Perspective on car data monetization
Car data monetization opportunities will grow incrementally for industry players along the mobility value chain as car data is likely to create new value leading to increased revenues, reduced mobility cost and improved safety.
What would you like your car to do for you?
Car data can be monetized in 4 ways:
- Sell car related services/features to consumers.
- Sell consumer insights to third parties.
- Reducing risks and operational costs due to more and better component usage knowledge.
- Additional safety and security benefits, e.g. improving safety response to a car crash, increasing road patrolling effectiveness.
It is clear that lots of organizations adjacent to automotive can seize these opportunities.
You would not be the first to explore possible gains from car data as major players are currently involved in a number of projects.
Some players are exploring software platforms and big data analytics services. A number of projects, mostly by service and mobility providers (such as Uber) and start-ups, focus on very specific features or use cases for car application to reach a large number of customers with the same application and value proposition.
However, up until now car manufacturers are mainly pursuing product optimization from ‘actual usage’ data, but there are much more upcoming use cases (see picture below).
In recent years, our dependence on road infrastructure has grown at a phenomenal rate.
While it offers immense advantages, this evolution has some major downsides as well:
- Permanent congestion on highways and urban centers.
- Energy waste.
- CO2 emissions with consequent impact on public health.
- High rates of accidents on the road networks.
It is clear that people refuse to give up mobility. So which improvements are possible to make our mobility smart and sustainable, and improve our lives?
Companies and users seem to be interested in augmented road weather information, especially when it comes to the safety of the users/drivers to warn the drivers for dangerous situations on their route. These could be traffic jams, road works but also changing weather situations.
Let's look at one use case related to real-time warning systems for dangerous road conditions.
- What are the hurdles today?
- What challenges do we face?
- Machine learning required
- So what can we conlude?
1. What are the hurdles today?
Advanced driver assistance technologies such as driver alertness and pedestrian detection are already being made available as standard options in vehicles.
However, there is still a need for improvement in various other areas, one of them is road weather conditions. Factors such as heavy rain, snow, hail, icy roads, strong winds and limited visibility are challenging drivers. The existing solutions for road weather services are limited in their scope and scalability. Furthermore they are technology dependent, work off-line and offer long delays. As a result, weather uncertainties and environmental conditions still pose a significant risk for driver safety.
To do so, employing and expanding the recent advances in smart vehicle technology is a must to improve road safety in the context of dangerous weather conditions. Many meteorological institutes and its stakeholders are interested in detailed analysis of local weather phenomena such as:
- Reduced visibility and fog.
- Heavy precipitation and hail.
- Dangerous conditions such as slippery roads due to frost, snow, freezing rain.
- Local temperature variations.
- Wind gusts (cross winds in particular).
The aim is to better understand the vehicle under various weather conditions that can lead to further enhancement of vehicle-specific safety recommendations integrated in real-time services like weather information for drivers, but it could also lead to broader applications.
2. What challenges do we face?
To address this need, it is required to create a complete integrated solution that benefits from a real-time analysis of the weather data gathered from various underlying technologies and provides an on-time appropriate reaction to the end user (see picture below).
This involves the integration of:
- Higher level of intelligence into the sensing and communication infrastructures.
- Decentralized aggregation and decision-making.
- Real-time predictive algorithms to provide accurate and timely warnings.
To make the innovations in weather data collection and advanced modeling useful in the field, a challenge will be to assert that the warnings are correct and delivered in a timely manner by determining the relevance of the data (f.e. if contradictory measurements are made by multiple vehicles), as well as assessing the relevance of data points in time (f.e. the weather conditions might have changed).
Large amounts of data can be captured from cars such as:
- Mechanical information: throttle, brake, steering position, suspension positions, wheel speeds, GPS, linear and angular acceleration, selected gear, engine RPM, engine torque and ABS/ESP.
- Environmental information: exterior/interior temperature, heating settings, engine temperature, windshield wiper frequency, lights on/off or setting info and derived or ‘virtual’ quantities from mechanical information.
Many of these parameters are only meaningful when captured at a high sampling rate thereby producing large amounts of data. Transmitting such vast amounts of data proves to be problematic as the available bandwidth on mobile nodes are limited.
Significant data reduction and analysis is essential before one can derive insights into vehicle behavior and performance. Incorporating additional regional meteorological data with
in-vehicle measurements provides a link between these two distinct datasets:
- Vehicle behavior correlated to certain weather conditions can be quantified to improve vehicle design in future applications.
- Vehicle behavior correlated to specific weather phenomena can be used directly to improve real-time weather sensing and now-casting.
3. Machine learning required
Extracting knowledge from large amounts of time-series data is another active area in the field of machine learning algorithms. Elaborating on the correlation analysis of a vehicle versus weather phenomena does not exist yet. In addition, vehicle data is largely contaminated by driver behavior (e.g. forgetting to turn on/off fog lights, wipers etc.).
Advances in data interpretation, aggregation and sensor fusion are needed to improve the reliability of collected vehicle data. Unsupervised learning algorithms will need to cluster time-series data and evaluate their performance regarding the value of the extracted knowledge. These contributions will impact the vehicle safety and performance.
There is also an additional challenge which lies in predicting the weather along a trajectory. This will not only require incorporating the real-time data collected from the vehicles, but also the integration of a predictive aspect and advanced correlation between the various data sources. Yet the complexity of spatial and time sample alignment is a computing power-intensive problem.
The GSMA Connected living programme, mAutomotive, predicted that the connected car market will be worth $151.8 billion in 2020, whereas its value was $13 billion in 2012.
To provide human interpretable output and real-time road weather warnings, various obstacles will have to be overcome:
- Real-time mapping and superposition with satellite and radar images and integration in existing systems.
- Automatic generation and dissemination of short range warnings for hazardous weather conditions.
- Investigating how to integrate vehicular sensor data in now-casts and warning systems.
- How to improve the quality (accuracy, timeliness) of the now-casts and warnings.
- Comparison of warnings based on vehicular sensor data only and warnings based on merged data (cars + weather observations + now-casting systems + road weather model forecasts).
Gartner reports that by 2018, one in five cars on the road will be self-aware and equipped with a combination of sensors, V2V communication and ever-increasing system intelligence.
4. So what can we conclude
The weather information is very accurate at the position of weather stations, but the reliability is strongly affected by the distance from the sources.
To solve all of these issues – acquiring reliable data to meet the objectives for warning drivers in real-time about upcoming dangerous road situations – significant advancements are required in the:
- Time and spatial data sampling and alignment
- Aggregation techniques of CAN-data
- Potential fusion of the CAN-data with additional sensors to improve the reliability of the crowd-sourced data.
From a business perspective, car data-enabled business models could trigger a new wave of opportunities, increasingly focusing on what happens ‘during’ personal transportation. These opportunities will be shaped by technological and business-related choices that will be made in the next years. A set of strategic questions need to be addressed to shape the scenario going forward, e.g.:
- How can customer value be created and communicated, in order to ensure that data sharing will be perceived as valuable?
- What are the different system building components and the required technologies for each subsystem that will be needed to develop the new offerings?
- When could investments start to pay off? When should pilot programs be killed?
As these questions are relevant to incumbents and new players in the arena, there may be different industry strategies. It is for sure that the industry competition and dynamics will increase strongly.
First, it will become more and more important for players to build digital capabilities (e.g. big data analytics, human-machine interface design and services set-up), gather digital new product management expertise and secure relevant technological and commercial partnerships.
Secondly, the potential impact of these new value propositions to the customer will become more and more transformative. Industry players will need to be ready to challenge even more fundamentally the way they look at value creation for the individual driver.