Zendrive has partnered with leading actuarial firm Milliman to analyze 30B miles of driving data and power one of the industry’s most predictive risk models.
With up to 6 times the predictive precision of the auto insurance industry’s leaders, we’re already ahead of the curve. And we’re just getting started.
While most industries have witnessed massive transformations in the way information is understood and analyzed, the world of auto insurance has for decades been largely unchanged.
The first commercial insurance policy was issued in 1898, to protect drivers and passengers from financial ruin in a possible auto incident. Fast forward over a century later: roads, vehicles and drivers have changed, but auto insurance hasn’t changed much.
The stagnation in the insurance industry has led to a surge in auto loss ratios. In 2016, they reached their highest levels in 15 years, meaning that insurers are paying more and more in claims. For 2017, things are looking even worse, with S&P predicting combined ratios of 106% and 110% for personal and commercial auto insurers respectively, meaning that insurers lose $6-$10 for every $100 of auto premium they write.
Panicked by depleting reserves and burdened by legacy policies, most insurance companies do not have the luxury of digging down to the root cause of the problem. Instead, they raise premiums across the board to make up for previous years’ losses, or exit the auto insurance category altogether.
As a company focused on saving lives by using data to improve risky driver behavior, understanding the root cause of why and how crashes happen has always been the crux of what we do. This year, we decided to put our hard work to the real test when we partnered with Milliman to construct a model predictive of collisions.
Knowing how powerful our tool is for identifying and coaching risky drivers, we had high hopes. Yet even those were surpassed when Milliman helped us improve our calculus, resulting in a 13.8x best to worst relativity — meaning that Zendrive’s driver scoring system reveals that the worst 10% of drivers are 1380% more likely to get into a collision than the best 10% of drivers.
Zendrive and Milliman constructed a risk score (“DeepData Score”) using Zendrive data on over 3 million drivers and several billion total miles driven between December 2016 and March 2017. A relativity was then calculated based on the relative difference in crash frequency per million miles by risk score decile.
Milliman found this is up to 6x more predictive than current leaders in the industry.
To this, the contrarian might wonder: ‘If it is so obvious, why hasn’t anyone done this already?’. To be honest; they have – just not as effectively. Progressive Insurance introduced “Snapshot” in 2011, a new way to price car insurance that gives drivers a custom rate based on actual driving behavior. Progressive revealed that “Drivers with the highest-risk driving behavior have loss costs that are approximately 2.5 times higher than drivers with the lowest-risk behavior”.
Since then, Progressive led this area with great success, growing both in size and in profits, jumping 450% in market share based on premiums since 1998 when the first few miles were logged. Others followed: competitors like Allstate created similar offerings and invested hundreds of millions of dollars in building competing datasets and actuarial analytics for pricing and underwriting.
When we began working on this project, we knew of some of the inherent advantages we had over current leaders in the industry. In contrast to hardware-based telematics companies such as Progressive, Allstate, and LexisNexis, Zendrive collects information directly from a user’s mobile device. While hardware solutions are only able to measure the vehicle’s movements, and are typically shipped back to the insurer after a few weeks of monitoring, Zendrive is able to understand the driver’s actual behavior, and monitor it continuously, including distracted phone use.
This turns out to be a crucial differentiator. Sheri Scott, Principal & Consulting Actuary at Milliman, noted, “The inclusion of smartphone-based information not captured in traditional telematics data contributes to the increased predictive power of this model.”
Zendrive’s ease of deployment allows us to expand at an exponential rate. Just a few months ago we announced that we had hit 15-billion miles; now we’re at 30-billion miles. As a point of comparison, it took Progressive 18 years to collect 15 billion miles.
Our massive data fuels our accuracy and helped us surpass industry leaders.
“A higher volume of data and the ‘freshness’ of the data also increases the predictive power of the data. Traditional U.S models have been based on crash data from 10 or 20 years ago to be able to amass a sufficient volume of data to create a telematics score. Older data is less relevant than data collected within the most recent year. Such a large, fresh dataset allowed for more granular analysis and to explore many combinations of behaviors, while maintaining statistical credibility.” Scott writes in Milliman’s comprehensive analysis of Zendrive data.
Accelerating at a rate of over a billion miles per week from our millions of users across the U.S. and Europe, there is still much more potential to improve the accuracy of our model. The more miles we gather, the more equipped our AI system and data scientists are to perfect our already very accurate risk-assessment algorithms and build out new ones that were once completely unattainable to the industry.
We had to step back and take a deep breath at that. This means that insurers can now leapfrog decades of costly data collection by partnering with Zendrive. It means that the negative trend in combined ratios can be reversed — making auto insurance a money-maker rather than a loss-leader.
One incumbent that has been quick to catch on is Atlas Financial Holdings, Inc. (NASDAQ: AFH), a leading passenger fleet insurance company with whom Zendrive announced a partnership on August 22nd. Atlas is deploying Zendrive’s extensive platform to its customers via mobile apps, so it can measure dangerous driving behaviors in real-time. Atlas plans to use Zendrive’s granular data to reinvent coverage for fleet and transportation drivers.
“Granular analytics at the driver level enable us to quantify potential risk with a level of precision that would not otherwise be available,” said Scott Wollney CEO of Atlas. “The Zendrive platform and their enormous data set, coupled with wide scale deployment via mobile app, are big reasons why this is possible. The platform can even motivate positive behaviors that cannot only reduce insurance claims, but help prevent crashes.”
Zendrive’s CEO, Jonathan Matus agreed, “By working together, we can enhance economic value for insurance companies and their customers, while helping make roads safer for everyone.”
Although we do not claim to know what the future will hold, we do have a hunch that the next few years hold an inflection point in the way that auto insurance companies operate. As more and more insurers continue to adapt, those who resist change will not only continue to lose money by not being able to price risk correctly, but they will also be adversely selected against.