How insurers can make still make use of “personality” data
16 November 2016
In early November, a leading UK personal lines insurer announced the launch of an app that would use Facebook data to “better understand first time drivers and more accurately predict risk”. This was intended to award discounted premium rates to safer drivers, who would be identified on the basis of specific personality profiles. Shortly afterwards, Facebook announced that this contravened its privacy policies, triggering a change of plan for the insurer.
While the social media company’s response has made headlines, the intent behind the insurer’s actions is far from new. For quite some time now, insurers have been talking about making greater use of data garnered from social media in order to segment customer bases more accurately. They have also made use of customers’ social media posts to successfully identify instances of fraudulent claims.
Concerns around data privacy have resulted in a number of regulatory interventions:
- In late 2015, the Financial Conduct Authority undertook its own consultation with the UK insurance industry as to how Big Data might be used in relation to underwriting and setting premium rates. Particular concerns have included the exploitation of vulnerable customers and certain types of customer being denied coverage or subjected to very high premium rates.
- The EU General Data Protection Regulation has been issued in draft form, and insurers are currently working their way through the significant implications of this before it comes into effect in May 2018. Whilst the levels of fines for a data breach have attracted much attention, there are also a number of other salient features, including the portability of data and the need to identify where all customer personal data is held within internal systems and back-up facilities, to enable “the right to be forgotten”.
If the hurdles around personal data permissions can be navigated, there are a number of other powerful ways in which customer segmentation can be achieved. Examples might include the following:
- The ability to utilise information on where individual customers go on a day-to-day basis, by using the “location services” features on smartphones. Location data could provide insurers with the ability to understand where their customers work; their shopping habits, leisure activities, health and fitness regimes; and their lifestyles more holistically. These data points would give a clearer insight into the personality and inherent riskiness of any individual.
- Analytics in relation to data (the web pages viewed by a visitor and the “succession of mouse clicks” made by a visitor etc.) will allow those individuals who do not provide permission for “location services” to be activated on their mobile devices to be analysed in greater detail. While much of the initial focus may well again be on online shopping habits, key lifestyle indicators based on websites visited etc., an analysis of clickstream data will also give some indications around customers’ behavioural characteristics, notably around attention spans, engagement levels, propensity to switch etc. In relation to insurance quotations, the analysis of clickstream data will also provide insights into how quote manipulation may have been attempted by specific types of customer.
- Alternatives to the relatively expensive black box approach to telematics for private motor insurance (designed to monitor driving styles and when journeys are undertaken) include the use of smartphone apps, although the level of sophistication is compromised compared to bespoke black box technology. The advent of eCall technology, which must be available within all new motor vehicles sold in the EU from 2018, might address this issue, if GPS and on-board computer data is transmitted continuously. Other solutions, such as smart windscreen attachments, will also provide telematics technology at a fraction of the cost of the black box. The telemetry will provide indicators of driving styles, as well as ancillary information such as journey times, commuting versus leisure journeys etc.
- Splicing mobile phone data with telematics data will additionally give vital information on whether motorists are texting or making calls while driving, which is a significant contributor to accidents in the UK.
- Wearables have already garnered a wide level of interest, with some insurers offering incentives to undertake a healthy exercise regime and to reward these good behaviours with discounts on premiums.
- The ‘connected home’ is starting to offer a source of valuable customer information from a pricing perspective, including when the home is occupied, room temperatures (especially in cold weather when burst pipes are a risk), whether burglar alarms are set when the house is unoccupied etc.
- The ability to make greater use of credit score, credit card transaction, banking and other financial health type information continues to be an area in which insurers have been investing. For insurers (and banks) the ability to understand household balance sheets, household profit and loss accounts and overall disposable income, combined with the application of behavioural economics, will give much greater insight into the decisions being taken by individuals (e.g. around switching and extent of income drawdown from pensions) and their appetite for risk.
- Market research data from third party vendors (such as Axciom) can also be used to gain a fuller picture of socio-economic, as well as a wide variety of other profiling, information on a geo-coded basis.
- While there is a strong belief that posts on popular social media sites and apps provide insights to the characteristics and personality traits of individuals, sociologists have also argued that there is a lot of pressure on individuals to fabricate or exaggerate elements of the lives they profess to lead via social media. As such, any personality insights which insurers might believe that they are gleaning may prove to be fallible and not necessarily entirely reflective of reality.
When insurers seek to make use of information for risk selection and pricing purposes that has not been sourced directly from customers, they must be able to differentiate between information that is predictive and that which isn’t, and also between information that is truly factual and that which is open to subjective interpretation. Those insurers that get it right will benefit from their superior risk segmentation and pricing capabilities, while customers will benefit from premium rates which more closely reflect their risk characteristics and so lead to less cross-subsidy of poorer risks.