Editor’s note: Spotam, an Israeli consultancy and research firm focusing on telematics, location-based services, and the Internet of Things, recently interviewed Avner Freiberger general manager of the Verisk Innovation Center in Tel Aviv. In the following article, Avner explains what the center is doing to advance Verisk’s telematics initiative.
Please describe Verisk’s Innovation Center.
The Innovation Center is dedicated to the development of Verisk’s telematics and IoT algorithms and actuarial models. We use predictive modeling, data science, and machine learning methodologies to develop models that predict driver risk. In addition, we participate in the definition of the telematics products. Models created in Tel-Aviv were filed and approved by 43 U.S. jurisdictions for underwriting and are used for driver pricing.
Can you tell us about the [Verisk Data ExchangeTM]? How is it structured, who is expected to connect to the exchange, and what services can it give its customers?
To explain the Verisk Data Exchange, it’s important to understand the state of usage-based insurance (UBI) in the United States and throughout the world. UBI allows insurers to price drivers according to how they drive and how much they drive.
Currently, to run a UBI program, an insurance carrier has to invest in infrastructure, communication, devices, analytics, filing, and operations. On top of all that, the insurer is asked to give discounts to drivers (currently surcharges are not common), making the ROI unclear. To add to the confusion, insurers are asked to select the technology from a variety of choices, none of which are safe. Dongles are expensive, smartphones are driver-centric and less accurate, and connected vehicles are not common enough.
On the other side of the equation, data contributors invest in collecting data for various purposes. They want to offer a good experience to their customers and to also monetize their data quickly and effortlessly. Unfortunately, there are hundreds of insurers in the U.S. alone, so it’s nearly impossible to sign with all of them and make the best of the data.
The Verisk Data Exchange provides a solution to most of the problems, including the “many to many” problem.
On the contributor side, we integrate with the various data vendors—but not before we use our experience to assess their data and approve its quality. We harmonize the data, store it in our database, and analyze it. On the insurer side, we provide pre-collected data at the point of sale or claim in several different products:
- Driving Data: This product enables insurers to retrieve full telemetry data as it was collected by the vehicle or device. Here we are merely a reliable conduit.
- Verisk Driving Score: This driver score is trained on insurance premium and loss data. The score also has a driver feedback component so that drivers who are not highly scored are able to understand and improve their score. This product was developed by our team.
- Verified Mileage: These are actual odometer readings sent by the vehicle. It can be done in near real time or as a monthly aggregate. There’s a fantastic demand because verified mileage replaces the insured’s statement.
- Model-Ready Data: This is an exciting product for large insurers. It allows their actuaries who want to study the telemetry data and build their own models to receive anonymized telemetry data (their insured) along with premium and loss data. They can use it to study the data and decide if they want to use our score or build their own.
- Instant First Notice of Loss (FNOL) - Auto: This product provides insurers with near-real-time notification of loss, which is extremely important in claims management. We work with OEMS on pushing claim notification along with the car’s “black box” data. Currently, insurers are not exposed to such data.
It goes without saying that the value of the exchange will be in the amount of drivers in the pool. As you know, we have an exclusive agreement with GM, and we’re in negotiations with several more OEMs as well as TSPs and MSPs. We have several millions active users, and the numbers increase daily.
What types of data are stored in the Verisk exchange, and how is the data being collected?
The Verisk Data Exchange is designed to receive data from various data contributors. Each contributor is capable of providing different data elements, resolution, and frequency, making each data vendor unique. To set order, we defined a six-level data ladder so that each contributor can fit its device to one of them. We make sure that each layer will be backed with the most appropriate model. Our analytics suite is capable of analyzing various data sets—from a thin data set that has no location to a very rich data set with location and sub-second accelerometer data.
How does Verisk plan to handle privacy issues?
Privacy is critical. Everyone is concerned with Big Brother monitoring every move. Verisk requires two consents from consumers before sharing driving data: the first consent is to actually participate in the program and to allow collection of data; the second consent is for sharing data or a score with insurers. This way, the consumer approves his or her data at the point of sale just before it exits our exchange.
Insurers have had the option to work with telematics service providers and use telematics data for their vehicle insurance plans for a few years now. What are the advantages of the new Verisk service over legacy insurance telematics services?
It’s clear that usage-based insurance, in its current implementation, is not reaching the tipping point everyone is waiting for. I think the main reason is that we haven’t seen a significant success story that insurers can reference and follow. I don’t think that’s because insureds prefer not to join these programs—insurers report very high user enthusiasm. I think that building and maintaining a successful UBI program backed with a reliable predictive model is not in
\insurers’ DNA, and they prefer not to jump in the water. Currently, most insurers are just exploring or piloting; just a handful are actually running large-scale commercial programs. I think that Verisk’s exchange resolves most of the obstacles and allows insurers to test and then scale their programs easily.
The recent growing activity in the area of smart homes creates new opportunities for telematics-based insurance in the home insurance area. Is Verisk involved in this area? Can you tell us more about it?
Verisk has already started exploring connected-home, wearables, and other IoT data sets; and we’re aware of the challenges they introduce. As opposed to driving telemetry data, which is rich and relatively consistent, other IoT data sets are thin and extremely diverse. It’s a real challenge to model smoke detector data that registers an event every couple of years together with a door sensor that sends messages dozens of times a day (and I have many more examples). I trust my team to come up with resolutions to these challenges.
This article originally appeared in Spotam and is republished in modified form with permission.