Our Chief Information Officer, John Davison, talks to CIO UK about emerging data science techniques to augment traditional forms of risk assessment.
Insurers have used data analysis since ancient Greek and Roman merchants would purchase loans for their voyages at rates based on the risk of storms and sinking, but in recent years more sophisticated methods have entered into their equations.
First Central Group CIO John Davison applies these modern techniques to augment the standard underwriting process.
"The way the business works is you have an underwriting assessment of a risk and you perform a series of calculations. When they're different, we have a team of people who do statistical modelling. We're using analytics and AI for that process," Davison tells CIO UK.
"We continue to perform the standard, actuarial calculations that are done and the modelling that's done, and we then augment that with data enrichment machine learning to help us pick the right models."
This blend of traditional actuarial theory and modern analytical technique fits the business model of First Central Group, which functions as both a motor insurance broker and a technology organisation built on digital foundations. The company was founded in 2008 to exploit the insurance opportunities that were emerging from the boom in price comparison sites.
Davison joined three years later, initially as Group Chief Architect, before progressing to the role of Chief Technology Officer. In January 2017, he was appointed CIO, responsible for the company's data and IT systems. He also helms the organisation's change strategy, which centres on the goal of having no legacy IT.
To achieve this objective, First Central Group produces its own bespoke technology products that it can continually develop, such as Focus, a retail insurance platform that can capture and analyse data from quotes, bookings and claims in real time.
"It helps us have no legacy, because we're continually changing it," says Davison. "We invest a very considerable sum each year in change, and we're already this year on our 13th major service pack."
Every system in the business is assessed each year and anything that is underperforming is promptly replaced. Last year, Davison brought in a new replacement claims platform after deciding the previous system was not up to the task.
"There's something to be learned from the InsurTech community," he says. "Those businesses start because there's a need not being fulfilled. The way First Central views that is there's a suite of opportunities in our market that other people are spotting that we think we're very well-placed to take advantage of because we have got that heritage of great underpinning technologies."
First Central Group processes a huge volume of information to identify the context of a transaction and react to user behaviour.
It analyses insurance policies after they are bought and what led to the purchase from the moment that the customer landed on the website.
"It's trying to proactively understand why a customer is contacting you and how you can better service them," says Davison.
"The approach that we take is to determine sentiment by textual analysis, so understanding the relationship between the words that are in the web chat and the corresponding outcome. We can attach those two things together so that you are learning as you go."
This analysis can help improve the website by identifying the information that customers struggle to find online or areas where the service isn't working, such as IDR telephone systems.
The unstructured data from web chats and phone call transcriptions is mixed with the structured information accumulated from policies, customers and identifiable transactions. This is then mixed with a deep data lake. Machine learning algorithms then dive in to find the patterns that reveal how each segment of the business is performing.
"It's almost akin to an online retail approach," says Davison. "When we have a brand or we have a particular offer that we want to progress, we can use some of our testing in real time to understand if that's working."
To separate the valuable data from the trivial information is no easy task.
"The approach that we take involves what we call certified information, which is information that we need to operationally run the business or supply to the insurance company, and then we have everything else.
"And the 'everything else approach' is why we have a data lake, not a database. It's that principle that means it can be both structured and unstructured, and available for analysis at a later date so we can identify trends or patterns in information that we had maybe suggested wasn't useful or was duplicated. Naturally, by capturing it all, you have the ability to look backwards and decide whether it is now useful and if it wasn't before."
A recent report by Accenture revealed that more than three-quarters of insurance executives believe that AI will significantly alter or completely transform their industry in the next three years, but many of them struggle to grasp the practical uses.
Davison gains widespread faith in data science by demonstrating its value, whether it's underwriting profits or making contact centres more efficient.
"That value can be more than just financial value," he says. "It can help the employees' working life and it can make our customers happier, so our NPS scores [Net Promoter Scores] can go up. All of these things are barometers and fundamentally it's about being able to derive value from the information and demonstrate that that value exists."
He ultimately views AI as just one of the tools that can allow the insurance sector to deliver better products.
"I think the key deliverable for AI is about being able to be more accurate in the way that you price and the way that you deliver products. AI machine learning, data science, data lakes, are all tools that can help you do that. And I guess FCG's view is that that tool will become part of our core business, and will become part of our DNA going forward."
Davison's objective for the last year was bringing the same data science capabilities used when underwriting premiums to the wider business.
Now that has been achieved he's focusing on expanding First Central Group's market by investigating the demographics of people who are buying insurance dealing and the more general trends in mobility such as the rise of shared vehicle usage.
He's also keeping his eye on a range of emerging technologies and techniques, from blockchain to non-linear modelling.
"We'll keep looking at changes as they come up," he says. "We have a considerable change budget that allows us to stay current and take advantage of those things as they become mainstream."