Rise of Social Analytics in the Auto Insurance Industry

Background: Increasingly, companies are starting to find value in social media data, including Twitter, Facebook, Reddit and other social data sources by employing robust filtering techniques and machine learning algorithms to uncover the signal within the noise. Then, they often correlate financial data with social media conversations to add value to their business. This is especially true given the recent access to social media data sources through Application Programming Interfaces, as well as options to pay for secure data access. This trend can be seen applied by one of the largest insurance companies in Britain, Admiral, as they attempt to use Facebook data to price their auto insurance.

Action Steps:

Admiral insurance plans to analyze the Facebook accounts of first-time car owners searching for personality attributes that may be suggestive of safe driving. For instance, those individuals who would be identified as well-organized, diligent and meticulous may have better scores. The company plans to look at posts, as well as likes of Facebook users, however they will not use photographs. An example of how they plan to generate a “safety score” is by parsing through an individual’s posts to see how frequently they have said the word “always” and “never” which may be indicators of over-confidence, as compared to the word “maybe”, which may suggest a more cautious individual. The name of this product is going to be “first car quote” as it would help first-time car drivers the option to receive a discount on their car insurance (~$400) by opting in to share their Facebook data. In this instance, the algorithm to identify the safety of a driver would not be based on a specific model but instead on thousands of combinations of likes, and phrases which would continually adapt and learn as the data source grows. Since there would be no response or success variable to begin with, the company would start with an unsupervised learning approach and as it obtains more data move on to a supervised learning approach. Advanced natural language processing algorithms will also be part of the detection process.

Privacy Discussion: At initial glance the move by Admiral may seem like an invasion of privacy. However, there are two key aspects of the deal that must be considered. First, each consumer would have the ability to voluntarily opt-in to such a data share agreement. Second, such data would only give first-time drivers the option to receive a discount on their car insurance, and not a price hike. Thus, the data share would provide the insurance applicant a direct incentive to share the data.

Implications: Currently, more than 30 billion content pieces are shared on Facebook each month, over 24 hours of videos are uploaded on YouTube each minute, and the average tweets per day in 2010 were 110 million. It is clear that there is a massive explosion in social media data, and companies are engaging with consumers through various social media platforms. Various companies across industries (e.g. auto insurance) are attempting to find value in such social media data and try to find correlations with financial trends. The challenge that still remains is appropriately filtering the social media data to only capture significant signal and less of the noise. Moreover, there is no clear process to obtain continuous value from such data. However, some companies like Admiral are testing out new and innovative products to uncover value for the insurance industry.

 

Works cited:-

1. https://www.theguardian.com/technology/2016/nov/02/admiral-to-price-car-insurance-based-on-facebook-posts
2. https://medium.com/privacy-international/social-media-intelligence-and-profiling-in-the-insurance-industry-4958fd11f86f