An Interview about The Floow with Ptolemus's Frederic Bruneteau

This interview was finalised in September 2016 and published in Ptolemus latest Connected Insurance Analytics Report

Dear Aldo, could you please tell us briefly about the Floow’s position now?

As an organisation, we completely focus on mobility and are inspired by the vision of making individual mobility safer and smarter.

Currently, our team counts 70 people, built over a period of 5 years. We are very proud of our scientific heritage, as the majority of our team is made of scientists from Sheffield University and other leading academic centres. The technical team includes computer scientists, data scientists and phycologists. This make-up provides the organisation with a unique blend of hard sciences and social sciences working under one roof.

We serve a large and growing number of leading insurance organisations worldwide. Our major markets are the UK, Ireland, Nordics, Continental Europe, and North America. We have been investing heavily in the Chinese market to build an early presence there. I am pleased to say that today sees the launch of our first Chinese pilot with one of the top 10 Chinese insurance companies.

We are focused on UBI. In fact, almost all our clients are insurance companies. And, we help them bring insurance telematics to market through a device agnostic platform. We are completely neutral to the source of data as we work with all devices: smartphones, OBDs, conventional black boxes and original devices embedded by OEMs.

That said, smartphones have a special place in our offering. We pioneered smartphone as sensors since 2012 and our telematics app is considered an industry leader, featuring efficient auto start/stop, minimal battery use, severe crash detection, engaging end user services etc. This was confirmed by the recent Ptolemus N.1 smartphone telematics service provider award for Europe.

We have arrangements with car makers such as Renault-Nissan and PSA to help them monetise data generated from vehicles, effectively exploiting vehicles’ original sensory capabilities and removing the need for aftermarket devices.

Today it seems you are positioned firstly as an analytics provider, do you also provide the telematics part?

Yes. We cover the full value chain, from data collection to data processing, management and interpretation. On data collection, excluding smartphones where we promote our proprietary app technology, we partner with high quality hardware specialists such as Meta Systems, Trackm8 and others. From our first day as a business we opted to work with hardware providers and take responsibility for the transformation of raw data into insurance grade insights and services.

The UK has been a disappointing market for UBI in the last 2 years, as it seems to plateau at around 500,000 policies. How can analytics help them go to the mass market?

I think disappointment is relative to the expectations of an observer.

We do not believe UBI will experience a “big bang” dynamic. Instead It will expand very gradually and in fits and starts, at a pace well below current  expectations.

However, a market of around 500,000 policies provides sufficient volume for good quality analytics to predict the probability of an accident based on how you drive relative to how others are driving in the same location.

Our sense is that we are still at the beginning of a journey towards mass market telematics. Telematics is still very much a “push” proposition where insurance companies are reaching out to their client base and trying to entice them with discounts. The challenge and the opportunity is in to transform this “push” into a pull proposition. This calls for a range of rich and diverse services that engage users more, and leverage the incredibly granular understanding of individual mobility patterns afforded by telematics.

Telematics data and analytics can also help bringing other stakeholders into the ecosystem and open up new opportunities. For example, analytics can be used to charge road usage based on driving behaviour, or in estimating the value of a vehicle to be sold in the used-vehicle market based on its validated historical usage. Such concepts are on the horizon but it will take time to bring them into the public domain.

How efficient are your driving scores in predicting losses? To what extend are you able to calibrate your scoring with real claims losses?

The correlation between our score and people who have claims is very high. We are very close, based purely on GPS data and in percentage terms, in being as predictive as an insurer who uses their knowledge about all of the customer’s other risk factors. The incremental benefit regarding the accuracy of prediction when using our telematics scores is between 10% — 30% and is about 10 times as predictive as other new factors like credit scores.

Our scores are calibrated against external or 3rd party data sources and then against insurers’ own claims data. 

Event-counting and thresholds are now often described as insufficient. What is your recommended approach to rating driving behaviour?

We have avoided event based approach from the very beginning of our business. We follow a continuous contextual analysis approach where we look across the entire customer journey.

What are the biggest analytics-related challenges that you are facing today?

Getting good quality and consistent data from the different devices in our portfolio whether it is Android or iOS smartphones, OBD devices or black boxes and gaining consistency across manufacturers is always a challenge.

Then making sure that context is properly taken into account and  understanding how different road infrastructures such as motorways or city roads affect driver behaviour.  

Having enough claims data to train our new scores used to be an issue as well. However, we are now at the point where this can be done as we have collected granular data from the very beginning and claims data are flowing through given volumes and exposure levels achieved.

Ultimately our objective is to create behavioural scores that are highly predictive of future accidents. However, these scores should not be taken too far or overcomplicate things and leave the insurers with over fitted models. Scoring methods should be simple and understandable by the customers both end users (drivers) and insurers.

Some statistics show that driver distraction is the number one factor in fatal car accidents. What are your thoughts and plans to address this endemic problem?

Distraction is a massive problem and one we focused on since day one. Our smartphone based scores include distraction, measured both in terms of call state and phone manipulation. However we observe that people opting for our telematics services tend to not use their smartphone much while driving perhaps because they are aware of the monitoring function in place.

Today do you recommend your insurance customers who use an app for driver monitoring to also use a beacon in the vehicle?

