Mobile Telematics

Looking Back at 2018: The Rise of Mobile Telematics

Softbank’s massive venture investment in Cambridge Mobile Telematics (CMT), one of the largest venture capital deals of 2018, has signaled a growing interest in mobile telematics among enterprises and investors. CMT’s marquee product, DriveWell, uses mobile telematics to identify distracted driving. In addition, CMT has set its sights on behavior-based insurance. TrueMotion, which offers smartphone-based telematics to the insurance industry, added three major insurance carriers as customers in 2018, following an eight-figure deal with Progressive in 2015. Columbus-based Root Insurance was catapulted to unicorn-status last August as well, after having raised over $100M in Series D funding from Tiger Capital and Redpoint Ventures to deliver low-cost, mobile-first, usage-based car insurance. Root’s business relies on mobile telematics and machine learning to understand driver behavior and assess risk.

Silicon Valley is rallying around a mobile telematics-driven future, but what does that really mean? To start, mobile telematics refers to the data that is passively and actively collected through several sensors in your phone and is selectively shared with applications. To understand what this data includes, think about all the transportation applications that you carry around in your pocket: Google Maps, Uber, Gig, etc. Despite each having a distinct purpose, all of these products rely on some form of spatial movement and location tracking. Given that smartphone penetration has hit maturity, we’re seeing more and more companies build enormous businesses around this type of data. But is mobile telematics the right data source to usher us into the golden age? Here are a few relevant considerations:

Hardware: Mobile telematics is centered around four core sensors in your phone: the accelerometer, compass, gyroscope, and most importantly, GPS. In theory, these sensors can be fused together to draw insights about your movement and handling of the phone. As we’ll demonstrate in a later post, the lion’s share of useful information as it pertains to location tracking comes from GPS. That’s not to say that the remaining sensors can’t be useful — in fact, one can still extract powerful information from the remaining sensors. For example, Sentience has built a machine learning model to detect distracted phone use while driving that has achieved as high as 96% accuracy using just accelerometer data. The combination of the accelerometer and gyroscope is ideal for an area like game development because they can sense motion along six axes: up-down, left-right and forward-backward and as well as the rotation of roll, pitch & yaw. However, generating basic linear movement statistics such as velocity and displacement is nearly impossible without GPS due to the inherent sensor error and lack of standardization across devices. This process becomes complicated in places with limited network access or with applications that require tracking granular movements in small spaces (e.g., factory floor management). Normal wear-and-tear and discrepancy across device make/year can have a significant impact on the results as well. Smartphones today rely on several different operating systems and platforms, each of which can change at an accelerated rate as new phones are introduced. Reconciling the discrepancy in software and hardware across models makes collecting data in a consistent manner problematic.

Monitoring: Additional challenges arise from the fact that operators can simply turn off either the application running the GPS tracking on the phone or turn their phone off altogether. Do a quick Google search and you’ll find a myriad of clever hacks on how to disable phone GPS without being detected (here’s an Android example). How does an enterprise monitor this kind of user behavior? In the case of small fleets, identifying anomalous behavior is relatively easy, but at scale it can become costly and require a dedicated workforce.

Data Quality: Many vehicles produced in recent years already harvest a myriad of useful data through on-board connected features. This presents a more useful alternative or complement to mobile-generated telematics. In particular, connected vehicle telematics can accomplish more than just real-time mapping and location tracking. We can expose the pros/cons of solely mobile versus vehicle data with the example of fuel economy. Connected vehicles have access to odometer and fuel consumption readings, which are required measure to fuel efficiency. Because cell phones and tablets are not connected to the vehicle’s on-board computer, they are unable to determine information on how the vehicle was used, including the odometer and fuel sender/gauges. However, this doesn’t mean that mobile telematics aren’t useful. Ultimately, it comes down to the use case. If you’re not looking for much more than workforce tracking or location-based event detection, mobile telematics are probably adequate. On the other hand, if you want to generate more advanced metrics, such as those related to productivity, fuel consumption, compliance, and safety, mobile telematics fall short in dimensionality and granularity.

Privacy: The use of mobile telematics introduces critical privacy issues that technology vendors have yet to fully resolve. The underlying assumption behind mobile telematics is that a user has opted-in to allow companies to collect sensitive location information. But more often than not, users aren’t really aware that it’s happening or, even more dangerously, the software providers aren’t behaving responsibly. While the threats that technology presents to personal privacy are well articulated, there are less obvious implications worth pointing out. For example, usage-based insurance policies can allow for extraordinarily low insurance premiums. This poses an unhealthy choice for consumers: relinquish privacy, or accept higher prices? In a later issue, we’ll dive deeper into the ethics of telematics data.

In general, telematics data could have an extraordinary impact on how vehicles are driven, financed, and optimized in performance. Mobile telematics offer a low-cost, low-fidelity pathway into this future. Interestingly, the players with the greatest traction in this space have centered their efforts around driver behavior. While Google, Uber, etc. mobile telematics to track where you’re going, CMT and others are using this data to learn about how you’re getting there.