Blogs

Auto insurance gamification can save lives and fuel, while helping society

Digital Operations / Internet of Things and Connected Car Executive Consultant, IBM

https://kapost-files-prod.s3.amazonaws.com/uploads/direct/1436979142-1931-8233/AutoInsurance_Game.gifMany of today’s cars are rolling sensor platforms. They have hundreds of sensors generating large amounts of data—for example, late-model Hondas can generate several gigabytes in an hour. With over 80 million new cars being made each year, their sensors may likely generate hundreds of exabytes/year.

Today, the vast majority of this data is discarded for a number of technical and economic reasons including mobile data plan costs. However, there are many valuable applications for using Internet of Things sensors and real-time analytics technology to collect and make sense of this data including predictive maintenance, intelligent transport and so on.
 One such opportunity is usage-based insurance (UBI) and its pay-as-you-drive (PAYD) and pay-how-you-drive (PHYD) variants.

UBI began as an effort to collect data to enhance insurance rating and billing. PAYD applies to the quantity of a person’s driving—that is, the amount of the insurance charged is based on actual use. This pricing model is analogous to metered billing for electricity. PHYD is based on the quality or style of a person’s driving. This model considers driving style and whether the driver is aggressive or makes good decisions while driving. And UBI is where I see an opportunity to improve on current approaches to save lives, make a positive impact on society, help the economy and help the environment.

Going beyond driver data for insurance billing

Typically, UBI collects vehicle telemetry and brings it to the cloud for analysis. The in-car portion of this system is usually an onboard diagnostics (OBD) dongle or an aftermarket telematics black box that publishes telemetry to the cloud through a mobile network. Don’t be surprised to see automakers moving to collect and distribute this telemetry to partners and insurance companies through built-in telematics systems and the automakers’ service delivery platforms. Telemetry from the car is sent, data is analyzed and the drivers are scored. If any insight is provided back to the drivers, it is done in a retrospective fashion through email and the web that tells them their current score with limited—if any—explanation.

UBI today offers black-box solutions—literally and figuratively—that are not part of the driving experience. Drivers gain no insight into what this solution is doing and what it determines about their driving until they get their insurance bill. Then they discover that their rates have been bumped up or reduced, or in some cases their coverage has been cancelled. And some insurance providers have started backpedaling away from previous promises that the telemetry will only be used to reduce rates, not raise them. As a result, we’re in danger of a shift in consumer perception of UBI as a little black-box snitch. The way the telemetry is being interpreted is usually just that—an interpretation and not necessarily a fact-based assessment. Decisions have been made to code some assumptions:

  • A bit heavy on the gas pedal? Rate as a high risk.
  • Chronic hard braking? Increase the premium.

This method of coding attempts to infer risk from telemetry, but it may sometimes use assumptions without a fact-based model developed from mining such telemetry and correlating it with events. Instead, in these early days of UBI what you may be seeing is a codification of opinions. For example, such models may score my wife as a high-risk, aggressive driver but score me as an extremely safe, conservative driver—because according to my wife and kids I drive like a grandma. But these assessments don’t reflect the reality that my wife is hypervigilant, possesses cat-like reflexes and has never been in an accident, whereas I’ve had a few fender benders—thankfully, none serious.

Regardless of how the telemetry is interpreted and the validity of the underlying assumptions, there is a bigger issue at stake than opinionated coding. We’re taking real-time data and using it for historic analysis. But in the real world, timeliness has value; latency has cost. We’re taking this real-time data and translating it into aggregate scores that are presented long after the fact, which is akin to giving your children generic feedback on something they did weeks ago and couldn’t possibly remember. “Timmy, your mother and I didn’t say anything at the time, but do you remember when you chewed with your mouth open three weeks ago? No? Well, we’re docking you your allowance because you did.”


Innovation through immediate, real-time feedback

A way to apply UBI does exist as a force for good that can help consumers, insurers, society and the environment while saving lives. Engage drivers and give them real-time feedback. Timely incentives help challenged drivers learn to become good drivers, and help good drivers become great drivers. In addition, encourage more efficient driving styles than are currently practiced and at the same time reduce fuel use and the carbon footprint.

Some great examples demonstrate how real-time feedback causes people to drive differently. For instance, the Toyota Prius presents real-time visualization of energy use that causes people to adopt a more efficient manner of driving. Prius drivers are very aware of the relationship between their driving style and their energy efficiency—mentally translated into gas cost—and it becomes a game of “how efficiently can I drive?” These drivers can see the immediate and obvious relationship between changing their driving style and the practical result, which is primarily efficient fuel consumption. OEMs take this concept a step further and also visually represent fuel cost and carbon emissions.

The key here is the immediacy of the feedback. Drivers can experiment with adjusting their driving behavior to find the efficiency sweet spot, and they are also learning a safer style of driving by adopting efficient driving habits. I suggest that efficient driving is also safe driving.

Make a game of safe, efficient driving

What if we were to take that real-time feedback into the social realm by using gamification such as Waze, a community-based navigation app. We could even add this real-time feedback to Waze. Waze has been quite the phenomenon. It crowd sources map and traffic data and turns every Waze user into a traffic probe. It uses game mechanics and dynamics to drive the collection and quality assurance of traffic information that the Global Positioning System (GPS) alone can’t provide such as construction zones, traffic cameras, police incidents and accidents.

People could compete to be the safest and most efficient driver of the day in their city, their family, their neighborhood, their sports club and so on. The game can even handicap drivers based on the type of vehicle they’re driving—just like in sailboat racing—by comparing driving styles used in dissimilar vehicles. Timely feedback can be critical for effective driver engagement.

IBM has determined that 60 percent of Internet of Things sensor data loses value within milliseconds, and much the same can be said for the driver’s telemetry in this context. Provided feedback based on historical events—such as rolling through a stop sign a dozen turns ago or the previous month—lacks the context to be helpful and may even be viewed as nonsensical.

To work properly, we need to know the driver by determining who is in the driver’s seat. Such an application would fall apart if it couldn’t distinguish among different people. The resulting analysis can become just as schizophrenic and useless as movie recommendations in my house based on aggregating the different tastes of all family members. Attempting to infer identity based on driving style is ill advised. Smarter ways exist to establish driver identity through facial recognition, or a combination of seat and mirror settings, in-seat weight sensors, mobile near-field communication (NFC) and so on.

Accomplishing the analysis of data from driving behavior at high volume with low latency to be a responsive experience is not a trivial challenge. IBM technologies can tackle such challenges through the Internet of Things using apps on IBM Bluemix with IBM InfoSphere Streams that can perform geospatial analysis in 25μs at high volume. In addition, open source Internet of Things Eclipse Paho Message Queuing Telemetry Transport (MQTT) protocol JavaScript with IBM Mobile helps deliver that information rapidly for a responsive driver experience.


Insight into convergence

UBI started as an effort to get better data for insurance rating and billing than was possible previously. But like most things in our increasingly interconnected world, other opportunities abound for it to have so much more impact for consumers, society and the environment. IBM Automotive 2025 offers additional insight into other ways convergence is expected to affect us all.