Predictive analytics is fueling an insurance revolution

Fast forward to the present and Jean’s role as head of analytics I&A at Collinson allows her to leverage her broad knowledge in insurance and technical skills to drive the machine learning, AI, and big data agenda forward at Collinson.

Her aim? To “transform the data landscape, and to maximise the utilisation of the data assets for I&A.”

Volume and value

As the volume of data increases, Jean believes the personalisation will become ever more granular.

“There is now a vast array of data – historical; real-time; social; data from wearables and machines, and increasingly, data not just about the customer, but the specific context they are in. Insurance providers are able to gather more information about individual customers than ever before, allowing the development of more specific, tailored insurance offerings.”

She offers an example of an ordinary online travel insurance offer provided within an airline booking path. Behind the scenes APIs, she says, allow data to be shared and certain trip characteristics to come together to create a persona-based product that takes into account details such as past conversion and transactional experience.

“Not only does it enable travel insurers to design a more suitable product for airlines, it also presents an opportunity to show dynamic imagery and messaging tailored to a customer’s trip which enhances the overall booking experience, makes the insurance offering much more engaging and increases the chance of conversion.”

Fighting fraud

Jean says that while the techniques for data modelling can vary from company to company, most are evolving rapidly towards deep learning and its application to a range of classic insurance problems. Neural networks and machine learning models, for instance, have proven particularly powerful for fraud identification, classification and prediction of fraud likelihood.

“For example, data scientists in my team at Collinson have been working in a graphical database that allows the fraud department to visualise claims relationships by different parameters and identify a fraud ring. This model and algorithm can be used over any claims data and has the potential to bring benefit for multiple product lines.”

“Deep learning has also started to benefit the insurance industry by tackling problems in call centres with the application of natural language processing, and increased security in apps by implementing facial recognition.”

The key is gathering enough data (not any data, but quality raw data) to build AI that is tailored to each product’s needs. “For example, data from connected devices can help predict and prevent accidents in telematics operated cars or self-driving cars,” says Jean.

Future trends

Regarding the future of business insurance predictive analytics, Jean cites an article published by Mckinsey that outlines four key future trends.

“The explosion of data from connected devices that pair with sensors in equipment will give a very good penetration to a multitude of life circumstances and essentially huge amounts of data. This will help insurers understand their customers better, the type the risk they are taking and even changes in behaviour that are likely to result in increased risk.”

Another trend is the increased prevalence of physical robotics: how the insurance industry will transform core products and coverage to insure a growing number of robotic-based operations/systems and companies.

With all this surveillance, it’s imperative that insurance customers receive tangible benefits, such as risk assessments that genuinely improve safety. “I would love for us to be able to take customers through their everyday life knowing that their insurer has their best interests at heart,” she says.

“For example, by having a variety of IoT devices connected to a big data account, we could know how certain activities might increase the risk for our customers and be able to advise them on the best ways to reduce this risk, offer them better coverage and reward them for their loyalty.”