Dispatches From InsurTech Survival Island: Five Takeaways From Statutory Financials (Part 1)

April 2, 2018 by Matteo Carbone and Adrian Jones
A version of this article, originally titled “Five Dispatches from InsurTech Survival Island,” was previously published on March 12, 2018 by Matteo Carbone and Adrian Jones on their LinkedIn pages.

All the insurance players will be InsurTech” is a phrase we have uttered on many occasions in the past few years, but some InsurTechs have chosen to be insurers. Real insurers. Which means they file detailed financial statements.

Executive Summary

With permission of the authors, Carrier Management is republishing this analysis of the statutory financials of InsurTech insurers Lemonade, Metromile and Root, originally published on LinkedIn. Originally published as a single article, we present it as a three-part series.

In Part 1, the authors present overall loss ratio, expense ratio and combined ratio results from statutory filings of the three insurers.

For the complete original article, visit the LinkedIn pages of Matteo Carbone or Adrian Jones.

These obscure but public regulatory filings are a rare glimpse into the closely guarded workings of startups.

Full-year 2017 filings for U.S.-based insurers were released earlier this month. Here’s what we found:

  1. Underwriting results have been poor.
  2. It costs $15 million a year to run a startup InsurTech carrier.
  3. Customer acquisition costs and back-office expenses (so far) matter more than efficiencies from digitization and no legacy systems.
  4. Reinsurers are supporting InsurTech by losing money too.
  5. In recent history, the startup insurers that have won were active in markets not targeted by incumbents.

We explain and show data for each of these points in this article.

Context and Sources

The most notable recent independent U.S. property/casualty InsurTech startups that operated throughout 2017 as fully licensed insurers are Lemonade, Metromile and Root.

Matteo Carbone, one of the co-authors of this article, recently published a book titled “All the insurance players will be InsurTech,” a phrase he has spoken often at insurance industry gatherings. Carbone also authored a popular article for Carrier Management last year, “UBI Is a Failure, But Telematics Insurance Is Working Extraordinarily Well.
Out of hundreds of U.S. InsurTech startups, only a few have taken the hard route of being a fully licensed insurer. The lack of interest in becoming a regulated insurer is evidence against disruption evangelists who say that insurers are going to disappear or be killed by GAFA (Google, Amazon, Facebook and Apple).

Despite the lack of interest, being a fully licensed insurer may prove to be a more durable business model than the alternatives like being an agency or otherwise depending on incumbents, and the three carrier startups that we analyze all have strong teams with powerful investors.

One of the co-authors of this article, Adrian Jones, has previously argued that being a fully licensed insurer may actually be the best long-term strategy. See his LinkedIn article, “Six Questionable Things Said About InsurTech.”

The statutory filings we reviewed provide many of the traditional KPIs of insurance companies. Startups may use additional internal measures as they scale their company. Statutory statements typically do not include the financials of an insurer’s holding company or affiliated agencies, and companies have some flexibility in how they record certain numbers. But for all their limitations, statutory figures have the benefit of being time-tested, mostly standard across insurers, and measuring critical indicators like loss ratio and net income.

The Statistics

Here are the key stats on three companies—from top line to bottom line:

The ratios are typically calculated on net earned premiums, but we also show a row where the expense ratio denominator is total direct premiums written, which may be more appropriate for growing books.

Lemonade Insurance Company

Metromile Insurance Company

Root Insurance Company

Additionally, a few other notable companies are worth a mention:

Berkshire Hathaway Direct Insurance Company sells online via biBERK.com but isn’t a venture-backed startup. They wrote $6.4 million in premium last year—their second year of operations—most of it workers compensation. Their loss ratio gross of reinsurance was 124.

Oscar is a startup brought to you by Josh Kushner and others who have plowed in $728 million already, with more funds being raised currently. Oscar shows no signs of profitability in its three main states:

Bottom Line: Underwriting Results Have Been Poor

All three companies have a gross loss ratio of near 100 or higher. (For reference, the industry average in 2016 was 72.) That means they have paid $1 as claims for each $1 earned from policyholders in the last 12 months.

An insurance startup has to prove two critical things.

We’ve been in debates over which is more important, and we always start with underwriting, since it takes no special talent to distribute a product that has been poorly underwritten (i.e., selling below cost). Underwriting quality/discipline is one of the “golden rules” of the insurance sector. Disrupt it at your investors’ peril.

Building an underwriting model is like playing whack-a-mole with a year’s time lag. Sometimes it’s difficult or impossible to address even widely known underwriting issues.
Underwriting returns have generally been poor, in some cases awful. (Double the industry average counts as awful in our book.)

Poor results are to be expected at first, even for the first several years. A single big loss can foul a year’s results in a small book. It can be hard to tell if that single loss is an anomaly or a failure in the model. Building an underwriting model is like playing whack-a-mole with a year’s time lag. Sometimes it’s difficult or impossible to address even widely known underwriting issues.

Other articles in this series:

For the complete original article, visit the LinkedIn pages of Matteo Carbone or Adrian Jones.

The authors’ opinions are solely their own, and only public data was used to create this article.