Delivering effective pricing is still today a core challenge for many insurers. The main reasons are due to:

1. Asymmetry of information. Buyers of insurance products possess better information about their own desires, preferences, and behaviours than any insurer would ever do. As a result, insurers rely on predictive models to infer premiums that are based on a heterogeneous number of risk factors for each future potential new insured.
2. Post sales risk profiling. Once insured customers are on cover, insurers gain insights on their behaviour to better understand high-risk from low-risk profiles post sales, and hence use this information to refine future pricing strategies.

The good news is that these challenges can be addressed in a much more informative way due to the rise of InsurTech startups and the possibility to observe customer behaviour better through the digital footprints each customer leaves behind when using connected devices.¹ Based on individuals’ behavioural patterns, insurers can develop personalization triggers to “nudge” them towards specific actions and test whether these interventions are effective to reduce potential claim incidences.

There are lots of InsurTech solutions out there, and many of the applications I have seen still seem to offer one-size-fits-all mechanisms, which very rarely consider customers’ differences. Personalization, when it exists, is restricted to the onboarding process, to optimize product selection based mostly on sociodemographic characteristics and self-reported attitudes and habits. Alternatively, by collecting data from social networks and web applications, insurers can go beyond demographic segmentation and infer individuals’ psychographic criteria, such as personality traits and non-cognitive skills that determine behavioural patterns.

When these factors are incorporated into customer analyses, the accuracy of risk assessments improves and, consequently, prices are optimized. In this chapter, I shall cover what these factors are and ways to identifying them across pre- and post-sales engagement processes, highlighting how to learn from customer behaviour and continuously improving ongoing price strategies.

InsurTech and the Role of Psychographics during the Pre-sale Process

Insurers’ profitability is highly dependent on the way they price and identify relevant factors that are associated with claims. In theory, insurers should have strong incentives to go beyond demographic characteristics to profile customers and identify new factors that describe their customers’ habits and ultimately their attitudes and behaviour. While demographic criteria are related to the structure of target segments, psychographic characteristics deal with individual personality, values, attitudes, and opinions² (see Table 1). These factors can be mined through large-scale observational data sets to feed better risk models. In fact, the search for user-level data is not specific to the insurance industry, it is in the mindset of many marketing departments that want to combine data collection and digital channels to achieve personalization at scale. For example, today, the advent of automation and data analytics allow many companies to “learn and test” using a variety of tools.³ As a result, these practices form part of some of the most advanced innovations that insurers can possibly use to redesign their value chain, improve their risk modelling techniques, and deliver significant new value-adding predictors.

It is important to note that observational data can now be more easily collected because of the online social interactions and shopping habits we now more readily undertake through mobile devices. Such engagements generate large amounts of digital records that are changing the way behavioural scientists observe individual actions and determine statistical patterns supported by clearly defined principles.⁴ For example, recent evidence on human personality traits has shown that computer predictions of digital behaviour are better predictors than judgements made by friends and family.⁵ Personality traits are one kind of psychographics defined as patterns of thoughts, feelings, and behaviour that serve as explanatory variables to predict a variety of social and economic outcomes.

The Big Five personality traits⁶ created by Costa and McCrae provide broad dimensions of non-cognitive abilities and emotional skills that shape human behaviour and preferences (see Table 3 in Appendix). Such measures have been applied in empirical studies to explain, for instance, health-related outcomes,⁷ risk propensity,^8,9 financial behaviour,¹⁰ and lifetime unemployment.¹¹

The body of evidence supporting these psychological factors as predictors of life outcomes has naturally led to their application within digital ecosystems. For instance, the myPersonality project¹² addressed the relationship between personality and digital footprints. Like Facebook, it has collected data from millions of psychometric tests from users. Its database is available on the website of the Psychometrics Centre of the Cambridge Judge Business School.¹³

An insightful review of the opportunities and challenges of using Facebook as a research tool to infer psychometrics was done by Kosinski et al. in 2015.¹⁴ The article follows the idea that the Internet of Things, through our smartphones and other digital devices, is operating like a psychological questionnaire, which we are constantly completing both consciously and unconsciously.

Among the psychometric measures currently available on myPersonality, there are proxies for personality traits and their facets, measures of cognitive skills (IQ), self-monitoring, and life satisfaction. Notable business applications are being developed in the area of behavioural micro targeting, one of which by the data science firm Cambridge Analytica.¹⁵ They use psychographics to try to anticipate customers’ needs and predict how their behaviour will change over time, to build relevant services that customers will engage and interact with.

In 2016, an attempt to use digital footprints was made by Admiral Insurance, a well-known motor and home insurer in the UK. It started to use an algorithm to identify safe drivers from Facebook “likes” and posts. First-time car owners who were identified as conscientious and well organized would get a discount on their insurance quote.¹⁶ This assessment was based on research evidence showing that conscientiousness is negatively associated with the crash risk of newly licensed teenage drivers.¹⁷ On the day of the launch, Facebook said that Admiral could not determine discounts based on its users’ social media behaviour, but only as a login tool during its onboarding process. An optional personality quiz is promoted to first-time drivers who have received a quote, which they may like to reduce. Every customer who takes the quiz is given a discount ranging from 1% to 10% based on the responses they give.

