Drive Cab Data Findings

Presentation on DriveCam Analytics and Data Findings Related to In-Cab Driver Distraction

©2010 DriveCam

Bryon Cook, DriveCam

My Background
• DriveCam:
– – – – Joined DriveCam in January, 2006 (4.5 Year Tenure) VP, Product Management and Analytics for 2 years VP, Operations and Analytics for 2.5 years Developed first three iterations of intelligent triaging algorithms as well as multiple business intelligence studies for internal and external uses

• Prior Work:
– 10 Years in Software and Service Development Prior to DriveCam (15 Years Technology / Service Experience Overall)
• • • 3 Years in ERP, CRM, Project Management Applications with Business Intelligence Tools On Each System (Dataworks – Now Epicor, Equative) 2 Years in
Workers Compensation and Automobile Medical Claims and Bill Review Systems with Business Intelligence Tools on Each System (HNC Software: Insurance Solutions Group – Now Fair Isaac). 5 Years in Automobile Medical and Physical Damage Claims and Bill Review Systems with Business Intelligence Tools on Each System (5 Years with Mitchell International)

– Researched, designed and developed 27 mission-critical B2B products over the past 15 years encompassing over 200 functional releases and 500 data updates in a multitude of environments including financial, insurance, and IT sectors.
• Education:
– BSBA / MBA: Washington University in St. Louis
©2010 DriveCam

Agenda
• • • • • • • Who is DriveCam? How Does the Technology and Service Work? Why is Video Important? Employee Acceptance What data is collected and how is it collected? Analytics Strategy and Data Sets Distracted Driving
– – – – What is Distracted Driving? Distracted Driving Behaviors Captured Distracted Driving Analytics / Results Impact of Coaching Behaviors

• Summary / Q&A
3
©2010 DriveCam

Who is
DriveCam?

©2010 DriveCam

Bryon Cook, DriveCam

Why does DriveCam exist?
Historically … Fleets have improved risky driving through: • Training • DMV Record Checks • Safety Meetings • Ride Alongs However … • Over 43,000 people continue to lose their lives in vehicle crashes every year • $245 billion are wasted annually • Traditional fleet safety efforts rely on lagging indicators • By the time you know there’s a problem … it’s too late
5
©2010 DriveCam

In Vehicle Video is Knowing the Truth

Open Directory

Following too Close

Drowsy Driving

Distracted Driving
6
©2010 DriveCam

Cut Off in Traffic

DriveCam FactSheet
• Headquarters: San Diego, USA • Founded: 1998
– Privately held – 10 years in business with significant, consistent growth – Funded by 4 leading capital funding firms: Insight Venture Partners, Integral Capital Partners, JMI Equity and Menlo Ventures

• Milestones
– – – – – Installed Base: 135,000 (55,000 Under Managed Service) Risky Event Database: 17.7 million events Collisions Captured: 18,500 Minor to Severe Collisions 3 years on INC’s list of 500 fastest growing, private U.S. companies #30 on Wall Street Journal’s List of Top 50 Venture Backed Companies

• Best of Breed Technology & Service Partners
7
©2010 DriveCam

Major Players in Major Industries

8

©2010 DriveCam

How Does the Technology and Service Work?

©2010 DriveCam

Bryon Cook, DriveCam

How the Video is Captured
•

Video Event Recorder • Mounts on windshield • VER Continuously Monitoring and Providing Feedback to Driver • Events Triggered by excessive forces • Erratic driving • Collisions
•

•

•

Digital looping memory captures before and after event Documents force and records video and audio looking forward and inside the vehicle Downloads to DriveCam via Cell

©2010 DriveCam

How DriveCam Works
Upload triggered event via Cell

2

1
Capture risky driving

3
Review, analyze and score
Driver Risk Management Program

6
Safer driver returns to the field

4
Access confidential website for events, dashboards and reports

5
11
©2010 DriveCam

Coach driver

DriveCam Online
Connect to DriveCam Online

12

©2010 DriveCam

Why is Video Important?

