WHITE PAPER. Preventing Collisions and Reducing Fleet Costs While Using the Zendrive Dashboard

Transcription

1 WHITE PAPER Preventing Collisions and Reducing Fleet Costs While Using the Zendrive Dashboard August 2017

2 Introduction The term accident, even in a collision sense, often has the connotation of being an event that happens by chance or without apparent or deliberate cause. At Zendrive, we believe that most collisions simply aren t chance occurrences. After analyzing more than 15 billion miles of driving behavior, we can measure risky driving behavior that causes collisions, like phone use while driving, speeding, hard breaks and rapid acceleration. In fact, 90% of collisions are caused by measurable human error. What s worse, is that an analysis of 3 million drivers and 5.4 billion miles of data reveals that 1 the worst quartile of commercial drivers account for roughly half of all commercial collisions. According to the National Safety Council, the average cost of a car collision was $7,500 2 in Since the riskiest drivers cause most of these collisions, these drivers are also the most expensive from a social and insurance perspective. As a mission-driven company working to improve road safety through data and analytics, we can make the greatest impact by focusing on risky drivers to improve their driving behavior. The Zendrive Dashboard helps fleets do just that. The Zendrive platform measures driving behavior through sensors on drivers smartphones. This data is fed to the Zendrive Dashboard, and categorizes drivers into Excellent, Fair and Risky segments. The fleet manager can see both aggregate and individual performance, and can message drivers with feedback and coaching. This gives the Fleet Manager the control to reduce the overall insurance risk of the fleet. Zendrive conducted the Preventing Collisions and Reducing Fleet Costs While Using the Zendrive Dashboard report to assess the validity of the Zendrive platform, particularly the Zendrive Dashboard, as a tool to reduce insurance risk and improve driving safety. This study examined the driving behavior of 8 fleets and 21,299 drivers that have used the Zendrive Dashboard for Loss Control purposes. In particular, we highlighted the riskiest of drivers those most likely to cause a collision and examined the improvements in their driving behavior after being coached through the Zendrive Platform. 1 Commercial drivers include ridesharing, on-demand delivery companies and others 2 This figure represents the average cost of a car collision (of types non disabling minor injury and physical damage to the vehicle costs) to a company/business. It includes wage and productivity losses, medical expenses, administrative expenses, uninsured costs, and repair costs to vehicles 1

3 Content I. Executive Summary II. How Zendrive Measures and Improves Driving Safety III. Targeting Risky Drivers Targets 53% of Commercial Collisions IV. Driver Safety Improved with the Zendrive Dashboard V. Estimated Cost Savings from the Zendrive Dashboard VI. Improvement in Overall Fleet Safety Over Time VII. Conclusion VIII. Appendix A. Data Overview B. Methodology 2

4 Executive Summary Backed by a dataset of over 15 billion miles and over 5 million active users, Zendrive has determined that the worst 25% of commercial drivers cause over 50% of commercial collisions. Using the Zendrive Dashboard to identify and coach them can improve their driving, prevent collisions, and reduce costs. In this study, we looked at a commercial auto driving behavior dataset of 8 fleets and 21,299 drivers, all using Zendrive. The Zendrive Dashboard lets fleets target their riskiest drivers for coaching. On average, the worst drivers reduced their likelihood that they would get into a collision by 49% over the first 250 days. From the perspective of a fleet manager, this translates into $874K to $2M of estimated reduction in losses for Bodily Injury and Property Damage Liability per 1000 risky drivers each year, depending on the type of vehicle, business and limit in plan. The bottom 25% of commercial drivers account for 53% of commercial collisions. These are called Risky drivers Fleets using the Zendrive Dashboard to control risk saw a 11.4% improvement in their riskiest drivers over the course of the first 250 days, as measured by the Zendrive Score, a comprehensive measurement of unsafe driving behaviors including distracted driving, swerving, and more This would reduce the likelihood of these drivers getting into a collision by 49% - nearly half - during those first 250 days, as calculated based on industry standards using commercial auto frequency relatives This safety boost translates into $874K to $2M of estimated reduction in losses for Bodily Injury and Property Damage Liability per 1000 risky drivers each year, 3 depending on the type of vehicle, business and limit in plan as shown below : $100K Limit $1M Limit Taxi $1,248,832 $2,123,014 Ride Sharing $874,182 $1,486,110 Safety improvements were sustained over the course of the study, and are expected to persist over the driver s lifetime 3 See additional details and notes to calculation in Table 4 of this report 3

