Connected Cars: Data To Dollar$

Technology Perspective February 2018 Connected Cars Data Dollars

We have come a long way, in terms of evolution of technology, including wireless communication that has progressed from 1G 17 years ago to 5G, which we will see in the near future. However, the elusive “killer app” has evolved for the automotive industry in the form of Data. This report is an excerpt from Cresttek’s high level perspective on converting data to dollars.

Data always existed in OEM, dealer and supplier databases. The key was to create tools and technologies to discover, identify, manipulate and process this data in a real-time scenario at low-cost using modern cloud solutions. Now powerful catalysts like Artificial Intelligence (AI) and Machine Learning (ML) are available from the likes of Google and its Tensor Processing Units (TPU) at a really low cost. It is so affordable that even a small start-up can manipulate and create killer apps using data. Couple this with the crowd sourcing of algorithms/ predictive models by tapping into the brilliant minds of global scientific talent, and it becomes priceless for any company; providing them with the opportunity to leapfrog competition.

If you use private equity (PE) and venture capital (VC) investments as a proxy for what could happen in the automotive space, it is clear that autonomous and in-vehicle technologies have received about $ 45 bn of investments in 2016-17, according to Brookings and Crunchbase. Less than 10 % of these investments are from automakers. So, who knows if new start-up bets could displace any major OEM in the next decade?


Cars have evolved from mere metal on four wheels to sophisticated, connected, smart and soon autonomous cars. They can now also drive themselves using the increasing electronics content (1). This journey took over 100 years and Detroit was at the centre of it all. After the great recession of 2009, Detroit is being transformed into a connected car technology hub. In this paper we are using “Connected Cars” as an all-encompassing word that could mean autonomous, internet of things, telematics, etc.

(1) Vehicle Electronics Content Growth

Increasing electronics in cars has created challenges for OEMs in managing and integrating new and advanced electronics. As shown in (2), the integration challenge increases up the pyramid. OEMs now need technology suppliers to roll out new products to the market faster and at lower costs.

For example, take an OEM using an ECU for crash safety in-car. It would need semiconductor chips from Samsung, ICs from TI and processors from Intel, all assembled on a PC board by Flextronics. Along with that, Autoliv may integrate the hardware with flashed software to meet OEM requirements with months of testing. This means the ECU has to function in extreme weather conditions, from -40 °C to over +125 °C. If you do not believe how difficult it is to pass this environmental test, take your smartphone and place it inside a car on a hot summer day. Rest assured, the smartphone will not work until it cools down to about 50 °C. These stringent requirements are necessary for automobiles due to the harsh environments in which they operate while maintaining customer safety.

(2) Automotive Electronics Industry Structure

So, what is a connected car? In simple terms, a connected car is an automobile that is intelligent and connected to the internet – constantly sending and receiving data just like any other smart device connected to the internet. In the future, this could also mean using techniques like ‘fog computing’, where processing happens closer to the edge of the network to avoid bandwidth issues when going back and forth to the cloud, (3). This means the network becomes an extension of the car, necessary for the vehicle to function efficiently and optimally. Of course, for safety reasons the cloud is not necessary to do manual driving.

(3) A Connected Car Architecture

If you are not a car person, it can be difficult to understand all the data a car generates through its communication interfaces today, let alone a connected car that could arrive over the next few years. (4) is an example of all communication interfaces that are used to transmit data in cars today. More data is also being generated by the driver/ passengers’ daily interactions with the car, its features, car dealers and other service providers. Now, with all the autonomous driver assistance features, data sets are growing exponentially. Therefore, understanding data volume, variety and velocity is crucial.

(4) Communication Interfaces in a Car


There are many capabilities needed to be a successful automaker. Just like any capital-intensive business, these capabilities include assets and competencies. The key assets and competencies that were required in the automotive industry for the last 100 years may not be enough to propel an automaker to the connected car world.

We believe the necessary capabilities at play in the connected car market are evolving rapidly. At a minimum, it is driven by the agility and speed to integrate and execute technology in a connected car. In (5), we have shown how the same set of assets and competencies need to evolve to play in the connected car “sandbox”.

(5) Capabilities to Play in the Auto Industry


The car of the past was only connected to the world through local radio stations for consuming music and news. The only way an automaker or any dealer could get feedback was when the customer brought the car to the dealer showroom.