Psychometrics is the field of behavioural science that is devoted to the measurement of these psychological metrics, where it can be combined with a variety of individual digital footprints. It is important to differentiate between active and passive digital footprints. An active digital footprint is generated when personal data is made accessible online, through deliberate postings or sharing of information by a user, while a passive digital footprint is obtained with no deliberate intervention from the individual. An analyst working on the data would still have to identify how some patterns in the digital footprints are related to selected psychographics. In summary, the process involves the use of unsupervised machine learning algorithms, first to discover patterns and relationships in the data, and second to validate the psychographics as predictors in a supervised algorithm.

Behavioural Design in the Post-Sales Process

The identification of relevant applications of behavioural economics to the insurance industry is not something new.¹⁸ Today, such applications can be enhanced through emerging digital ecosystems. Once a digital user has become a customer, insurers should continuously interact with him by understanding his usage of connected devices and sensor-based technology to unveil unspoken needs. Such interaction will mitigate the risk of losses for the insurer and the insured. Behaviour-led product design might help insurers reduce the risk of fraud and churn. From an insured viewpoint, this would lead to the reduced likelihood of claims through preventative triggers and recommendations. From such learning, insurers can improve pricing strategies deep-diving into customers’ true behavioural patterns.

Because InsurTech relies on the ubiquitous utilization of social media and digital channels, insurers can enhance the results from their behavioural work through the study of computers as persuasive technologies, called captology.¹⁹ These persuasive technologies might also be understood as devices that help users to monitor their risky choices in order to achieve a desired goal. For example, insurers can send timely nudges²⁰ to customers’ smartphones and email addresses to influence the context-based determinants of decision-making and behaviour. The MINDSPACE framework²¹ provides a checklist of potential behaviour change techniques that insurers could potentially test. The options and commitments that result from the work could improve the effectiveness of an insurer’s business model. Despite the heterogeneous nature of psychographic criteria, unique user experiences can be optimized through the delivery of personalized interventions.

We can find examples of ways digital technologies are used to create simpler and automatic customer interactions that improve data collection but also drive compliance (see Table 2). Renter insurance provider Lemonade uses timely pledges and moral reminders to reduce the risk of dishonesty within a peer group before a claim is paid. Daniel Schreiber, co-founder and CEO of Lemonade, talks at length about the behaviour economics that Lemonade uses within its algorithms and web pages. Among many other findings that apply to InsurTech startups, Ariely’s research²² shows that when an individual signs up at the top of a web page, it is likely that the individual will be more honest with the information he or she discloses thereafter.

Aviva, the largest general insurer in the UK, launched in 2012 an app called Aviva Drive that records information on a driver’s journeys with smartphone GPS geolocation. Once a driver has driven and recorded 200 miles of journey, the app gives a score (out of 10) based on the driver’s cornering, braking, and acceleration skills. This information is later used to price car insurance and personalized future discounts. The app provides feedback on driving skills the driver needs to improve, while also utilizing rewards mechanisms through gamification techniques for safe driving, ensuring that the driver visits the app on a regular basis.

Both Lemonade and Aviva provide insightful examples of how InsurTech could exploit the opportunities brought by behavioural design in the development of their solution or platform, particularly during after-sales processes, such as claims prevention and servicing. Such practices help increase brand awareness for those using the techniques on a regular basis and scale up relevant and targeted customer communications.

Some Ethical Concerns

As we would expect, nudges need to be utilized carefully, specifically impersonal nudges, when segments are very heterogeneous or when the audiences cannot take decisions autonomously. If it is essential and feasible for a segment to learn new behaviour through educational mechanisms, it is more effective to produce informative digital content that improves the audience’s competence while monitoring risky behaviour.

Delivering large volumes of personalization triggers might be perceived as invasive and raise questions about privacy and customer manipulation.^23,24 Nudges are supported by the majority of European countries,²⁵ but the US prefers that, in specific cases, these interventions be supported by the display of value-added information.^26,27 These very recent findings highlight the importance of evaluating users’ satisfaction to ensure that the information they receive is not seen as manipulative.

Conclusions

The main idea of this chapter is that insurers should go beyond demographic segmentation and infer individuals’ psychographics through the use of active and passive data collection. When these factors are incorporated into customer analyses, the accuracy of risk assessment is improved and pre- and post-sales engagements become more personalized on an increased scale. As insurers can optimize prices and users can better monitor their behaviour, value-chain interactions are enhanced.

However, even though users can gain benefits from sharing data to access better services and better prices, practitioners should expect an ongoing debate relating to the ethical concerns regarding the use of psychographics through persuasive technologies, and how invasive such technological advancement might become with the wide use of interconnected devices.

Appendix A