©2010 DriveCam

Bryon Cook, DriveCam

Can you predict without video validation?
Below are four ‘Hard Braking’ events as determined solely by the vehicle information and accelerometer data. All of these would be considered ‘Risky’ from a black-box perspective and considered correlative and possibly predictive of risky driving behaviors. Which ones do you feel are the most predictive of risky driving? Which one is the most severe? What behaviors did these drivers exhibit? How would you correct that drivers behavior?

A B C D
©2010 DriveCam

Can you predict without video validation?
A
Many of our customers do BRAKE TESTS EVERY Morning and Before shifts as part of typical vehicle maintenance and testing. These come through as severe events, but are actually good safety checks and NOT considered risky driving.
©2010 DriveCam

Can you predict without video validation?
B
Many of our commercial vehicles make FREQUENT STOPS (eg. Buses, waste trucks, distribution vehicles). While the driver stopped a little abruptly, this event is considered ‘wear and tear’ on the vehicle, but it is NOT considered risky driving.
©2010 DriveCam

Can you predict without video validation?
C
The driver in this event was only maintaining a 1.25 second FOLLOWING DISTANCE. This is not enough time to see a problem and smoothly respond to it. The driver had to react abruptly to avoid a collision. This IS considered a risky driving event.

©2010 DriveCam

Can you predict without video validation?
D
In many cases, we find that a driver could have done better DEFENSIVE DRIVING. In this case, the driver could not have anticipated the car from the left lane trying to cut in front of the bus. The driver showed tremendous reaction time to avoid an accident. This is NOT a risky driving event.
©2010 DriveCam

Can you predict without video validation?
Only one of the four events would have been considered ‘RISKY DRIVING’. The other three events are ‘Wear and Tear’ on the vehicle, but not considered risky driving.
Brake Test

A
Frequent Stops – Wear and Tear

B
Following Too Close

C
Good Defensive Driving

D
©2010 DriveCam

Why would an accelerometer detect high g’s?
There are many reasons why a device might trigger:
– – – – – – – – – – – – Pot holes Unpaved roads Rail road tracks Turning hard in a large vehicle Rocking an unloaded tractor-trailer Waste truck throwing a trash bin into the vehicle hard Jack-rabbit start Vehicle Maintenance Defensive Driving / Evasive Maneuver Risky driving maneuver Collision etc.

©2010 DriveCam

Combination of Science and Humans is Required to Identify Risk!!!
• Driver risk management systems require audio / video and human review to verify:
– – – – – Why the device triggered Who triggered the device How to improve on future triggering algorithms Identify risky behaviors so they can be corrected Provide a cost effective solution

•

Our event analysis consists of three key elements
– Trigger Algorithms (Accelerometer Based Algorithms)
• • • System monitors 24x7 and looks at G-Force data continuously. Multi-Axis Accelerometer Systems to Identify High g-Force Events. System filters out obvious road-noise and identifies potentially risky maneuvers (Accelerations, Braking, Cornering, Collisions, etc).

–

Wave Form Analysis (Decision Tree Algorithms)
• System evaluates wave-forms to detect what type of event occurred and filters out obvious noise and nondriving events (eg. Maintenance events).

–

Human Review (Human Analysis)
• • Evaluates remaining events to separate out high g-force noise events from risky events / collisions. Isolates root behaviors and causes of risky events.

•

Overall, this three step process ensures that we capture the risky clips and collisions while providing a cost-effective solution for our customers.
©2010 DriveCam

Employee Acceptance
How do we deal with employee concerns?

