their car stolen, by comparing their behavioural data with that of thousands of other drivers in their database. This data is sometimes captured and transmitted from a specially installed box fitted to the car or, increasingly, from an app on a driver’s smartphone.
US insurers Progressive offer a great example of a business which has committed to working with data to enhance its services. It has created what is calls its Business Innovation Garage where technologists known as “mechanics” produce and road-test innovations. One project involves rendering 3D images of damaged vehicles using computer graphics. Images are scanned in from cameras to create 3D models allowing structured data to be recorded on the condition and damage to a vehicle.
A similar revolution is underway in the world of health and life insurance due to the growing prevalence of wearable technology such as the Apple Watch and Fitbit activity trackers, which can monitor a person’s habits and provide ongoing assessment of their lifestyle and activity levels. According to research by Accenture, a third of insurers are now offering services based on the use of these devices. One of these is John Hancock, which offers users discounts on their premiums and a free Fitbit wearable monitor. Customers can work to reduce their premiums on a sliding scale by showing that they are improving on their unhealthy behaviors. Brooks Tingle, the company’s SVP of insurance marketing and strategy, says “customers don’t mind giving up some data if you’re transparent about what data you’re asking for, and they are getting real value back for it.”
Of course, in both of these cases, certain ethical concerns, such as privacy, need to be taken very seriously. Driving telemetry works by monitoring every road which a driver takes their vehicle onto, with each being graded with a danger rating, based on the number of reported accidents either on that particular road or on others which fit its profile. This doesn’t mean that an insurance company needs to keep a log of every journey which a driver makes (although, due to a current lack of transparency on the issue, some of them may very well do). To some extent this can be done anonymously – with the GPS locational data being encrypted as it enters the insurers’ analytics system, and then output in a form which doesn’t breach a user’s privacy. For example it doesn’t need to directly tell the insurer which particular roads a driver uses (which would be unstructured data and therefore not particularly useful for analytics). Instead it can simply state that a driver spent 40% of their time travelling on roads which are categorised as “Grade A” for safety, 20% on roads categorised as “Grade B” etc. This is an example of unstructured data being transmuted into structured data which is far more likely to provide actionable insights.
In health insurance, the potential for problems around privacy and ethics are more complex. Most would probably agree that it is fair, or efficient, for people who make unhealthy lifestyle choices such as smoking or never exercising, to pay higher health insurance premiums. However science has shown that, unfortunately, the biggest decider of our state of health is often the genetic makeup we are born with. In an age where the human genome is mapped and genetic profiling is becoming more and more common, it would be more efficient for those of us unlucky enough to be born with bad genes to pay more, but would it be fair? It is right that throughout history healthcare has been more costly for people born with genetic disadvantages. But now Big Data technology – genetic profiling and behavioural analytics – give us the chance to draw an empirical distinction between when ill health is caused by bad choices, and when it’s down to genetic factors beyond our control, I believe it’s important to make sure safeguards are in place to make this distinction and insure as fairly as possible. This is a situation which is more likely to be resolved through legislation than market forces alone.
Beyond setting fairer, more efficient premiums, Big Data has also proven itself to be of use to the insurance company in cracking down on fraud. In fact, according to the FBI, fraudulent claims, not counting health insurance claims, cost the average US family between $400 and $700 per year in the form of increased premiums.
Insurers use Big Data to tackle fraudulent claims through profiling and predictive modelling. Variables within each claim are matched against the profiles of past claims which were known to be fraudulent. When specific matches which are known to be indicative of likely fraud show up, the claim is flagged up for further investigation. These matches could involve the behavior of the person making a claim, as fraudsters will exhibit certain behaviors detectable by computer analysis which might slip past a human tasked with manually assessing every claim. It could also involve examining the network of people they associate with, through social media or other open sources of demographic data such as credit reference agencies. It will also involve looking at partner agencies involved in the claim, for example auto repair shops, where patterns of behavior could indicate that a particular establishment is involved in nefarious activity which a claimant may be complicit in.
A third important use of Big Data within the insurance industry is in marketing. By gaining a more complete understanding of a customer by analyzing all of the available data, insurance companies can become more efficient in offering us products and services which will meet our needs. Increasingly this deeper understanding is being achieved through sentiment analysis of customer feedback and even social media activity. Algorithms comb through the unstructured data in telephone calls, emails and the information divulged on social media networks about what they do or do not like, to create personalized marketing strategies for each customer. And behavior while logged into their insurer’s web portal – such as length of time browsing FAQ and help sections, as well as participation in user forums and message boards, can be recorded and built into a customer’s unique profile. Justin Cruz, VP of strategic data and analytics at American Family Insurance, says “We have always been focused on providing value to the customer, but we are zeroing in on doing that in a more rigorous fashion, using data, using technology to prove whether we’re impacting the customer in a positive way.” His company has licenced analytics software from Applied Predictive Technologies called Talk & Learn, designed to increase understanding of customer data and make suggestions about products and services.
Marketing departments within insurance companies have also begun to use Big Data to identify customers most at risk of cancelling or leaving. As with policy underwriting or fraud detection, this is done by comparing the data on a customer’s activity to that of customers who have cancelled their policies in the past. For example, if an analytic system flags up a high (or low) number of calls to a helpline as a possible indicator that a customer may soon leave, then effort can be committed to attempting to change that person’s mind. This could be done by offering discounts or lower priced premiums, or, most likely more effectively, by dedicating more customer service time to sorting out that customer’s issues.