In 2018, home insurance provider Homeprotect reached a huge milestone by completing the first version of an in-house data science platform designed around the skills of its top IT talent, breaking from the limitations of off-the-shelf software. It was the result of chief technology officer (CTO) Dan Huddart’s freedom to innovate, after swapping the large corporate insurance world for a scaleup business.
During an 18-month period, the core system – a data science platform known as Cortex – was built to go beyond calculating premiums for insurance products and is now transforming how Homeprotect’s business operates in key areas.
Cortex is the result of the thinking brought into Homeprotect when Huddart joined as CTO in 2015. Reporting directly to the CEO, he arrived just as the home insurance specialist received a cash injection following a private equity takeover, which brought with it an ambition to grow.
This ambition, along with the company’s profile, matched Huddart’s career cravings, after building significant “big corporate” experience in the insurance sector.
“You know what big corporates are like,” says Huddart. “I had a lot of experience at that point, with lots of ideas and energy, and I just wanted to get into a smaller business where I could get on and do things, and be accountable for things whether they worked or not.”
It was 2011 when Huddart first took a job in the insurance sector, joining RSA Insurance at a time when it was “investing a lot in digital platforms”.
At the time, RSA was expanding and acquiring insurance companies internationally, which also gave Huddart experience working abroad in Scandinavia and Poland. During this time, he got to work on projects internationally and then return to the UK when complete.
He spent six years at RSA, including two years working on a core system overhaul.
After meeting the Homeprotect team and its private equity owners, he was sold on “a small, but really exciting” journey.
At the time, Homeprotect had just over 100,000 customers, with six IT staff out of about 60 people overall, but its owners were ready to invest in growth. “It was a scaleup challenge to take the business and make it a mainstream insurer,” says Huddart.
Homeprotect was a specialist in non-standard insurance, offering customers something they couldn’t get online from most insurers.
When he started his role, the main challenge was scaling the IT team, says Huddart. “There are two challenges for small businesses. The first is you need to automate to scale stuff, because otherwise you won’t be able to deal with more customers and staff. As a result, we had to invest a lot in enterprise technology for staff and automation.
“You know what big corporates are like. I wanted to get into a smaller business where I could get on and do things, and be accountable for things whether they worked or not”
Dan Huddart, Homeprotect
“Secondly, you have to get more value-add stuff in any given week than your competitors, otherwise you are not going to outcompete them,” adds Huddart. “We are in a market where technology really drives most of the difference between the customer experience of one insurer and another.”
This includes work on pricing algorithms, websites and databases, which Huddart describes as “the life and death between you and the competition”.
“Getting more done was the challenge,” he adds.
Growth ambitions
Today, the IT team comprises about 18 people, and the overall company has more than 100 staff, with another 110 working for it at an outsourced call centre.
It was in 2016, a year after he joined, that Huddart had an opportunity to make a huge impact on the company’s growth story when it needed a new pricing engine, known in the sector as a rating engine.
By then, pricing was becoming more sophisticated and data science was emerging in the sector, says Huddart.
“Because we had been insuring the whole market, including non-standard, for about 10 years, we had been collecting much richer data on customers than our competitors,” he says. “When data science came along, it was a bigger opportunity for us than for your average insurer, because as a non-standard insurance provider we ask customers more questions, therefore we have more data.”
Pricing has become more sophisticated because insurers ask more questions, usually through intermediaries, and there is more external data for insurers to use.
“The arms race of accurately pricing people has become more advanced,” says Huddart. “In the past, algorithms would be used to calculate prices. Now, you have to be quite an advanced data science company, and the bar to compete is much higher.”
Tech for talent
To become competitive, Huddart looked at the pricing engines available in the market and set out looking for top data science talent.
“When we tried to hire the best pricing people available, we found that there are quite a few data science and pricing people who are really talented, but not just in insurance,” he says.
Talking to candidates, he found that one of the problems with insurance pricing systems is that they “are so far behind other industries”.
Huddart says buying an off-the-shelf insurance pricing system would limit the company, like all the other insurers, and the people it wanted to hire wouldn’t want to work on technology they see as behind the market.
This prompted Huddart and his team to build their own platform that can keep up with “talented staff” rather than buy something that falls behind the market.
“We asked our team to come up with a platform that they felt would be attractive enough to bring in the best minds,” he says.
This was the birth of the Cortex platform, which Huddart describes as an intelligent platform designed around the company’s people. It uses open source technology components, although there are parts that the company has written which are not open source.
“There were common trends, such as for messaging [where] we saw companies like LinkedIn using Kafka, which we thought was probably the most competitive messaging system,” he says. “We also saw lot of companies using Kubernetes and placed a bet on it for our compute power.”
Extracurricular Cortex
Huddart says his team “cobbled together a platform”, which is now called Cortex. It began life as a pricing engine, but is “an open platform that runs in the cloud that ingests data and allows us to make decisions very quickly”, according to Huddart.
“In the next few years, we want to be able to retrain all our data models, automatically, in near real time, based on the claims data coming in” Dan Huddart, Homeprotect
He says Cortex is not just limited to pricing and is already used in other areas. It is even training Google.
When businesses spend money on digital marketing with Google, they have to tell it what kind of people they want to attract. Huddart says Cortex gives feedback on customers to Google and, over time, it teaches Google which customers are more valuable. It has plugged Google into Cortex, which has enabled it to direct its marketing spend towards customers it knows are the most attractive, based on previous relationships.
“Cortex gives Google a guide to the customers that are most attractive to Homeprotect through data on their spending,” says Huddart.
Next up is an attempt to close the gap between data coming in and updating data science models. Due to claims being made constantly, premiums need to change, but according to Huddart, it take months, even years, for insurers to update data. This makes insurers slow to react to changing claims behaviour. But Cortex is being used to change this for Homeprotect.
Huddart says: “We are planning for Cortex to monitor claims data in real time. In the next few years, we want to be able to retrain all our data models, automatically, in near real time, based on the claims data coming in.”