How Actuaries Became Data Scientists

Tell someone at a dinner party you’re an actuary, and you may as well spike their salad with chloroform. Tell them you’re a data scientist, though, and they’ll be lapping up every word you say. Which is strange, because the two professions have much in common, especially when it comes to Insurtech.

Actuarial work was, in fact, the very first data science task to be carried out. It’s the magic bean that grew into the beanstalk of data science. Ever since the first ships were insured in 14th Century Italy, actuaries have been trying to assess risk by looking at data. How did they do it? I have no idea. I’m not a historian! But I do know how they do it today. And I’d be willing to wager that in many ways it’s very much the same as it was back then.

Although insurance is one of the world’s oldest professions, technology and data have, in recent years, begun to have an impact. Marine insurance, though, has – with few exceptions – stayed stubbornly the same. Now we’re starting to change it.

To learn how, imagine two drivers, let’s call them Sarah and Zara. They’re both lawyers aged 35 without any history of accidents or trouble with the law. Sarah travels 50km to work each day, accelerates and brakes aggressively, and has a tendency to tailgate anyone who overtakes her. Zara is calmness personified; she gets the bus to work, and only drives to the shops and back at weekends; she sticks to the speed limit and has never hurled a word in anger at another driver. If they have the same model and make of car, of the same vintage, and if they each have a history of no claims, they will pay the same for their insurance. That’s because the insurer doesn’t know that although their profiles are similar, they drive in very different ways. The same applies to ships – just with a few more zeroes at stake.

But at Windward, we know a lot about how ships operate. We know how the ship sails, where and when. We know if it favors entering crowded ports at night; if it sails faster than most; if it often sails through stormy weather; what oceans, ports and rivers it tends to frequent and a whole bunch of other features whose inclusion here would make this blog longer than a supertanker.

In short, we have a lot of data per vessel. A lot! And because there’s so much of it, and because there are 100,000-plus vessels plying the world’s waterways, we use machine learning to model the data and create a prediction of vessel risk. This is nothing short of revolutionary. Until now marine insurers – and even sometimes vessel owners – didn’t really know where and how their vessels were sailed.

The immediate outcome of such risk modeling is to assist underwriters in adjusting premiums, and, with the help of Windward Insurance, to enable vessel-owners to prevent or mitigate losses. Imagine what would happen if Sarah’s insurance company told her she had to pay double the premium to account for her risky driving? It would either be less exposed to her potential losses, or she’d moderate her driving to reduce her premium. In marine insurance, the same will apply, ultimately leading to safer seas and lives saved. You’re welcome!

Yonit Hoffman, Windward Data Scientist