Insurance Telematics:
Big Data, Big Potential, Big Headache
Dave Huber, President
Kairos Solutions
IFSUG March 2012
Big Data
One of the few products whose price is set
before costs are known
Known costs
Unknown costs
3
O
Loss adjustment expense
O
Operations
O
Advertising
O
Underwriting
O
Commissions
O
Pure premium (freq x sev)
O
Bodily injury
O
Comp & Collision
O
Regulatory
O
Trends
Known costs
Unknown costs
Premium
Pricing sophistication is a competitive
advantage and depends on data analytics
O
Granularity
O
The number of pricing cells per question or variable
O
Age: 16-19, 20-25, 26-30…vs. 16, 17, 18, 19….
O
Dispersion
O
The range of rates for each of the variables
O
$450-$900 vs. $225-$1375
O
Interactions
O
The lift when combining variables
O
Vehicle symbol & territory – pickups in suburbs
O
Variables
O
New questions and/or external data
O
Credit, occupation, prior limits
Insurers generally use the same data to price
5
Age Gender Marital status Violations Points Homeowner Prior insurance Credit Vehicle 31 M S Speed 4 Own Y 611 YMM 31 M S Speed 4 Own Y 611 YMMThese drivers look like Pure Premium Carbon Copies and are priced identically
But imagine knowing something about drivers
that no one else knows
6
31 M S Speed 4 Own Y 611 YMM 10,651 4.9 31 M S Speed 4 Own Y 611 YMM 13,182 6.1$800
$1200
Age Gender Marital status Violations Points Homeowner Prior insurance Credit VehicleVerified Annual Miles Trips per day
Usage-Based Insurance is all about
segmentation & pricing
O
How, when & where you drive
O
Driving data’s not readily available &
expensive to collect
O
Need a lot of driving data
O
Beyond insurers’ core competency
O
Insurers would really like a driving score
8
The pricing advantage of UBI data is big
O
Granularity
O
The number of pricing cells per question or variable
OAge: 16-19, 20-25, 26-30…vs. 16, 17, 18, 19….
O
Self-reported mileage buckets vs. verified continuous mileage
O
Variables
O
New questions and/or external data
OCredit, occupation, prior limits
O
How, when & where, self-selection, personal driving score akin to a
credit score
O
Interactions
O
The lift when combining variables
O
Vehicle symbol & territory – pickups in suburbs
O
Miles x time of day, frequency & magnitude of speed changes, speed x
traffic
O
Dispersion
O
The range of rates for each of the variables
O$450-$900 vs. $225-$1375
Where does driving data come from?
9
Smartphone
apps
OBD data
loggers
How big is Big Data?
O
Time-stamped trip start/stop, engine on/off
O
OBD - vehicle speed every second
O
GPS - lat, long & heading every second
O
Accelerometer – 3 axis acceleration
10
O
5,000 GPS-enabled devices
O
8MM journeys & 15B journey points
O
20 million new rows of data daily
How might all this Big Data show up?
11
Annual mileage Avg trip duration Avg trip length Trips per day
Trips per time of day Journeys
Miles by time of day Miles by day of week Weekdays
Weekends
Miles in speed bands Time in speed bands Average speed Trip regularity (miles) Trip regularity (time) Aggressive acceleration
per 100 miles
Aggressive braking per 100 miles
Road type Relative speed
Miles in territory Drive time in territory Idle time in territory Cornering Lateral acceleration Rolling stops Self-selection Lane changes Acceleration events in speed bands
Braking events in speed bands Frequency of speed changes Magnitude of speed changes Commuter profile Errand-runner profile Coffee drinkers YMM relativities OnStar subscription Cruise control Driver score Driver “footprint” Left turns Speed variation
Trip type (speed vs time) Territory by time of day Holiday driving
School zone
Violations by trip type Trip radius Student profile Intersections Turn signal Seat belt Lights / wipers Vehicle maintenance Time between trips/journeys Congestion index Summer car