Now envision a future where the car is connected to the internet just like your smartphone, and it could drive itself to your destination. While your car drives you around, you could be productive: make video calls; social: have a family gathering; or relax: get a motivational talk, while you share the ride with your personal advisor.

In this future scenario your car will be connected and aware of all the devices, public information, local ecosystem and roadway issues. The cloud will be the centre of connectivity.


Smart city is an urban or suburban area connected by sensors to supply information used to manage the area efficiently. This includes data collected from devices, and assets that are analysed to manage traffic/ transportation systems, law enforcement, information systems, schools, hospitals, and other community services. Sample ecosystems from a connected car perspective are shown in (6).

(6) Smart City Ecosystems

Smart city ecosystems provide huge monetisation opportunities. These opportunities are not just limited to cars. Of course, cars and transportation (public, shared, multi-modal, etc.) are the focus of this report, but would include industry, cloud and the public.


One does not have to be a math whiz to use AI and ML. Today, these are tools at your disposal to do predictive automation in the cloud. This is not to minimise the advanced work companies do in these areas but to make a point that these tools will become commonplace just like using office spreadsheets. So in the future, you can use intelligence as a service.

Connected cars and smart cities will be pumping out enormous amounts of data. What do we do with the data? How do we analyse and make sense of the data? The data could be in the form of text, binary, video, audio, images, etc. There are numerous AI models already available as a service to build, train and deploy, including the following – location-based pricing search, context matching in conversation, video analysis, image processing, speech processing, language translation and predictive analysis of driver/passenger needs.

In the next section we discuss how we can use these tools during the data discovery and analysis phase, to fine-tune the application deployment rapidly during the launch of a product or service from a data platform. An important caveat to keep in mind while developing and using learning models is that they are like a black box with quirks. Despite all the cloud tools that are available for AI, industry/ domain expertise is necessary for fine-tuning the models - primarily to correct and fix Type-1 (false-positive) and Type-2 (false-negative) errors.

Based on our deep industry experience in working with leading automotive companies, our methodology uses relevant use cases during the training phase of model development. A process flow of this methodology is illustrated in (7) to give a high-level concept for model selection. There are many ideas and concepts published about how to reduce bias and improve confidence. This requires understanding of both industry and the ability to apply the appropriate training dataset, as no model is fool-proof for every future scenario.

(7) Machine Learning Process Flow


How do we innovate and bring new ideas to the market? Every major company asks this question but very few have been as successful as a focused start-up. The key difference is that a start-up made up of a few founders (usually 2 or 3) is an ideal number to create and launch something fast without the typical organisation inertia. This is difficult to replicate in bigger organisations. Couple this with the technology disruption, and the outcomes become unpredictable – especially non-linear outcomes.

Looking at a couple of major ideas like the Google Search and iPod, they were launched by connecting the dots using trial and error. Since it is difficult to connect the dots and force intuition, we have developed a framework that has been proven in many client situations.

We have worked with many clients in customising a framework for their needs. A typical framework is made up of a combination of ideation and lean start-up approaches. This framework is a set of tools and principles that will help our clients discover ideas in the connected car market. It will also help them implement solutions and add value to their customers that will generate new revenue streams. This process is lean by nature – go to market fast or fail fast, (8). This is not a standalone framework; this will be combined with our ideation funnel, (9) to drive the ideas through a fast validation process.

(8) Framework For Agile Innovation
(9) Framework Driven Innovation Process Flow

Furthermore, we use a technology roadmap to look for non-linear outcomes, (10). When we combine multiple disruptive technologies, the outcomes become difficult to predict without an agile innovation process. Our data discovery process is broad-based and is part of our agile framework to rapidly ideate and deploy revenue generating ideas, (11).

(10) Technology Roadmap to Connect the Dots to Outcomes
(11) Data Discovery-Deployment Process


The beauty about car data is that any player in the supply chain can access it, with varying levels of difficulty. The natural owner, the OEM, has absolute control of data and is in the best position to take advantage of converting this data into dollars. However, this strategic advantage could be squandered, if OEMs are unable to deliver value-adding data-to-dollars products and services in a timely manner. Many players are already starting to access that data through smartphones and aftermarket products/services by providing valuable services directly to the vehicle owners and their local ecosystem.

It is our hope that this high-level perspective on converting data into dollars, and our proven framework for launching products and services faster with less risk, will help your company develop a competitive advantage using connected car data.


MADHU S NAIDU is Co-Founder at Cresttek – An ALTEN Group Company in Troy, Michigan (USA)