Empowering a data science team to drive results in mid-market businesses

Effective data science teams can revolutionise businesses with detailed insights. Inflexion’s portfolio was invited to hear about the experience of building an in-house team at The Acorn Group at the recent Data & AI Exchange.

Insurance firms operate in an information-heavy and competitive market where customers are highly price sensitive. The operating model of insurers means they are inherently data-rich – yet don’t always make the most of this.

The Acorn Group was one such example. The specialist insurer, who have updated their name from The Granite Group, comprises a number of brands which distribute specialist motor insurance, with a market leading position in UK taxi insurance as well as specialist cover for private cars and vans. Its services are marketed through retail branches, price comparison websites, brokers, its own website and a call centre, and it has control of the full value chain through underwriting, distribution, claims handling, premium finance, investment management and vehicle hire. While this all bestows a wealth of granular data to Acorn, it wasn’t maximising it.

This has been changing since a minority investment from Inflexion in 2019 provided funding to grow the business. Efforts were further boosted in 2023 with the arrival of Duncan Procter as Data Science Director from a larger insurer, bringing with him experience in harnessing data to drive growth. The short time he’s been with Acorn has seen him create a small but effective team which illustrates the positive impact a data science team can have.  

"Acorn’s stronger competitors were leveraging data science profitability, and we knew we had substantial and niche datasets waiting to be exploited,” Duncan explains. The first step for Duncan was defining the role of the data science team. "We needed to solve hard problems, automate complex tasks, and extract signal from noise," he explains. "Our first use case was pricing — a critical aspect of our business,” he says, adding that getting this right is essential in insurance as it’s such a price-driven market.

What a data science team does (and doesn’t) do

But what exactly is a data science team, and what does it entail? It is not just about creating dashboards or automating tasks, Duncan clarifies, but is in fact a strategic investment. “A data science team is not a short-term fix – it requires substantial expertise and resources to deliver quantifiable solutions."

The requirements for a successful data science team are multifaceted. "We need data — lots of it," Duncan emphasises. "If you have data, the data science team wants it – all of it, as soon as it comes in, well documented, stored well and long-term, queryable at speed, reliably imported, granular, and raw. Clean data is valuable, but we also need the 'dirty' data to train our models effectively." Understanding the data science team’s need for this level of detail required a shift in mindset, Duncan admitted.

Technical infrastructure is equally crucial. "We need access to high-quality cloud machine learning (ML) suites and well-documented services," Duncan explains. "Integration into the tech estate is essential, but we also need the freedom to innovate and serve different departments."

Communication and collaboration are key aspects of Duncan's role. "I interface with various stakeholders, from the CTO to product leads," he says. "Understanding the value of data is critical. Collecting and storing it in raw and granular form is important to generate value."

Building the right team is paramount to success. "We need a small, efficient team with diverse skill sets," Duncan notes. "Data engineers, ML Ops engineers, data scientists, and data analysts — all working together to deliver high-value solutions."

Quantifiable solutions are the proof points for data science teams. "We aim to deliver measurable outcomes," Duncan explains. "Whether it's improved estimates of things you estimate, or providing answers to previously unanswerable questions, our goal is to drive tangible results." He shares an example of running an A/B test which showed a 5% uplift in objective, resulting in a six-figure monthly revenue boost. This feedback to the rest of the team helps to solidify buy-in and encourage further data sharing. “To establish a data science function, you need to have a single large use case to justify its cost.”

The timeline for projects varies depending on complexity and infrastructure. "First projects can take around six months, but once the infrastructure is in place, iteration becomes quicker," Duncan says.

In-house or consultants?

To outsource or insource is a critical decision that depends on several factors, according to Duncan. ”We started off with an external partner to kick-start our journey” he says. "Ultimately, we needed in-house expertise providing the necessary niche understanding to drive meaningful and sustainable results. In our industry pricing is constantly changing, requiring us to adjust all the time."

Acorn’s journey with its data science team is an ongoing success. "We've delivered substantial uplifts in profits for our key products," Duncan states. But he and his team are not stopping there.

Inflexion’s expertise and support were invaluable to Duncan's journey. "Having Inflexion’s in-house data expert Jan Beitner on my side provided much-needed support," Duncan acknowledges. "His experience from building other data science teams within the portfolio was very helpful."

The support should continue given the mighty results from Duncan’s small team of four: they’ve delivered solutions for Acorn’s two biggest products, representing c70% of income and delivering a c. 10% uplift on profits. There are solutions for other products queued up, meaning Acorn should continue to see the boost from harnessing its data and a savvy team driving it.