©2010 DriveCam

Bryon Cook, DriveCam

Employee Acceptance
It’s natural for employees to be uncomfortable with DriveCam coming to their worksite…

• Most concerns are based on misconception, misinformation or lack of experience with the program
Left unaddressed, these concerns can fester into employee resistance that can impede program success…

• Open, early communication is the key to gaining employee success
DriveCam has the experience and the tools to help you successfully introduce the program to your employees…

• Orientation video, guide book on how to introduce the program, driver & supervisor FAQ handouts, onsite assistance
©2010 DriveCam

Driver Acceptance
Common Concern
It’s Big Brother

Details
Manager’s can look in live or can activate it remotely. Technology does not have this capability. On average less than 10 seconds of video are captured per day per vehicle. Website is secure. Only those with proper permissions can access. Coaching will take place in a private place

Video will be used to embarrass me

The video is just an excuse to get me There are no new disciplinary policies. Program intention is to improve safety. fired
Video often exonerates driver.

It’s an illegal invasion of privacy

Employer does have the right to monitor the workplace. Video cannot be covert and without employee knowledge

©2010 DriveCam

Driver Acceptance
• Best Practices:
– Communicate the program early to all stake holders – Emphasize benefits (protects drivers, improves safety) – Include DriveCam in management vehicles – Don’t forget positive reinforcement! – After deployment, communicate, communicate, communicate!
©2010 DriveCam

What Data is Collected and How is it Collected?

©2010 DriveCam

Bryon Cook, DriveCam

Event Review Process
• • A certified DRA reviews and analyzes each event for potential risk Once meaningful behaviors are identified they are validated using the DriveCam behavior definitions and rules Finally the event’s status is set to Face-to-Face Coaching when the selected behaviors meet certain thresholds of concern
– Collisions and extremely concerning events are escalated to the Client for prompt notification

•

27

©2010 DriveCam

Sample Guidelines for DRAs
Distractions While distractions are common, it is usually when they contribute to an event trigger or significantly increase risk that we score them.
Activities such as checking a mirror or scanning to the side or glances in the vehicle such as looking at the instrument panel for 1 second or less are normal driving activities and should not be considered as distractions. There are 3 situations in which a distraction should be selected:
1. 2. 3. If the behavior is noticed during the event. If the distraction was the cause of the event being triggered. If the driver reaction is within 1 second of the risky situation, this is not considered to be the cause of the trigger. If the duration of the distraction is exposing the driver to substantially increased risk. * Guidelines to aid in qualifying as a distraction exposing the driver to substantially increased risk are: eyes are off the road while moving for 2.5 consecutive seconds, 4 or more seconds in the first or last half of the clip or a total of 6 seconds or more throughout the
clip.

Cell Phone –Anytime a driver is using a hand-held or hands-free cell phone select this behavior. Under all circumstances, cell phone should be marked as risky when the use of the cell phone and the associated inattention led to an abrupt action or clearly increased risk. Follow the same guidelines listed in “Root Cause – Distractions” above. General guidance to determine if the use of the cell phone is risky.
• • Hand-held: If driver is holding the cell phone up to the ear or they are actively using the phone/speakerphone (texting, dialing, obviously in a phone conversation). Hands-free: If you can see the driver wearing the device and they appear to be in a phone conversation.

Select Cell Phone Hands-free if you cannot see the device yet the driver is obviously in a phone conversation (clearly responding to a caller). Indications of this can be: the driver has conversational pauses and/or verbal ©2010 DriveCam acknowledgements such as uh-huh,hmm, yes, no, etc.

