Insurance as Lemonade

Allianz invests in Israel-based player Lemonade. This was big but not surprising industry news last week, and Accenture has Lemonade on the radar since it was founded last autumn. In fact, Lemonade might be the world’s most-hyped Insuretech company of today. Allianz strategic investment makes it time to take a sneak behind the hype and try to understand how Lemonade is innovating the traditional business model of property and casualty insurance.

The traditional P&C insurance business model, put simply, is to compensate their customers for their loss or damage in specified events and to take monthly premiums from their customers in return. Without any loss event actually occurring, this business model would apparently be a very profitable one. This potential profit is partially eaten up as soon as the customer files a claim, and the insurance company grudgingly pays the contractually agreed compensation. Each customer claim adversely affects the P&C insurers business case and the key metric of combined ratio, which figures the total claims cost as a percentage of premium income. This is why claim settlement usually involves a lengthy process with the insurer examining if all contractual conditions are really met by the customer.

From a customer experience perspective, this has always been the unresolved riddle of P&C insurance: how can a positive customer experience be delivered if in the key moment of truth – the claim – the primary interest of the provider lies in not fulfilling the customer’s need? So far the answer was: not really. Despite all customer experience efforts, the diametrical opposition of interests remained a fact.

Lemonade aspires to resolve this opposition by killing the combined ratio as key metric: rather than trying to minimize the combined ratio (that is to maximize the excess of premium income over total claims cost), they are taking a fixed fee of 20% and give away the remaining excess of premium to charity. By consequence, the profit will be fully based on customer acquisition and retention, but the combined ratio will always be 100% – as long as claims cost do not exceed premiums (this is a very problematic assumption and seems to imply heavy reliance – or even dependency – on reinsurers, as is explained elsewhere) To make it even more appealing to customers, the customer itself is to choose which charity should be supported by his premiums.

Lemonade hopes that this approach will radically reduce fraud (and thus in turn significantly cut claims costs, as traditionally a major part of claims is believed to be fraudulent): customers who no longer believe that the insurer will actively work against them and try to find reasons for not having to pay for the claim will themselves feel less inclined to cheat. This psychological phenomenon of mutual cooperation is actually well established in Game Theory and Behavioural Economics – now wonder that Behavioural Economist Dan Ariely has joined Lemonade’s advisory board last year.

In addition, customers who know that of their premium dollars not spent on claims will be donated to their chosen charity, will also not be encouraged to trick the system.

Lemonade uses the charity construct to group together peers who are supporting the same charity. This opens up the possibility to add a third driver of anti-fraud motivation: group pressure. If I file a fraudulent claim, I will not only cheat the charity, but all peers supporting it. In the future, this peer group element might be elaborated; at the moment however, it seems mistaken that Lemonade is called a peer-to-peer insurer by most sources. A proper peer-to-peer model is exemplified by German player friendsurance which motivates actual (i.e. people knowing each other for real) social groups not to file claims by guarantee a significant payback to all group members in case no (or only little) claims were filed.

On the face of it, I find such peer-to-peer models more appealing than the charity element Lemonade has constructed, as I would expect it to address more effectively the three twists of customer economics which insurance traditionally faces:

1. The first twist of fraudulent claims has already been addressed: yes, the charity aspect might mute fraudulent energies, but the P2P idea of harming actual friends appears more effective
2. The second twist is about so-called Moral Hazard: to keep premiums and claims in the right balance, insurers calculate premiums based on the actual risk. The problem is that this risk is not a static number, but typically increases as soon as insurance coverage comes into play: if you know that any scratch on your car will be paid for, you will be less thoughtful on where to leave your car unattended for some hours. Again, the P2P idea that careless behavior is potentially harming your friends appears more effective.
3. The third twist is about Asymmetric Selection: since risk, as we have just seen, depends on personal behavior, it is highly individual: some people are by nature more careless than others, and some are forced to live a riskier life, for example by living in an unsafe neighbourhood. For a given premium and coverage, an insurance contract is the more attractive for you, the higher your personal risk is. Consequently, lots of bad risks (i.e. customers with high risk profile) will buy the insurance contract, lots of good risks will not. If the insurance company had calculated the premium on the basis of some representative risk average, she now has a big problem, as claims cost will be much higher than calculated. Here I don’t see at all how the Lemonade model will help to resolve this twist, while pure P2P models are also effective on this one: if you know your fellow is a big risk-taker, you might go parachuting with him, but maybe not commit to co-covering his risk.

While the charity component will not be effective, it seems that Lemonade has a different remedy in place: as you would expect an InsurTech-player to do, they aspire to employ state-of-the art Artificial Intelligence and Analytics capabilities to assess individual risk profiles. Any insurance company taking the lead on this one will truly be a game changer, and it has yet to be seen if Lemonade is really championing this discipline beyond marketing talk. Championship of marketing talk has already been accomplished by boasting about allegedly setting the world record of settling a claim super-fast. Quite mysteriously they claim to employ 18 anti-fraud algorithms in this rocket-fast process, without specifying what kind of checks the algorithm is performing. Among industry insiders there is a rumor circulating that Lemonade uses latest technology to detect signs of fraud in voice patterns and facial expression. Since part of their customer interaction is executed via Video Chat this seems plausible.

So after all, Lemonade might overturn the insurance business by virtue of charity and mutual trust, but by continuously profiling their customers in a depth we have not seen before. This could work.