It depends on the variable that our insurance partner is trying to solve for. For example, many insurance companies deploying telematics are constrained by the need to capture data for every instance of vehicle movement. In this case, smartphone solutions, which are  great to collect behaviour information, are not sufficient.

So, in the presence of such binding constraint, we may encourage our clients to consider a tethered solution which gives them the ability to collect data from vehicles and minimise the cost of data transmission thanks to the smartphone’s cellular engine. However, “tethered” has complexities from a customer experience standpoint as it demands end-users to perform multiple actions such as installing the device, downloading the application and paring the device with the smartphone. If an insurer needs to solve for vehicle movement data and slick customer experience, a traditional OBD may well do the trick, though at a higher cost.

We are still in a world of sub-optimal solutions, and this will change when vehicles are equipped with original telematics capabilities. However, it will take years for in-vehicle telematics to become a substantial proportion of assets in circulation.

What is your method to accurately rate different vehicles e.g. a Fiat 500 and a BMW series 5?

Driving behavioural scores can be tailor to the vehicle if required - i.e. you would expect that a BMW may be able to accelerate faster than a Fiat and can do more damage if it were to hit something. However, to a large extent the differences between claims frequency and costs is already allowed for in an insurers standard rating and may therefore not need to be explicitly allowed for in the behavioural scores depending, of course, on how you are integrating them into the overall price.

However, there are only limited factors where this is the case; speed, time of day, fatigue etc are all the same (relatively speaking) in any vehicle. Even harsh braking (despite possibly being safer in better equipped vehicles) is generally an indicator of lack of attention to the road conditions ahead and therefore could be equally penalised regardless of vehicle.

Ideally you may look to create scores based on how close to the edge of the vehicles safety envelope you drive but since most driving is not so much limited by the vehicle as the traffic around you this may be unnecessary.

As more claims and exposure data is collected, the way in which telematics products are rated is likely to move more towards the behavioural scores with minor adjustments for traditional factors vs the current approach of mainly traditional factors with a minor adjustment for telematics and when this happens there is likely to be a requirement to interact the scores with some of the other factors such as the vehicle being driven.

Several insurance companies have recently indicated their willingness to build their own driving scoring algorithm. what would you tell insurers who are tempted to do the same?

There is an interesting polarisation between insurers who believe they can do everything internally and those who choose to collaborate with service providers and technology vendors like The Floow.

Call me biased, however I believe the latter group stands a much better chance to integrate telematics productively and cost-effectively in their operations.

It takes a lot of experience to acquire, store securely, cleanse, process, contextually enrich and extract meaning from unimaginable quantities of raw, noisy GPS and  accelerometer data. Our insurance partners allow us to play this role and work collaboratively with us to benefit from our exposure to global markets and different experiences.

In addition, we have designed our scoring methodology to be customised based on individual clients, and we encourage clients to use it as a tool to generate unique algorithms specific to their requirements. Our insurance partners use The Floow technology to build their own unique IP.

Could you please tell us more about your analytics partnership with Agero in the US?

Agero has a license to distribute our technology in the US market and  they are currently engaged in a number of UBI pilots. They are a value-added partner who can  either resell our technology on a standalone basis or incorporate it with its proprietary road safety and emergency services.

Could you please explain the role you are playing in the Renault-Nissan data hub? When should we expect it to be active?

OEMs like Renault and Nissan are key partners to The Floow. They own the vehicle infrastructure and control access to data of high quality and of great interest to our insurance partners. Our role consists in helping them extract value from the potential embedded in their original raw data. The transformation of raw data from the vehicle into insurance grade scores for underwriters is at the heart of our relationship with the Alliance.

The relationship is fully functional and live. We expect that we will sell our first policies in the UK market by the end of this year. The data will be generated directly from the vehicle using Renault’s R-Link devices and would be sent to us for scoring as part of a UBI proposition.

It is a global arrangement designed to be implemented in any geography where we find willing insurers. However, the first market will be the UK as we have clients there who are eager to bring this solution to market.

Your company managed to forge a partnership with CarSmart in China last year. What is coming out of it?

We are in the process of forming a joint venture with United Electronics Group, parent of CarSmart, one of the China’s largest system integrators with a market capitalisation of almost 3 billion euros.

We are contributing our scoring technology, app technology and expertise with insurance companies worldwide and our partner is providing the data centre infrastructure and adapting our app to cater to Chinese customers. We have already built the joint platform and are starting our first pilot, effectively from today, with one of China’s largest insurers.

What do you expect from your participation to the MOVE project with JLR (Jaguar Land Rover) and Direct Line in the UK?

This project is the largest UK government funded initiative for research in autonomous driving. It is a great platform for us because it gives the opportunity to collaborate with organisations of the calibre of Bosch, DirectLine Group and Jaguar Land Rover amongst others. It has offered our team a very clear perspective of what our role as a TSP will be in a future of increasingly autonomous vehicles.

So how relevant do you think telematics and analytics will become in the autonomous age?

The challenge for an insurer in the world of autonomous vehicles is to price the risk of a journey where the vehicle governance is outsourced to a system. To overcome this challenge, you need a new generation of risk models that inform on the probability of something going wrong. You need to be able to monitor the vehicle operating in context whether driven by a human being or by a system. That is a core capability at The Floow. We are equipping ourselves to support vehicle manufacturers, tier-1 suppliers and insurers to deal with the new categories of risk that will arise from autonomous vehicles.