Human Review
• Certified Analysts Review All of the Events that Pass the Wave Form Analysis
– New Analysts must pass a certification test to review driving events – After they pass the test, there is a probationary period where they review clips with a certified analyst – Once the analyst has shown proficiency, they will become certified – Over 30% of our initial hires do not pass the screening process as we have strict guidelines that all clips are reviewed to our standards

• Analysts will review the event and identify the trigger, root cause, and risky behaviors within the event
– Reviewing for commonly accepted safe driving practices based on Smith Systems

©2010 DriveCam

Analytics Strategy / Data Sets

©2010 DriveCam

Bryon Cook, DriveCam

Analytics Strategy
Classifying Risk

• Trigger Algorithms • Wave Form Analysis • Additional Sensors (Speed, GPS, Vehicle Information) • Neural Networks for Vehicle Specific Information • Human Review

Profile
• Profile Non-Collision vs Collision Drivers Against Leading Risk Indicators (Frequency & Severity) and Claims • Report on Drivers that Meet the Profiles

• Predictive Models to Identify Underlying Driving Patterns and Changes in Driving • Predict Collisions Based on RealTime Driver Performance Against the Profiles (Which Driver?, When?, Probable Cause?, Collision Severity?, Estimated Claims?)

Identify

Predict

©2010 DriveCam

Analytics Strategy: Key Partners Academic Government Commercial Insurance
©2010 DriveCam

• VTTI, Univ. of Iowa, Univ. of Montana’s Western Transportation institute, Univ. of Michigan, Univ. of California – San Diego

• FMCSA, CVSA

• Major auto manufacturers, Analytics experts, suppliers and fleet operators (confidential)

• Major industry service providers, brokers and carriers (confidential)

Analytics Strategy: Industry Associations
– – – – – – – – – – – American Trucking Association (ATA) – *Recent Study Being Handed Out (Fatigue) Truckload Carriers Association (TCA) National Private Truck Council (NPTC) National Tank Truck Carriers (NTTC) American Public Transit Association (APTA) American Gas Association (AGA) Edison Electric Institute (EEI) Environmental Industry Association (EIA) National Ready Mix Concrete Association (NRMCA) National Retail Federation (NRF) American Society of Safety Engineers (ASSE)

©2010 DriveCam

Data Set
• • • Triggered Events: 27.7 Million (10 Million VER Reviewed Events) Human Reviewed Driving Events: 17.7 Million (15.5 Million Behaviors) Distracted Driving Events:
– Breakdown:
Behavior Records Cell Phone - Handheld 37,142 Cell Phone - Handheld [Risky] 159,417 Cell Phone - Hands Free 22,709 Cell Phone - Hands Free [Risky] 3,510 Electronic Device 2,701 Food/Drink 18,379 Food/Drink [Risky] 21,304 Other 139,150 Other Communication Device 16,053 Other Communication Device [Risky] 1,584 Passenger 3,766 TOTALS 425,715

•

Growth
– – Driving Events: 425,000 Events Each Month (5.1 Million Annual Run Rate at Current Base) Distracted Driving Behaviors: 14,500 Events Each Month (174,000 Annual Run Rate at Current Base)

©2010 DriveCam

Distracted Driving

©2010 DriveCam

Bryon Cook, DriveCam

What is Distracted Driving?
• Definition: Diversion of attention away from activities critical for safe driving toward a competing activity
1

• Comes in 3 different forms:
– Visual: Eyes off the road – Physical: Hands off the wheel – Cognitive: Mind off of driving

• Distracted driving has always been an issue but mobile technologies have upped the ante

1 Regan,Lee & Young 2008
©2010 DriveCam

Distracted Driving
• Drivers texting take their eyes off the road nearly 5 seconds • At 60 miles per hour a vehicle has traveled about 147 Yards (1.5 Football Fields)

©2010 DriveCam

The Myth of Multi- Tasking
• Brain does not perform 2 tasks at the same time – it switches from one to the other • When overloaded, brain selects what to pay attention to – incomplete information • And reaction time slows due to switching
©2010 DriveCam

Distracted Driving Examples

Cell Phone – Roll Over

Texting- Almost Head on

Other Distraction: Lipstick - Spinout

©2010 DriveCam

Hands Free – Traffic Violation

Food / Drink – Near Collision

Distracted Driving –What We’ve Learned
Top Behaviors
35% 30% 25% 20% 15% 10% 5% 0%
Following Too Not Looking Far Traffic Distractions** Failed to Keep Close Enough Ahead Violations* Out Other

