan to give up revenues or efficiency of size as a means to qualify for an acceptable A.M. Best rating“.

You can’t grow in line with their ambitions on a Demotech rating, especially if you have to retain more business if reinsurance gets too expensive.

The Tail Risk

As mentioned, there is a risk that the underwriting goes sour. In this case, the reinsurers would likely abandon ship and Lemonade would be in a fight for its survival as it doesn’t have the capital to assume all the underwriting risk, so it would likely have to significantly scale back growth. A possible outcome would be giving up on home and other lines to reset as a renters only insurer.

The Tech Halo

“Future legislation may affect our ability to use artificial intelligence in our business and operations, and artificial intelligence is integral to our business model.”

This likely relates to concerns about disparate impact, particularly in New York. If it is found that Lemonade unintentionally discriminates in approving or denying customers or especially in how it prices them, they could have to change the algorithm.

The other risk is in the claims AI. If there is a demographic trend in types of customer who have lower approvals from the claims bot, that could trigger sanctions or even the removal of the claims AI process.

Of all the things Lemonade could have highlighted as key risks, it is interesting that the lawyers chose this to be the fourth one. It suggests investors should probably spend some time researching the issue and figuring out why Lemonade views it as a bigger risk than retention or the lack of a rating.

So Why Did It Double?

Easy. Lemonade is a slam dunk holding for an ESG fund. It has the Giveback as the sizzle, but the steak is the Public Benefit structure which requires them to act in the interests of stakeholders, even if it hurts shareholders at times.

This is obviously a slippery slope (see my concerns above about the 25% fixed fee), but for now it doesn’t matter. Every ESG investor in the world is going to want to show this as a top 10 holding. A large UK ESG investor already filed that it intended to buy $100M on the IPO.

Simply put, there is far more demand for core ESG holdings than there is supply. The merits of the investment are consigned to the file drawer for now to be revisited at a later time when fundamentals matter.

This certainly gives Lemonade a window to execute while investors remain patient. However, this window will not last forever, so they need to hope none of the above risk factors prove to be actual risks.