Contents:
- Introduction
- Market Dynamics
- Telematics & Underwriting
- Customer experience
- Customer acquisition
- Full stack tech stack
- Machine Learning
- Financials
- Claims Management
- Future Growth
Did you know that your driving micro-habits, how you turn, accelerate, brake, and how much you drive can all be used to give you a risk score? The full-stack insurer Root Insurance uses this type of risk score to determine how good a driver you are and then assigns you a premium value for your motor insurance policy. Root aims to raise more than $6BN in its IPO in 2021, and while that looks lower compared to its Fintech cousins, this is indeed the coming of age of insurtechs.
Insurance industry professionals all agree that they work in an industry that has been least impacted by technology disruption - for now! Things have been changing in the last few years, and we get a chance to peek under the hood of one such company - Root Inc, that has recently filed for its IPO in the US. I had to wade through 100s of pages of its IPO filing so that you would not need to!
The first point that drew my attention to the company is that it is headquartered in the great state of Ohio. That is an unlikely place to find a disruptive tech-driven business. Still, perhaps that remains the state's competitive advantage in accessing talent at a lower cost than in San Francisco. I have an affinity to the Buckeye state, having lived there for four years. Go Buckeyes! Having gotten that out of the way, let us dive deep into what makes Root tick.
Market Dynamics and what it takes to succeed:
The US Auto insurance market is around USD266 BN in size and largely depends on the law of large numbers for underwriting their risks. In this scenario, good drivers and bad drivers are lumped together and usually get a similar premium amount. The sound drivers subsidize the claims of the wrong drivers.
How good would it be if an insurance company had a way of attracting (positive selection) good drivers and offer them lower than market rates? In theory, you would get lower claims, and your customers will leave your traditional competitors. For this strategy to succeed, you would need to move away from the traditional distribution channel of agents in the US. These channels need to be paid high commissions for distribution. The insurer's technology stack needs to be built ground up, taking advantage of modern cloud-based microservices architectures that do not tie you down with legacy costs of operation. That is essentially Root's strategy.
Telematics & behavioral underwriting:
The use of telematics is nothing new, but Root has gone a step forward and avoids the use of onboard devices that have to be plugged into the vehicle. Instead, it uses mobile telematics that uses the sensors and accelerometers built into our mobile phones. This is at the heart of Root's risk modeling that uses a short test drive to determine a driver's risk score. Get less risky drivers, reduce the cost of your claim- simple as that! The data science behind behavioral underwriting tells us that the most dangerous 10 - 15% of drivers are up to two times more likely to get in an accident than a "good" driver as determined by the risk score.
Mobile-First Customer experience:
Root is a mobile app and chatbot driven system at the front-end. Onboarding can be done as quickly as 47 seconds and is paperless. This process provides an advantage in a lower cost of quote generation and policy issuance. The acquisition cost should be lower in theory, but Root spent 37% of its revenue on sales and marketing expenses if you look at the financials for 2019. As is the case with other fintechs, customer acquisition at an early stage is expensive, with dollars being spent on Google and Facebook rather than an agency network.
Direct Customer Acquisition:
Root acquires 75% of its customers directly through social and online channels, resulting in less commission payout and lower acquisition costs. Currently, the renewal premium is 47%, which is lower than the US industry average of approximately 80%, so there is an opportunity to improve. This could result from Root's Machine learning models getting better and rejecting a larger percentage of its existing customer base during renewals.
Full Stack insurance tech stack:
Root should have a lower operating cost and have the advantage of being quick to deploy technology changes to suit changing customer requirements. Root claims to have no dependency on external suppliers and long release schedules. The use of cloud architecture helps them to scale up with load without buying expensive hardware. This logic seems to hold; till 2020, the IT costs were 11.1% of the revenue.
Claims Management:
Root relies on its mobile app to complete claims settlement. Customers can upload photos of the damage, and Root generally pays out swiftly. They claim to have a five-day cycle time versus an industry average of 12 days.
