Shoot the messenger: Lessons from Trump’s data-driven election victory for the insurance industry
February 8, 2017 Chris Sandilands
I recently read an article about the use of data in Trump’s election victory. It was fascinating and terrifying in equal measure – and potentially has some lessons for the insurance industry and its regulators.
First, here’s a quick summary of the article.
The journalist argues that “Trump’s striking inconsistencies … and the resulting array of contradictory messages [were] his greatest asset.” This allowed “election management companies” to deliver a “different message for each voter”.
This is because election management companies use targeted digital marketing to deliver that message. As the CEO of the election management company in question notes, it is a “ridiculous idea” that all members of any one demographic segment should get the same message. Instead, these companies craft messages that resonate with individuals. Big data analysis has allowed them to build what the author describes as a “human search engine”.
This search engine is based on the analysis of huge amounts of data. This means that the “the personality of every adult in the United States [has been profiled]”. The analysis is based on a psychometric framework called “OCEAN”, which measures someone’s openness, conscientiousness, extroversion, agreeableness, neuroticism. This framework has existed for some time, but its limiting factor was data collection. Researchers require rich data on any individual to populate it and it was difficult to get large samples. This problem was solved with the rise of social media: those personality tests that float around Facebook are not as innocent as you think.
But personality tests are just one of the inputs. Other inputs include Facebook likes and third party data. In the article we discover that “liking” Wu-Tang Clan is a strong indicator of heterosexuality. No input by itself is not definitive – but in combination they are.
In aggregate, the analysis allows election campaigns to tell people what they want to hear – either positive messages about their candidate or negative ones about their opponents. In other words, the holistic view of Trump paints the picture of a confused, “unelectable” candidate, but this is irrelevant. The average, disengaged voter, sees a highly curated, favourable candidate.
The insurance angle
The most obvious read-across observation is that insurers and brokers are amateurs when it comes to the targeting of messages. A datapoint for this is the fact that companies like Bought By Many (link to Bitesize profile) and Digital Fineprint (Bitesize profile) exist. Both companies’ raison d’être is their superior – and therefore cheaper – acquisition of insurance customers. Big insurers buy clicks for “car insurance”; Bought By Many buys clicks for “car insurance for cancer sufferers”.
Insurers contend that they are too big to worry about niches. Historically this may have been true because it was hard and expensive to reach those niches, but in the context of US elections it seems like “old thinking”. If campaign management companies can work out what to talk to any single voter about on their doorstep, then a motor insurer should be able to target their ads more closely. Interestingly, companies like Visual DNA, already sell psychometrically-profiled digital audiences – note their a-typical landing page.
My second conclusion is that power has shifted from the “creator” of news to its messenger. Many, including the academic who pioneered the use of the OCEAN model, argue that this is “manipulative”. But could pioneering or unscrupulous (depending on your view) insurers or brokers shift markets from such practices? Could a targeted social media campaign see (fake?) negative stories being circulated about competitors along with (fake?) positive stories from one’s own customers? It would surely be a powerful customer acquisition model and cause incumbents a major headache.
A small and arguably cheeky rather than unscrupulous example of this practice might be Lemonade (Bitesize profile and update), which trades on the notion that “traditional insurance companies make money by keeping the money they don’t pay out in claims.” We comment in our Bitesize profile that we question the legitimacy of this claim, and we think the “Ministry of Information” (more on that below) might want to take a look at it. There is also a case study on the B2B side – consider the impact on Deutsche Bank of a PR campaign against it by hedge funds who held a short position.
This leads onto the third conclusion. There is no Ministry of Information in the western world. This allows the distribution of distorted or even fake news. In the insurance world there is also no such Ministry, but there is something close: the regulator. Often data is discussed in the context of underwriting and pricing in regulatory circles, but it is clear that much of the “emerging risk” could come from marketing approaches. The industry needs to ensure there are “rules of play” and the regulator needs to have effective sanctions against those who eschew them.
The insurance industry needs to confront the idea that it, like all other parts of society, could be beneficiaries or victims of highly targeted marketing campaigns which redefine the rules of play. “Fair play” rules need to be in place and the regulator needs to patrol those who don’t follow them.