Developing a scalable data platform to support evolving IoT-based insurance propositions
Telematics has transformed the vehicle insurance industry over the past few years. Insurers are increasingly offering usage-based insurance (UBI) products, based on the data gathered from GPS-connected telematics devices.
Telematics, a portmanteau of telecommunications and informatics, enables the real-time transmission, receiving and storage of data and information from remote objects such as vehicles. The information could include location tracking of vehicles, driver behaviour, maintenance requirements and data on accidents.
Now, developments in Internet of Things (IoT) technology present a further opportunity for insurers to move beyond connected vehicles and offer new insurance propositions in the home, commercial property, life, and other insurance lines. IoT can connect insurers to the assets they are insuring, whether that is a vehicle, a property, or a person. This allows insurers to understand the exact status of any insured items and potentially take action in response to an abnormal situation, such as detecting a fire, a vehicle collision, or an abnormal heartbeat.
To take advantage of these next generation IoT-based insurance propositions, Octo Telamatics, a market-leading telematics service provider, developed a new platform, which they called the Next Generation Platform (NGP).
A pioneer in vehicle telematics
Established in Italy in 2002, Octo Telematics was one of the first companies to support insurers in the then vehicle telematics space. Since then, the company has established itself as the global leader in the telematics service provider (TSP) space. It has one of the largest global database of telematics data, with over 186 billion miles of driving data collected and 438,000 crashes and insurance events analysed (as of 31 December 2017).
Octo Telematics captures a comprehensive set of data from a vehicle, including the speed, location, and journey duration, as well as aspects of a driver's behavior, such as how harshly they accelerate or brake and how quickly they corner. This data is combined with further contextual information, such as the local weather conditions, road type, and current traffic situation, and analysed to provide the insurer with a detailed profile of the true risk posed by a specific driver at any particular time.
Using machine learning, the company can make more accurate predictions and risk models, thus allowing the insurer to calculate a premium level that accurately reflects the risk and usage to the level of the individual policyholder.
Telematics also offers the ability to detect and respond to crash and claim incidents in real time, and to react in the most appropriate way, by alerting emergency services or dispatching roadside assistance. The telematics data captured during a claim incident also provides insurers with a precise view of a claim, enabling decisions about liability, the likelihood of fraud, and estimated cost of repair to be made in a fraction of the time. As claims expenses typically account for between 70% and 80% of an insurer's costs, this can have a significant effect on an insurer's profitability. In addition, it can dramatically improve the customer's experience of the whole process, and boost customer retention.
Challenges of dealing with growing volumes and variety of IoT data
Octo Telematics had developed a proprietary telematics platform that had evolved over the last 15 years. However, with the growth of IoT and an order of magnitude increase in the number and types of connected sensors, Octo Telematics foresaw that the current platform would increasingly become a constraint on the company's growth ambitions.
One of the biggest challenges was that the existing platform was designed around the needs of vehicle telematics and could not easily accommodate other types of sensors, such as wearables, smart watches, smart locks, smoke detectors, and surveillance devices, which are becoming increasingly important components in IoT insurance propositions.
The growth in sensors also implied the need to monitor data from potentially tens of millions of connected devices in the near future. Though the old platform was able to support over 5 million vehicles, the platform was reaching the limits of scalability.
The data management platform also needed to accommodate a broad range of inbound data types, ranging from real-time streams from in-vehicle telematics devices to bulk uploads from other sources, such as third-party weather data. There is also significant variation in the formats of stream data depending on the application and capabilities of the sensor. This can vary from relaying a simple journey start and finish time through to detailed crash reconstruction data from units incorporating sophisticated sensors, such as six-axis accelerometers with very high sampling rates. This data diversity will increase further, potentially to include images and video streams.
The increased number of items being monitored would also require significant growth in the compute and storage capability needed to support real-time analysis across many millions of sensors.
The data captured by the platform, whether from vehicles, properties, or people, needed to be closely coupled to the incident response and claims processes to enable insurers to offer policyholders a fully integrated IoT insurance proposition. The existing platform lacked this high degree of integration.
In addition to capablities of accommodating potentially orders of magnitude, the data management and analytics infrastructure would also have to ensure the total security of all inbound data or "data in motion," as well as that of data within the platform residing on disk and other storage media ("data at rest"). The challenge is compounded by the sheer volume of data from connected sensors, numbering in the millions, distributed across a wide geography.
Adopting and implementing a co-innovation approach
Due to the scale, complexity, and criticality of the development needed to realise the NGP in a time frame that would allow Octo Telematics to capture the emerging IoT opportunity, the company decided to adopt a co-innovation development model.
