We've recently acquired our version of The Stig at Compass Insure. A secretive, if personable, figure brought in to help us navigate the matrix of data science.
This is a field that has shaken insurance models globally and we are told, will continue to impact our business in myriad ways. Yet for many of us, it is still shrouded in complexity.
Therefore, since we can't see our data-Stig roll his eyes, we're asking him some dumb questions about data science.
The first question being whether it's really worth all that hype? After all, data has been the lifeblood of insurance modelling for the better part of the last century, right?
'Sure', says the data Stig in a neutral, non-committal voice. (I think I detect an invisible eye-roll. I get this repeat feeling throughout the interview!).
"Did you know", says our secret weapon, 'the insurance industry invests heavily in and developing data science models and processes internationally. Forecasts show that spend on digitisation by the big European insurers will reach over $11bn by 2020. About $2.5 billion of that will be spent on data analytics?'
No, we didn't know that - the insurers must be expecting a big return on that investment right!
Okay, so it's a big deal. But what is data science? It seems to have a language of its own and that language changes depending on who you speak to?
Very simply, data science is the use of various tools (hardware and software), algorithms and something called machine learning to discover hidden patterns and extract useful insights from masses of raw data.
Just so you know, machine learning refers to computer programmes that can access data and use this data to learn automatically without human intervention, adjusting their actions as a result.
That's a top-line view. But yes, be warned, a quick search of the internet will give you about as many slightly varying definitions as there are bytes.
Processing data to predict outcomes? How is that different from old-style data management systems?
The difference in 2019 is to capitalise on the humongous volumes of data. There is just an unprecedented amount of information, to be interpreted at scale, for a business outcome. To do this, we need the computing power of algorithms and machine learning.
Also, in the past, insurance companies developed models based on structured data. Structured data is the kind that the insurance industry usually works with – the information that you get when someone applies for a policy, for example!
But now there's a world of interesting, random unstructured data from new and emerging sources. Think everything from fitbit to geo-location services. Think streaming sites, social media platforms, mobile applications and Internet of Things technologies, like city traffic systems. Think global shared databases.
And that's just the start. The real payload for insurance is probably in predictive analysis capabilities – but that's for another article.
In short, because data science can 'sync' so many variables from such a wide variety of sources (structured and unstructured) – we can look at the data set from many angles, sometimes angles not even known to us yet!
Can you give us a quick list of where and how data analysis impacts insurance?
Sure, a quick list. Seven key areas immediately come to mind...
|•||Customer Segmentation (for marketing purposes)|
|•||Oversight of key risk and compliance indicators for insurers (KYC, etc)|
And that's just the short list...
Are we getting it in SA? Or are there challenges?
I think most insurers understand the impact of data science. And there's some innovative processes that are happening. But the deeper insurance eco-system needs to understand how this applies in their business. There's so much hype – that smaller companies think this is out of their reach. Yet there are myriad ways in which data science can add value – even without accumulating big data. The payoff – especially for specialist insurance and those who write complex commercial risk – could be huge.
Data is, and always will be, the lifeblood of the insurance industry. The challenge in South Africa is to put in place mechanisms to harvest data from the market, to understand the type and quality of data required and to access the right data at the right time.
What's the high dream for data and insurance?
The high dream is for insurers to embrace data to create a customer-centric insurance model that is real-time, predictive and flexible.
For example: There is talk of Internet-scale "data listening" that aggregates and updates petabytes of data to build risk models in the very near future. It's already happening in pockets. We can access data from public and private sources, from connected devices, from control systems, and even content such as news reports of power outages or data breaches. Can you imagine being able to process all that data in an instant to balance a risk portfolio or even, make an underwriting decision. That's already possible.
Ummmm…no, can't quite imagine it. But we hear the need! Last question. Are data scientists cleverer than actuaries?
"Of course not," he says in that inscrutable drawl!