To catch a glimpse of the future for our industry, we turn to an unlikely place: the 2016 U.S. presidential elections. In the months leading up to the election and right up until the eve of 8 November, the New York Times predicted that Hillary Clinton was the clear favourite to win the presidency. On the day that Trump claimed success, Cambridge Analytica released a statement to the press in which its CEO, Alexander Nix, is quoted as saying: “We are thrilled that our revolutionary approach to data-driven communications played such an integral part in President-elect Donald Trump's extraordinary win.” Although this bold claim has since been challenged, we still must ask: What did a little-known British company purport to do to influence the result of such a significant event? And what does this mean for the insurance industry?
The answer to these questions lies at the intersection of big data, analytics and behavioural science. Drawing inspiration from our counterparts in politics, let’s examine the substance behind the buzzwords and explore the potential future that these disciplines unlock.
More and more, everything we do is being recorded and stored digitally. This expanding digital footprint is what we think of as big data. Consider for a moment the kind of information that resides in the cloud for each of us, and what picture this might paint if it could be pulled together in a sensible way. From government databases, to medical records, transactional accounts, club memberships, loyalty cards, social media platforms, Google searches, streaming subscriptions, and smart devices: the list goes on and on, and the picture becomes startlingly clear.
Targeted marketing is not a new idea, but social media marketers have evolved this concept far beyond basic demographic stratification. Facebook offers marketers the ability to hone in on very specific characteristics of their membership, all within the bounds of confidentiality by using predictive models that make clever use of basic information shared on the platform. Have you noticed the kinds of banner advertisements that pop up on your Facebook timeline and how these differ from what pops up for family, friends and colleagues? This kind of data-driven, tailored messaging is not unique to Facebook. Search engines like Google are masters at ensuring that each of us sees search results which are tailored to our particular needs and interests. And this is just scratching the surface.
In 2013 Michal Kosinski, a psychologist and data scientist, published a paper entitled “Private traits and attributes are predictable from digital records of human behavior.” He and co-authors David Stillwell and Thore Graepel presented groundbreaking findings from their research to assess personality from liked, shared and posted information on Facebook. The results range from amusing (followers of Lady Gaga tended to be extroverts, whereas those that “like” philosophy tended to be introverts) to highly sensitive predictions about traits like sexual orientation, religious views and intelligence. The following results provide a practical appreciation for the predictive power of their work:
- With just 10 Facebook “likes” Kosinski and his team believe that they would know you better than your work colleagues.
- With 100, better than your friends and family.
- With 250, better than your spouse or partner.
- With over 300 likes, they may even know you better than you know yourself.
As astonishing as they are, these results are based purely on one, not particularly intimate, piece of information. The real power of Kosinski’s work is the ability to pull together hundreds, if not thousands, of individual data points from a range of sources to produce reliable predictions about a person’s personality.
It would be remiss to discuss future applications of big data without touching on the enabling capability of analytics. And yet, analytics is becoming a buzzword of the past. It has morphed into something more flexible and robust – something that can handle the ever-growing extent of the cloud and incorporate a continuous flow of data from a range of different sources, including pictures, voice and more.
The future of analytics is cognitive computing.
Cognitive computing represents the convergence of a variety of emerging disciplines including machine learning, artificial intelligence and natural language processing. As futuristic as it sounds, there are many examples of cognitive computing systems that we already regularly engage with, including search engines like Google, speech recognition apps like SIRI, and chatbots on social medial platforms. The cutting-edge developments in this space are being driven by organisations such as Google DeepMind, with its inspirational ambition to “Solve intelligence. Use it to make the world a better place.” In 1996, we were amazed to see IBM’s Deep Blue beat a world champion at chess, a game of logic. In January 2017, Google DeepMind’s AlphaGo achieved 60 straight wins against top international players at Go, a game of intuition. Not only have we developed the capability to compute huge amounts of structured and unstructured data in real time, but we are very nearly touching a future where systems can make complex and astute decisions better than we can.
Let’s bring this back to the 2016 U.S. presidential election. Politicians have been making use of data and voter profiling for many years. Cambridge Analytica claims to have elevated the art of profiling to a new level by incorporating research aligned to Kosinski’s work and incorporating behavioural science techniques. Not only can they target voters according to demographic characteristics, but they claim to have mapped the personality of every single voting American according to the Five-Factor Model of Personality (Extraversion, Agreeableness, Conscientiousness, Neuroticism and Openness to Experience). By understanding at a granular level how each individuals thinks, feels and behaves – and positioning political messages appropriately according to proven behavioural principles – they believe that they are able to influence votes more effectively than ever before. Consider the example of stimulating support for the Second Amendment right to bear arms. For individuals that demonstrate high levels of Neuroticism and Conscientiousness, in order to support the right to bear arms, Cambridge Analytica prepared a message positioned on fear, protecting oneself and taking control. For individuals who were found to be Agreeable with low levels of Openness, an optimal outcome was achieved by positioning the message on compassion and tradition.
So what does this mean for us? As noted above, many have questioned how influential Cambridge Analytica truly was in the recent election. At the very least, its story has highlighted a tangible truth: mass marketing is dead (or dying). More and more, all of us will experience customised messaging, products and services – and soon our insurance customers will come to expect this from us. Are we ready?