Backend of a prototype telematics system

Key Takeaways:

  • The client is one of the largest telematics organizations on the market, offering solutions for fleet operators and car insurance companies.
  • We took the role of an external innovation lab to develop the backend for a prototype smartphone-based telematics system to open up new markets for the client.
  • Our solution covered developing a fully functioning, scalable system that collected various data and translated it into individual driver score.
  • The system went live as we performed multiple production deployments. At the peak the system had 100k+ users a year, processing hundreds of requests per second.

In the right circumstances, R&D may seamlessly transfer from a side project to a competitive advantage integrated throughout the business. In the following case, we'll present an example of such a situation considering one of the major telematics solutions providers and their efforts to face the challenges of the modern Usage-Based Insurance landscape (UBI - a type of vehicle insurance focused on individual approach to each driver, where costs are dependent upon the vehicle characteristics, traveled distance, or road behavior.)

New territory to acquire

At the beginning of 2015, our client was already one of the largest telematics companies on the market. Their solution was best suited for fleet operators, as their telematics system was based on external devices that had to be installed inside cars. For individual policyholders dealing with external devices can be too costly and cumbersome.

At the same time, a brand new branch of telematics was on the rise – smartphone-based solutions. Being cheaper and much more ubiquitous, it was a perfect fit for individual customers wanting to use UBI.

Our client wanted to start operating in this area to extend their business offer. They assembled an in-house enterprise startup, with Sparkbit taking the role of an external R&D lab.

Mobile telematics experiment

We were presented with a high-end vision of the product, with no specific architecture, technology, or methodology requirements. We started the project knowing it'll be a textbook R&D.

We knew we had to build and implement a horizontally scalable, dependable system to collect trip data and design algorithms for evaluating users' driving skills. There was no clear roadmap, yet we had to release the prototype quickly to validate our initial ideas and gather user feedback.

Data collection methods and driver evaluation algorithms were carefully chosen and improved over many iterations. In 10 months the MVP was ready, and the product went live with all the main features implemented.

Going mainstream

We developed a feature-rich system to measure driving skills using heterogeneous data such as real-time GPS and accelerometer data, weather and road conditions, or road types. Our algorithms detected hazardous events such as speedings and harsh accelerations. Each trip was evaluated and scored based on various criteria. 

As more and more insurance companies were attracted to our client's solution, we started performing production deployments. The final product was customized many times, as some clients wanted to increase user engagement with additional in-app elements like social features, a fun factor - gamification, collecting badges, or presenting drivers with safety tips.

Our system was battle-tested with 100k+ users in a year, processing hundreds of requests per second, proving that our architecture perfectly suited the requirements.