5 How Zendrive Measures and Improves Driving Safety Zendrive technology detects vehicle trips and safety-related driving events using smartphone sensors like the GPS, gyroscope, and accelerometer. Unlike traditional hardware-based solutions, this technique measures the driver s behavior rather than the car s a technique better attuned to measure driver error, the number one cause of collisions. This data shows aggressive acceleration, hard braking, excessive speeding, and risky phone use, and combines them with risk-related data about the time of day, the type of road, the mode of transport, and any detected collisions. This information is used to construct three metrics Focus, Caution and Control which are combined to generate a single Zendrive Score, an overall measurement of safe driving behavior. The Zendrive Dashboard allows the Fleet Manager to view driving behavior details for individual drivers as well as fleet-wide safety assessments over time. It categorizes drivers into Excellent, Fair and Risky groups and allows the Fleet Managers to message all drivers in each category. This lets them recognize great drivers, and coach drivers who need to improve. The top 15% of drivers fall typically into the Excellent category, the bottom 25% typically fall into the Risky category, and all other drivers fall into the Fair category. After targeting drivers for safety coaching through the Zendrive Dashboard, the Fleet Manager can review progress by examining each individual driver s recent trips, as well as monitoring driver improvement over time. 4

6 Focusing on Risky drivers targets 53% of commercial collisions According to a dataset of 41,529 commercial drivers covering 233 million miles, we find 4 that the worst 25% of drivers account for 53% of commercial collisions. The worse the driver, the higher their relative percent of collisions. (See Chart 1). Therefore, targeting the bottom drivers means maximizing the effects of coaching to prevent future collisions. These statistics were further validated by our much larger dataset on our non-fleet consumer drivers. In a sample of 3 million drivers and 5.4 billion miles, we find that the 5 worst quartile of drivers account for roughly half of all collisions. 4 See methodology section for details on calculation 5 See methodology section for details on calculation 5

7 The Risky category of drivers as identified by the Zendrive Dashboard was designed to target the bottom 25% of commercial drivers, who account for more than half of commercial collisions. Driver Safety Improved with the Zendrive Dashboard Our analysis started by identifying drivers who were flagged at least once in the Risky category for commercial fleets that use the Zendrive Dashboard. Out of 21,299 drivers, Zendrive identified a group of 838 drivers that were flagged at least once in the Risky Category and had at minimum 14 days of driving activity. We identified the date when 6 these drivers were first flagged, along with the Cumulative Zendrive Score ( Zendrive Score ) of each driver on that date. We then followed these drivers on the application over their time up to date of study. We found that after receiving coaching through the Zendrive Dashboard, risky drivers 7 saw consistent improvements. (See Chart 2 ). Risky drivers started off with an Average Zendrive Score of 64 and saw their score improve to 71.3 over a 250 day period. 6 See methodology section for definition 7 The x-axis represents the number of calendar days after the driver was first flagged as Risky and the y-axis represents the Average Zendrive Score for that day. 6

8 Improvements in driving behavior were sustained over the driver s days of coaching and monitoring by Zendrive, although the trend was less typically conclusive with time due to a declining sample. On contrast, risky commercial drivers whose fleet managers did not use the Zendrive dashboard did not see any improvements in their driving behavior during the same time period. A group of over 50,000 risky commercial drivers started with an average score of 66 saw their score actually decrease slightly to 63 over a 250 day period. 7

9 Using the frequency relativities calculated on our commercial auto dataset of 41,529 commercial drivers (see Chart 1) and improvement in Zendrive Score over time while using the Zendrive Dashboard (see Chart 2), we were able to calculate the average change in the likelihood of getting into a collision. During the first 250 days, we found that drivers in the Risky category reduced their likelihood that they would get into a 8 collision by 49%. (See Chart 3). 8 See methodology section for details on calculation 8

10 Estimated Cost Savings by using the Zendrive Dashboard A reduction in the probability of collisions translates into direct costs savings from both the collisions themselves and insurance costs. Not only that, but drivers who are no longer getting into collisions can stay on the road longer, improving supply availability and maintaining the opportunity for revenue. Fleet Managers using the Zendrive Dashboard could see estimated loss reductions between $874K to $2M per 1000 risky drivers per year, depending on the vehicle, business and limit of plan. (See Table 4). 9

11 Improvement in Overall Fleet Safety Over Time To account for the possibility that the driver level results could be skewed by one large fleet with positive results, we also calculated the Risky drivers improvement on the Fleet Level. Additionally, we wanted to ensure that driving behavior improvements hold true for the days the driver actually uses the application after the driver was first flagged as Risky rather than the calendar days after being first flagged as such. We found that improvements in driving behavior over days driven on the application are 9 very comparable to the improvements over a calendar day period. (See Chart 5 ). The maximum improvement over a 220 days driven on the application (which translated to roughly a 400 day calendar period) was an 11.4% increase in Average Zendrive Score. There was no sizable differences between the simple and average trend over time. For instance the average driver improved their Zendrive Score by 10.81% during the 100th-109th day period on application, after initially being flagged as Risky. The average fleet improved by 10.15% during the same time period. 9 The x-axis shows the group of days the driver was on the application. The y-axis is the Average Growth in the Zendrive Score from when the driver was first flagged as Risky. The blue line in Chart 6 below represents the weighted average, which is comparable to the driver level line as shown in Chart 5 (the weight being the number of drivers in each fleet). The gray line is a simple average of all 4 fleets that fit the data criteria -- as outlined in notes 1 and 2 of the data section of Chart 3. 10

