"We combine now hundreds of algorithms simultaneously, and the strongest ones are the ones that survive," says Fiehn.
"The more data that we put through, the better they get. That means that whereas before our pricing teams may have been able to run two or three models a month, we can now literally run hundreds a week, which means that the degree of accuracy has improved a hundred-fold."
As an experiment, they pitted the machine against the hundreds of years of experience in the human pricing team to discover which could most accurately predict the prices at which insurance policies were sold based on MoneySuperMarket figures.
The machine took a week to process the data and made predictions that were 30% more accurate on average than the pricing team, which spent a month on the task. In a third of the cases, the automated prediction was exactly the same as the real sale price of the insurance.
Markerstudy can now apply this system to prices offered at the point of quote.
The team also used it to calculate their retention rates, by mapping the renewal price of an insurance policy against the occasions when they raised or reduced the price and assessing how this affected the customer's decision to renew their policy.
"As you'd expect, if the insurance premium goes up, you get less people that renew your policy, and if you drop the price more people keep it," says Fiehn.
"But what is interesting is we found that actually, it doesn't matter how much we reduce the price. The retention rate is about the same. So we can use the algorithms to calculate what the renewal sweet spot is, so that we don't give away more money than we actually need to."
There were 1,792 fatalities and almost 200,000 injuries on the road in 2016, according to the Department for Transport. Markerstudy has produced a digital product called VisionTrack to help reduce these figures.
It sourced a high-definition telematics camera from a manufacturer in Korea that contains a SIM card to transmit information about the driver in real-time.
They then partnered with Microsoft to create an IoT platform that connects VisionTrack with all of Markerstudy's drivers around the world. It can monitor their speeds, provide real-time images, and send text messages to the device that are then automatically converted to audio if they need to talk to the driver. The device also comes with a panic button that the driver can hit if they're in a situation of distress.
"This has had a profound effect on driver behaviour," says Fiehn. "The fact that they know we can see what they can see has really moderated their style."