Last 365 Days
31% 24% 20% 11% 4%

10%

©2010 DriveCam

DriveCam Analytics
Distraction Type Cell Phone - Handheld Cell Phone - Hands Free Electronic Device Food / Drink Other Other Communication Device Passenger ALL Distractions % Identified in All Risky Driving Events 2.9% 1.4% 0.3% 1.0% 5.4% 0.8% 0.1% 11.9%*

* Slightly Higher then Previous Chart because an event can have multiple behaviors. ** Driver Condition: Fatigue/Drowsy is 0.1% of Behaviors and a contributing factor in collisions 2.13% of the time in trucking (see ATA study).

©2010 DriveCam

DriveCam Analytics
Single/Multi-Collision Drivers were “x” times more likely than non-collision drivers to . . .
Not Checking Mirrors Not Scanning the Roadway Failed to Keep an Out Ran Stop Light Following Too Close 1 sec or less Cell Handheld Other Unsafe/Risky Violation Other Distraction Food/Drink Following Too Close 3 to 3.75 sec Ran Stop Sign Not Scanning the Intersection Following Too Close 1.25 to 1.75 sec Cell Handsfree Rolling Stop Not Looking Far Enough Ahead Following Too Close 2 to 2.75 sec 0.0 1.0 1.5 1.5 1.5 1.6 1.6 2.0 2.8 2.7 2.6 3.0 4.0 5.0 6.0 7.0 8.0 1.2 2.3 3.1 2.9 1.5 1.6 3.5 3.5 1.5 1.5 1.8 1.9 2.0 4.1 3.9 4.4 4.8 4.6 5.2 1.8 2.7 5.6 3.4 7.0 6.9 7.4

Multiple Collision Drivers

Single Collision Drivers

©2010 DriveCam

DriveCam Analytics
• Events triggered by abrupt action – our data is a subset of total # of cell use occurrences • Drivers identified with 3 or more cell phone events twice as likely to have a collision • Drivers identified with 6 or more cell phone events 3 times as likely to have a collision & 6 times as likely if prior events were deemed risky
©2010 DriveCam

In-Cab Video Impact
Frequency Drop in Distracted Driving - 1st Coach Date
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

Construction / Concrete: 88% Reduction Distribution: 87% Reduction Energy: 96% Reduction

Impact of Identifying and Coaching Drivers on Distracted Driving Behaviors.

Construction/Concrete

Distribution

Energy

91% Reduction Overall

©2010 DriveCam

Distracted Driving
• Many rear-end distracted driving crashes also include poor following distance • With little distance ahead, the distracted driver has no margin for error when things change ahead • Proper following distance increases the margin for safety…
Rear ender 28.DCE

©2010 DriveCam

Distracted Driving
Following Distance and Rear Ending Others
50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0%

47% 17%
< 2 Seconds 2+ Seconds Following Distance

36%

Veh Ahead Stopped

• Proper following distance = fewer distracted driving incidents
©2010 DriveCam

Summary

©2010 DriveCam

Bryon Cook, DriveCam

Summary
• Video-based solutions have tremendous advantages in identifying and eliminating distracted driving
– – – Is the only way to identify “why” (the root cause) of a an event– and so leads to superior analytics It is a powerful tool for changing driver behavior ( a picture is worth a thousand words) As an exception based system, it is still cost effective

•

Distracted Driving is highly correlated to collisions – so it is VERY dangerous
– – Note that other risky driving behaviors are more frequently identified. In particular, in combination with other risky behaviors is a recipe for a severe collision (following distance, not looking far enough ahead, etc

•

We applaud the efforts of the FMCSA and the MCSAC to address ways to make our roads safer.
– – It is also our passion! We are here to help!. Thank you for your invitation today

©2010 DriveCam

Q&A

©2010 DriveCam

Bryon Cook, DriveCam