An estimated 3.2 billion people – nearly half the global population – are now active users of social media worldwide for networking, collaboration and sharing content. The connectivity of this diverse mix of users creates a highly-aware, engaged population that has access to vast amounts of information. This creates both opportunities and risks to consumer-focused services like the insurance industry – especially in times of catastrophes when communication is key to recovery.
In continually-evolving natural, economic and political environments where major risks are often hard to predict, there is increasing responsibility on the insurance industry to be more active in improving the resilience of communities.
Social media has been an effective medium for reaching clients and communities during catastrophes, particularly natural disasters, in the following stages:
Social media plays a key role in improving community preparedness for the impacts of catastrophic events.
One of the most common uses is promoting resilience to natural hazards via social networking platforms. An example is U.S. insurers providing property checklists to highlight mitigation efforts prior to seasonal hazard periods. Encouraging customers to become more resilient via rewards has also been a successful approach for an Australian insurer through offering premium discounts for retrofitting in areas prone to cyclones: a benefit that was promoted via social media. By creating partnerships with leading disaster response agencies and marketing advice via social channels, insurers can build their brands as a trusted resource from the renewal process to during a catastrophe.
Taking a longer-term view in Florida, solutions for sustainability and climate resilience are being crowdsourced via social media to help frame policy ideas for climate strategy groups and energy companies. A similar initiative from Oasis Loss Modelling Framework is aiming to increase the availability of information on catastrophe and climate change risk. It was formed in 2015 to help bring together climate risk tools to assist the insurance and other industries to develop evidence-based climate adaptation planning. Recognising the power of social collaboration, they plan to implement a crowdsourcing function to help organisations find data and services that are not readily available.
The power of social media to help communities in disasters was nowhere more evident than during Hurricane Harvey in 2017. Nextdoor, a U.S. private social media platform for neighbourhoods, broadcast calls of help in rising floodwaters and facilitated rescues from neighbours. There were hundreds of other cases of people using Twitter and Facebook to call for assistance, even publishing their addresses because emergency services could not deal with the volume of requests. There was a case of a mother participating in a live television interview from the roof of her house after she posted messages and videos of her situation.
These examples led Nikki Usher, Associate Professor of Media and Public Affairs at George Washington University, to call Hurricane Harvey “the first major natural disaster of the social media age”. Following the Christchurch earthquake back in 2011, government organisations used Twitter to post critical information, a Google Person Finder was set up to collect information about missing persons and the University of Canterbury used Facebook to organise volunteering and humanitarian aid.
Social media platforms have been critical in distributing information about recovery and reconstruction, and identifying where the major stresses exist to focus aid efforts. There is evidence of communities having higher survival rates and greater resilience where there is active citizenship through strong local engagement – even though they received minimal assistance from government and business. As a result, many not-for-profit organisations focus on building social networks to strengthen communities, particularly in developing economies. Social media platforms play a crucial role in this success where technology permits.
The increasing use of social media for disaster communications has allowed many organisations to take advantage of the high volume of complex information. The Department of Homeland Security in the U.S. developed a social media analysis system called SMART, the Social Media Analytics and Reporting Toolkit. It uses topic extraction, word cluster examination and unusual event detection to provide situational awareness and improve decision-making for time-critical tasks. Insurers have begun using similar social media analytics to identify fraudulent claims activity during events; an example of which is identifying unusually large claims made outside of the known catastrophe footprint.
How can we analyse social media communication for practical insurer insights?
The first step in producing valuable insights from language used in social media is converting it into structured data. Natural Language Processing (NLP) enables computers to understand, analyse and interpret speech and text. Every day examples of NLP technology include autocorrection when writing text messages, pre-populating search engines or filtering spam emails.
To quickly understand large volumes of information, algorithms have been developed to enable natural language processing tasks. These range from simpler exercises such as assessing the frequency of terms and named entity recognition, to the more complex tasks of classification, sentiment analysis and unsupervised modelling of topics.
There are many examples of how the insurance industry is leveraging NLP technology to improve processes and develop products, including:
- ‘Bots’ that use artificial intelligence, which are increasing in popularity, to process claims, detect fraudulent activity and communicate directly with policyholders and potential customers.
- Profiling of customers and their web interactions to identify preferences and target product marketing.
- Web-scraping algorithms that mine textual data from the web, which are then integrated into pricing models and underwriting.
Advances in image recognition algorithms can also help classify natural catastrophe claims by validating the presence of meteorology or damage in posted images. Digital claims agents are also leveraging image recognition technology to authenticate damage to properties and improve the efficiency of loss assessment processes.
Case Study: Predicting Severity of Hail in Australia
The severity of hail events is particularly hard to determine because meteorological footprints often cover large urban areas. To help predict the severity of these catastrophic events, Aon has applied natural language processing to social media data for historical hail events in Australia.
The foundation of this predictive modelling is the language communicated by people on the ground experiencing the event. In extreme situations, people naturally use more descriptive language that can reveal valuable information about an event. By extracting descriptive information such as “tennis-ball” or “smashed window”, we can quickly determine there is damaging hail.
Combining this with geolocated posts or communicated location detail, event footprints can be refined and help insurers plan for claims response. Predictive classification algorithms can then be trained on this historical data and used in real time to predict future hail event severity – both in Australia and globally in hail-prone countries with a strong social network presence.
This approach has been successfully used to enhance understanding of other types of natural catastrophes. Investigations into Twitter activity during Hurricane Sandy in 2012 verified strong correlation with economic damage and the physical effects identified from insurance claims. This highlights the opportunity for insurers to identify areas of major claims activity in real-time as part of their recovery planning.
Where do opportunities lie?
- Becoming a reliable advisor in disaster situations can greatly build trust with customers. Insurers would benefit from partnering with crisis and technology experts to deliver reliable disaster response services to consumers via social media. Impact Forecasting, Aon’s catastrophe model developers, uses data from agencies such as NOAA, USGS, Japan Meteorological Agency, UK Met Office and Australia's Bureau of Meteorology to deliver catastrophe alerts with detailed, real-time disaster information.
- Detecting fraudulent activity and estimating the severity of natural catastrophes from social media discussions can also greatly enhance disaster response analytics through improving accuracy and speed of service.
- Beyond disasters, the insurance industry is increasingly realizing the potential of coupling social media data with technological innovations to advance other areas of their business. With operational efficiency being the core characteristic of many successful InsurTech firms, advanced language processing techniques are used to enhance claims handling, achieve questionless underwriting and form the foundation of artificially intelligent insurance agents.
- Social media interactions have also started influencing insurance policies. Predictive algorithms have been developed to identify a customer’s risk tolerance and classify whether their behaviours warrant an adjustment in policy terms at renewal. As we become even more connected through devices monitoring more of our actions – the rise in Internet of Things – online behaviours, such as posting photos online that reveal home security weaknesses or careless behaviour while travelling, could trigger social media-related exclusions in the future.
This high connectivity of our global community means the insurance industry needs to mitigate these evolving risks and capture the opportunities. Insurers must ensure disaster response, underwriting, marketing and risk management practices continue to adapt in line with changing technology and human behaviour. Embracing analytical and technological advances, implementing social media strategies and innovating risk management practices are important actions in helping to realise the benefits from social media.
About the Author
Tom Croshaw is a Senior Analyst at Aon’s Reinsurance Solutions within its Analytics team. Based in Sydney, Australia, his role focuses on delivering research projects and developing risk management products for the Australia and New Zealand region. Tom’s expertise is in predictive analytics, natural hazards, financial modelling and GIS analytics which he uses to deliver risk solutions to clients.