We interrupt this business…

…to bring you a message from our data-Stig.

He says: “My mission in this series is to help you – the insurers, underwriting management agencies and intermediaries, understand how to benefit from the data wave. It is a happening-right-now, competitive advantage for insurance. And the people closest to the policyholder are those that have the biggest opportunity – especially in the South African specialist and commercial insurance space. All it takes is a bit of foresight and collaboration within the insurance eco-system!” Ask yourself three simple questions:

  • What am I in the dark about? What do I not know about my clients and their businesses (as a collective) right now?
  • If I did know this, how would I manage their risk better?
  • How much money could I make or fail to make? What is the cost of not having that information?

Okay data-Stig. I think that means ‘get your act together like yesterday’. But can you give us some basic definitions– like buzzwords – that will make us feel a bit more on top of the data game? It does help if you know the language right?

“Yes,” he says. But then swiftly turns the tables on us, asking us to explain the data buzzwords first.

This is only mildly less embarrassing than pretending to a bunch of eleven-year-olds that we know what ‘flossing’ and a ‘dab’ means.

Well, here goes…

So, what are data sources?

Okay. This one I got. It is simple – all the possible sources of information available!

‘Yes’, he smiles, ‘but today this means accessing new sources of information – like apps that feedback behavioural data from insurance policyholders and installing connected devices in the homes of the insured – the equivalent of FitBits for cars, web-based fire alarms for houses and web-based monitoring devices for equipment maintenance – just to name a few!’

Obviously, this will also have benefits for claims. But it’s about predicting risk and modifying client behaviour as a risk management strategy. It’s about knowing where possible data points exist and cleverly incentivising policyholders to stay connected to the tech. Banking and insurance are already doing this widely in the personal lines space. The payoff for the specialist and commercial insurance space will be even bigger.

Right! Next question. What is machine learning?

Let me think! Logically, there are two options. Either it means me learning through machines like webinars OR the machine (computer) processing data and providing insights. I’m going with the second one?

‘Not bad but not quite’, says the Stig.

Machine learning is the use of algorithms that self-learn over time. Key word: self-learn! Therefore the machine self-improves its accuracy and predictive ability as more data is received over time i.e. the machine gets cleverer. How clever? Well, at this stage, it’s up to a point i.e. some machines can reach human equivalent capability but not exceed it, right now.

Applications for Insurance?

Of course there are lots of applications for insurance. One example is using machine learning to insure the logistics and transport industry. Machine learning allows the development of applications to assist with everything from driver behaviour modification to limiting cargo loss severity. Predictive analysis on transit types, product categories, and shipping destinations can significantly reduce risk and claims for insurance carriers.

Just to give you an interesting perspective. Progressive, an insurance company based in the United States, collected 14 billion miles of driving data in seven years from policy-holders signed up to its usage-based car insurance product. The result of this degree of data analysis means that driving style is almost as uniquely identifiable and predictable as a fingerprint. Can you imagine the power of that volume of data?

What is Deep Learning?

Um. No! I give up. It sounds scarily obvious – too obvious?

Okay. So I wouldn’t get too way-led by the concept of machine learning. It has enormous value but has also become a buzzword in broader business. The fact is that whilst it is a very clever model, It is one of a range of different types of data modelling that can be used. We should always consider what model is fit-for-purpose!

Deep learning usually involves simultaneous image, text and audio processing. The model is inspired by biological neural networks. Think neurons firing in the brain i.e. how the powerful human brain can simultaneously process similar varied data inputs.

Applications for insurance are continually evolving – as is the modelling capacity. Think about the use of biometrics in the banking sector and some forms of facial recognition and voice recognition?

Aside from streamlining the customer experience (especially for personal lines insurance) – there are also applications in claims, fraud detection, etc.

Deep learning is also enabling new product development in areas like parametric insurance (which does not indemnify the pure loss, but agrees to make a payment upon the occurrence of a triggering event such as weather events, delays in travel, etc).

Now we have a question for you data-Stig. What is Python? It sounds cool!

A python is a snake. (Smiles)! Python is a programming language. It’s been available since 1994. We (data scientists) like it for its ease of use. It also has significant ‘stretch’ for advanced users, especially if you access specialized libraries such as those designed for machine learning and graph generation.

Right! So next time I meet a data scientist in a bar – I can say: “do you code in Python?”

The Stig gives that silent eye-roll. So I give him the last word.

He makes a surprising comment: “The best business results from data science is when data scientists, actuaries and insurance specialists – in fact the entire insurance value chain – collaborates on a problem. Machine learning is enhanced by collaborative thinking.”

For the previous article in this series, click here: http://www.compass.co.za/2019/05/the-lifeblood-of-insurance-comes-up-against-the-data-stig.html