Are you thinking about venturing into the usage-based insurancemarket before your best drivers are poached byother carriers? Make sure you understand the data implications.Accurate driver data helps you attract and retain the best riskswith pricing that reflects the exposure underwritten. On the otherhand, inaccurate data is like driving on a dark, rainy night withno windshield wipers or headlights: You may not reach the expecteddestination. Read on about three key UBI data considerations:

1. Measure the Person or the Car?

Brilliant: Compiling a complete picture of thedriver's activity
Bombing:Compiling vehicle data that doesn't differentiate who isdriving

A vehicle is not risky until a person steps into the driver'sseat. That's why auto insurance underwriters place much greaterfocus on the driver than the vehicle driven. For example,let's look at a 23-year old California driver, Jack, and hisvehicle, a 2009 Honda Accord. Underwriters don't want to know howfast the Honda can accelerate. They want to know about Jack. Doeshe drive during high-risk times? Does he exhibit reckless behaviorssuch as speeding, hard braking and sharp turns? How many accidentsand violations has he had?

With the driver as the central underwriting focus, it's ironicthat the vast majority of telematics devices have focused on thevehicle. Onboard diagnostic devices that plug into the vehicle onlymeasure the car. Any driver can be behind the wheel.

Conversely, the smartphone facilitates the first insurance UBIplatform focused on measuring the driver's actions. The great thingabout smartphones is they're personal and rarely left behind. Infact, according to ZipCar's annual millennial survey, 40% ofmillennials believe that losing their phone would be a biggerhardship than losing their car, TV or computer/tablet.

2. Complete or fragmentedview?

Brilliant: Gaining an interrupted, continuouspicture of the driver'sactivity
Bombing: Measuringonly a fragment of what really happened

One would assume that a 360-degree view of driver activityprovides a more accurate basis for underwriting than a 180-degreeview. Let's examine this hypothesis using Jack's situation.

• Jack drives his own Honda Accord. Sometimes his girlfriendHannah also drives his car.
• Hannah drives a Toyota Forerunner. Jack drives the Toyotawhenever the two are travelling together.
• Jack also frequently drives his parents' Suburban.

What portion of Jack's driving activity can be viewed andanalyzed by an insurer?

If the insurer uses an onboard diagnostic device (OBD), it willcollect data when Jack drives his Honda Accord. It also willcollect data when Hannah drives the Honda Accord. The two sets ofdata will be integrated as one. Instead of seeing Jack, the insurerwill analyze, rate and price for Jack-Hannah. Jack's activity inthe other vehicles will never be considered.

If the insurer uses a smartphone-UBI platform, it will collectdata on Jack's activity in all of the vehicles he drives, as longas he has his smartphone along.

What if Hannah is driving Jack's car? Over time, throughsensor-fusion technology, machine-learning and pattern recognition,smartphone-UBI platforms will have the ability to develop a uniquesignature for each driver, and detect deviations from normaldriving pattersn for each driver.

Moving beyond the issue of multiple drivers and vehicles,significant data gaps occur with OBD monitoring, such as:

• Vehicle speed is often measured by on-wheel sensors, which canvary by up to 20% based on inflation levels and other factors.
• Speed values read by an OBD device are not time-stamped, makingit difficult to reconstruct what happened.
• Drivers can unplug the OBD device at any time.
• Data gaps occur when the OBD device fails to connect with celltowers at trip inception often due to variation in the OBDsleep-mode wake-up variances in different car models.

Smartphone monitoring has opportunity for data gaps as well.Most notably, if the smartphone app requires the user to turn theapp on and off, the insurer may not get the complete picture.Fortunately, some smartphone apps have auto trip detectionovercoming this challenge. Likewise, if the driver turns the phoneoff completely, tracking is interrupted. However, the UBI platformwill record the exact time and location the app was turned off andon again, allowing for time and distance gap handling.

Cell service availability is generally not an issue forsmartphone-UBI. The smartphone is equipped with an internal memorythousands of times larger than that of an OBD device, allowing itto store up to two weeks of average driving data. Smartphones aredesigned to seek out data connections when they are available, soUBI driver data is automatically uploaded to the cloud as soon asthe driver re-enters cell or Wi-Fi range.

3. Is Information Actionable orJust Accessible?

Brilliant: Starting with a data “control” groupand an established scoringalgorithm
Bombing: Startingfrom scratch and waiting two years for enough actionable data

Whether you measure speed from an OBD device or a smartphone, arecord is created every second. In 60 hours of driving, 216,000records are generated–and that's just for speed monitoring.However, this vast amount of information is useless until it istranslated into human parameters.

That's where actuaries come in. How long will it take them todevelop a scoring algorithm? How will they translate the data intorating parameters? What will they compare data against?

These are formidable questions, unless you have a UBI platformthat already has collected substantial driver data that you can useas your “control” and already has established a scoring algorithmthat you can use as-is or adjust to your specifications.

In the case of our driver Jack, you can present your actuarieswith hundreds of thousands of individual records and let themfigure it out. Or, you can present them with Jack's average drivingscore of 82 out of 100, along with supporting data and control datato compare against. When Jack's driving data is compared to that ofother 23-year old males in California, he can be punished orrewarded via a rate increase or decrease based on his deviation ofnorm.