Octo Telematics used its understanding of the evolving insurance market to define the functional requirements of an NGP capable of supporting a broad range of new IoT-based insurance propositions.
Key technology partners were identified for the development of the NGP: Cloudera, Software AG, Salesforce, SAS, and SAP. Using this co-innovation approach, Octo Telematics and its partners were able to accelerate the design and implementation of the NGP, delivering a complex and challenging development project in under 24 months.
The initial phase of formulating the approach and conducting a dialogue with the partners to refine and improve the architecture of the NGP took seven months. A jointly agreed co-innovation roadmap was created. The implementation took 18 months of development, with an initial prelaunch version being released to key existing Octo Telematics clients at the end of 2016. Following the beta testing phase, the full commercial version of the NGP was released in July 2017. All new Octo Telematics clients are now supported on the NGP, with a migration plan in place to move the majority of existing clients to the new platform.
Adding 11 billion additional data points daily
The resulting NGP enables Octo Telematics to store, process, and analyse data generated by over 5.3 million drivers totaling 175 billion driven miles, and that increases by over 11 billion additional data points daily. It also allows for complete flexibility in the selection of sensors, analysis and output of data for all insurance and automotive services.
And the backbone of this NGP is powered by Cloudera’s machine learning and analytics platform. The Cloudera Enterprise suite includes a set of tools to provide security, governance, and workload management functionality operating within an integrated data and platform model. The platform provides the underlying infrastructure to ingest, process, and analyse huge volumes of structured and unstructured data, while being able to perform analytics on both streaming and static data sources. All inbound data, data moving between multiple clusters, as well as data stored within the platform, is encrypted.
A "scale out" hardware approach was adopted, as opposed to “scale up”. Scale-up is done by adding more resources to the existing nodes of a system, while scaling out involves adding additional infrastructure capacity in the form of new nodes, which can be done through the use of commodity on-premises and cloud-based hardware. This avoids the need for investment in expensive high-performance servers, as storage and compute requirements rise.
The NGP also utilises Cloudera's Shared Data Experience (SDX) module to define and enforce unified user and role-based access and security policies, as well as provide auditing capabilities at the application, cluster, and environment level.
Using Apache Spark, Octo Telematics is able to leverage the huge volumes of data, the compute power of multiple clusters, and a resilient distributed data set (RDD) structure to quickly implement, train, and test machine learning models.
These models allow Octo Telematics and its insurance customers to better understand, model, and price risk, and can form the core of new innovative insurance products.
Inherent in the Cloudera Enterprise platform's distributed computing model is the ability to operate the NGP both on-premises and across private or public cloud. The ability to flexibly use major cloud service providers such AWS, Google Cloud Platform, and Microsoft Azure means the NGP can support transient but compute-intensive projects, such as testing new pricing algorithms or risk model development, on a usage-based commercial basis.
The NGP has resolved the capacity issues of the previous platform and is now continuously scalable. It will only require additional cloud-based compute and storage resources to accommodate the growth.
The enhanced functionality in areas such as CRM and incident analytics, as well as the increased capacity of the NGP, means that Octo Telematics can offer all insurance clients detailed, real-time crash reconstruction capability. This will allow users to drive significant efficiency improvement in claims processing, identify potential fraud, and enhance the customer's claims experience.
Octo Telematics' insurance clients also benefit from the additional functionality of the NGP by being able to introduce new types of IoT-based insurance products. For instance, one client introduced a property insurance product that uses a home hub, developed by Octo Telematics, that is equipped with smoke, heat, flood, and intrusion sensors. Another insurer introduced a pet insurance product using IoT-based GPS tags worn by the pet. Yet another is piloting the use of smart watches as part of a health and life insurance offering.
Furthermore, the NGP is reducing time to market for new product launches by more than 50%. The time to implement a new UBI product has been reduced from two to three months to four weeks.
Currently, most insurers implement UBI offerings as stand-alone projects requiring parallel core administration and claims systems. To address the inefficiencies and complications from this, Octo Telematics is working with core insurance software vendors to develop a range of connectors that will allow direct integration between the NGP and an insurer's core processing systems. This direct integration will significantly reduce the cost of entry and complexity for insurers wanting to offer IoT-based products.
As of November 2017, Octo Telematics had developed a connector allowing direct integration of the NGP with the policy administration and claims suite of Guidewire, a software for property and casualty (P&C) insurance providers.
Octo Telematics is also looking at extending vertical-specific functionality to the NGP beyond the telematics sector, in support of a wider spectrum of industries, such as the telecoms, energy and utilities sectors.