12 10 Chart 6 below highlights the average improvement in Zendrive Score for four fleets 11 that fit the minimum viable data criteria. (See Chart 6). All fleets improve consistently over the driver s days while using Zendrive, with the largest improvement seen in the most active users of the Zendrive Platform. For instance, Fleet #1 (See Chart 6) utilized 10 This chart follows each fleet s average improvement in the Zendrive Score for the group of Risky drivers, since they were first identified as Risky over those driver s lifetimes on the application. Similar to Chart 6, the x-axis in this table represents the group of days the driver was on the application and the y-axis represents the Growth in the Average Zendrive Score (from when the driver was first flagged as Risky ). 11 See methodology section and notes of charts for details on minimum viable data criteria 11

13 the Zendrive technology for both virtual and weekly in-person coaching of drivers and saw the largest improvement in their Zendrive Score of 15.39% over the first days. The Zendrive platform was beneficial for all coaching types observed but the best results are as expected for the most engaged fleets. Overview of Fleets included in calculation Fleet #1: O n-demand parking and car services mobile application platform Fleet #2: Ride service company designed to transport children Fleet #3: Online marketplace for personal drivers on-demand to drive the vehicles of their members Fleet #4: Ridesharing company 12

14 Conclusion Zendrive is working with communities, local decision-makers, advocates and driving coaches to use our data to save lives. Traffic deaths are preventable when we understand the behaviors and factors that contribute to collisions. The most impactful step is to take is to coach the lowest quartile of drivers, who account for over half of all collisions, to practice safer driving behaviors. Until now, Fleet Managers had no systematic and accurate way to control this risk by identifying these drivers. With the help of the Zendrive platform, and tools such as the Zendrive Dashboard, Fleet Managers now have the power to help improve their drivers behaviors, reducing their insurance risks and costs, as well as protecting the reputation of the company. Risky coached drivers expected likelihood of being in a collision dropped by 49%, and their safety improvements persisted through the course of the study. This can result in $874K to $2M of estimated reduction in losses for Bodily Injury and Property Damage Liability per 1000 risky drivers each year, depending on the type of vehicle, business and limit in plan. Zendrive s mission is to eliminate road deaths. Backed by one of the largest overall dataset in the industry of 15 billion miles and over 5 million active users, Zendrive is working hard to continue to improve our algorithms and coaching techniques to accurately measure and improve driving behavior. 13

15 Appendix Data Overview Driver behavior data set for study: A sample of 8 fleets and 21,299 drivers that use the Zendrive Dashboard for Loss Control purposes Time period for study: Over the lifetime of the drivers on our platform Total Zendrive driver behavior data set: 15 billion miles Total number of Zendrive users: 5 million drivers Methodology Data collection: Zendrive technology detects vehicle trips and safety-related driving events using smartphone sensors like the GPS, gyroscope, and accelerometer. This data is analyzed to show risky driver behaviors, including aggressive acceleration, hard braking, excessive speeding, and risky phone use, and combines them with risk-related data about the time of day, the type of road, the mode of transport, and any detected collisions. Zendrive Score calculation: Zendrive technology aggregates the information it collects to construct three metrics Focus, Caution and Control which are combined to generate a single Zendrive Score. Focus measures the driver s attention level based on phone usage; Caution measure s the driver s tendency to follow the rule of the road such as overspeeding; and Control measures the driver s tendency to accelerate of brake beyond normal limits. The Zendrive Score is calculated as a function of these three scores. Cumulative Zendrive Score calculation: The main measure of driving behavior referenced in this report is the Cumulative Zendrive Score, which is typically computed over a few weeks period depending on how much the individual driver drove during that period. This normalizes the variance in daily driving activity and identifies the systematic tendencies of that driver s behavior. The Zendrive Score referenced in this study is the Cumulative Zendrive Score. Selection of fleets in study: Fleet customer who use the Zendrive Dashboard and have the minimum amount of viable data were included in this study. Minimum amount of viable data is defined as having at least 25 daily active drivers that drove a minimum of 14 days on the application. 14

16 Calculation of quantiles based on Cumulative Zendrive Score: Quantiles were calculated by sorting the drivers by their Zendrive scores, dividing them into twenty equal groups, and then determining what percentage of total accidents each group was responsible for. Relative frequencies were then calculated for each quantile as the the collisions that group accounted for divided by the total number of collision for the sample. Quantiles were not adjusted to be on a per mile basis. The best group is referenced in the top quantile and the worst group was reference in the bottom quantile. For instance, the worst 25% of drivers account for the majority, 53%, of all commercial collisions statistic referenced in this report was determined as the percentage of total accidents the bottom five groups accounted for. Calculation of likelihood of collision statistics: The likelihood of a collision for a certain driver was computed in this study as the relative frequency of collisions of the quantiles that that average score for each driver fell into, with quantiles broken up by 20th s. The likelihood of a collision for a certain score was computed as the relative frequency of collisions of the quantiles that that average score for each driver-day fell into, with quantile broken up by 20th s. Average reduction in likelihood of collision was computed as the average of the difference in the driver day frequencies for driver scores over time. 15