Advantage: Reduction in claims handling costs by reducing the need for manual surveys. The other benefit of using its mobile app data (timestamp /GPS location) is that Root can weed out fraudulent claims. It is well known in the industry that the more the time difference/ distance from the accident site, the more the chance of a fraudulent claim. Root's loss adjustment expenses are 9.5% compared to the industry average of 11%. The quicker payout increases their net promoter score, and this drives more business their way. That is a virtuous flywheel right there.
Capital Light model:
Root uses reinsurance to take the risk off their balance sheet periodically. They have a captive agency that gains commissions for sales. They use RI quota share to cede premium to RI (Captive Re partner in Caymans is called Root Re). At the moment, they cede out a whopping 70% of their premiums, thus retaining very little of the risk. This model allows them to be capital efficient and could be changed to increase premium retention once their machine learning algorithms become better at identifying lower-risk customers.
At the moment, they can write four dollars of premium for every one dollar of capital. They earn commissions from the other Reinsurance partner. Having a captive RI company helps manage their balance sheet when other reinsurers hike up their rates or refuse to accept Root's contracts for any reason. The business model becomes robust with this vertical integration. This model alleviates the effect known as "Capital Drag" that can be particularly punitive for high growth insurers.
Use of Machine Learning:
Root’s use of Machine learning, drives better underwriting decisions and fraud prevention. The more policies they underwrite, the better their models get. It needs a long term vision to stay in this mode of operation. Initial years will have losses. To get you an idea of the way their models are improving:
- The first generation of their pricing model was built in 2016 and had four times the predictive power of an external benchmark. The first model used only GPS data and leveraged 100k trips with 2 Million miles of data.
- Their current (v 3.14 ), version of the algorithm uses 10 Billion Miles and has ten times the external benchmark's predictive power. This was built in 2020.
You can easily see the massive advantage of Machine Learning as more data is fed through the algorithms.
Growth into adjacent products:
Once you own the customers, you get into other products like Home and renter’s insurance. The US Personal Lines market is USD370 BN in size, and this is the next Target market for Root. Currently, Root is active in 30 states, and as it expands its presence in all the 50 states in the US, you can expect the premium volumes to grow. Therefore, there are substantial untapped avenues of growth.
Financials:
So the big question on your mind is- is Root making profits? Take a look at the financials. It is growing both in the number of policies and the Gross written premiums. Its loss ratio has reduced over the years, but it is still making losses. This should improve over time as it gains more data, and its machine learning models improve over time. The Renter's portfolio is small and can increase as the cross-selling processes get better over time.
Expenses:
The majority of Root's costs go towards paying and handling claims (92.6%). Sales and Marketing take another 21.7% of the revenue. "Other insurance expenses" are a decent chunk, and the company filings state that this is due to higher underwriting costs and premium write-offs. This is possibly due to the higher salaried data scientist and actuaries.
Their technology and development expenses are a reasonable 11%. So unless the claims loss-ratio is brought down to the 60s and other costs are reined in, this full-stack insurer is unlikely to turn profitable.
Future Growth:
1) Root is currently expanding into the home renters insurance market using their existing motor customer database. The only issue is that they are achieving low single-digit conversion (<5%)
2) Using their tech stack, they can expand into other P&C business lines in the US (~USD 400BN). This will require collecting specific data that drives risk, and that process is not always straightforward. The insurtech Vitality follows this approach for health insurance products by collecting customer fitness levels using trackers resulting in lower premiums for healthier folks.
3) Root can license their technology to other insurtechs and insurance players globally. This is a software product play that many others like EIS and others
Going by the success of Lemonade and liquidity in the market, Root's IPO should do well. This IPO will allow the current investors to cash out. Whether Root will do well in becoming profitable and gaining market share remains to be seen. They seem to have built the foundation for creating a strong insurance company in the competitive US insurance industry.
References:
https://techcrunch.com/2020/10/20/root-targets-6b-valuation-in-pending-ipo-a-boon-for-insurtech-startups/
https://www.sec.gov/