Key Highlights
Around 30,000 policies priced through one model in its first six months
Unified nine offices, 20+ territories and 12 products into a single platform
Delivered a year ahead of schedule and recognised with a Howden Flight Award
It’s been a huge transformation, but it was definitely the right one to do.
Catherine Farnworth
Chief Actuary, UK & Europe
About DUAL
DUAL develops specialist insurance products on behalf of the carrier partners who provide its capacity. Founded in 1998, it is one of the world’s largest international MGAs, writing business across 22 offices, and 70+ product lines. As it grew, it needed pricing that could scale with the business - something future-proof and well-governed.
The Challenge for DUAL
DUAL had grown into a global underwriting business with the data inventory to match: over 25 years of policies, many products, many territories, and growth ambitions that the existing tooling could no longer carry.
A patchwork of spreadsheets, by product and by territory
Each product line, and in some areas each territory, ran on its own Excel pricing model. Sub-products often added more variations on top. Policy admin systems held some of the rating logic. Offshore teams re-keyed the same submission data into the policy admin systems that underwriters had already typed into their spreadsheets. The maths worked, but it was difficult to scale, hard to govern, and almost impossible to roll up to a portfolio view.
Joel Smith, Head of Pricing, describes the version-control cost of that world: “If your pricing model is built in Excel, every time you’re releasing a model, that means there’s a new version of the Excel model floating around that people might end up using.” Updating a model meant chasing down which spreadsheet copy lived where, which slowed every change request from the underwriting team.
A portfolio-level standard DUAL set out to meet for its carrier partners
As an MGA, DUAL underwrites on behalf of carrier partners who provide capacity. Those carriers increasingly wanted portfolio-level evidence: rate adequacy, mix, performance. Building those views meant the actuaries pulling data out of disparate spreadsheets, conforming it, and rebuilding pipelines per question.
Speed in dynamic lines, with the underwriter in the broker’s office
In many product lines, market conditions move fast. DUAL’s underwriters needed to adapt base rates quickly and price complex risks confidently, including in person, with a broker presenting a risk on the spot. Stephan Eberlein, Head of Cyber and Financial Lines for DUAL Austria, sums up the prior state simply: “We had a huge fragmentation among our tools… we struggled to have a clear overview of our rate adequacy at portfolio level.”
How hx Helps
One Europe-wide model where there used to be nine
DUAL consolidated nine European offices, more than 20 territories, and 12 financial lines products into a single European Financial Lines model on hx. That build won the Howden Flight Award and ran around 30,000 policies through the unified model in its first six months. That is far more pricing data, in one place, on a consistent basis than DUAL has ever had before. “Rather than replicating nine different local models into one framework, it actually was one true global model,” Farnworth says.
The same pattern is now repeating outside Europe. With the structured library approach, the team has created a global cyber model that would have been impractical to maintain as a set of regional spreadsheets.
Faster, more confident decisions in the broker’s office
In financial lines, where Eberlein’s D&O rater carries up to 40 underwriting parameters, the same workflow handles SME and large-corporate risks without underwriters switching tools or losing their rationale.
Behind that, Joel Smith’s pricing team can release model changes far more responsively, because version control no longer lives in filenames. “When you go to price a new quote, you will automatically be on the latest version of the model… it gives us the confidence to release models that fix little bugs or make little changes a lot more quickly.”
Portfolio-level evidence for carrier partners
The unified model lets DUAL show carriers a real portfolio view (rate adequacy, mix, performance) rather than a stitched-together one. “From an actuarial perspective, one of the key outcomes for us is we’re able to evidence a robust and well-governed pricing framework to our carrier partners, and then also evidence the profitability of the portfolios that we write, proving our credibility,” Farnworth says. Batch rating, the ability to rerun a live portfolio of risks through an updated model and compare the result, is now routine on the pricing team.
AI that augments underwriters instead of replacing them
DUAL is building an end-to-end underwriting architecture with hx at its centre: AI-based ingestion at the front, enrichment and triage in the middle, hx for pricing, and the policy admin systems on the back, all connected over APIs. The codified Python models hold the technical pricing logic. The underwriters keep the judgment.
The hx Difference
DUAL ran a formal RFP and considered building a Python pricing framework in-house. “At the beginning of our project, we did consider whether we should build our own Python-based framework,” says Mark Cook, Group COO, “but we rapidly realized that if we were going to do this, we would have to build out a number of critical services. So, for example, error and exception management, model version and lifecycle management, and also environment management.” The decision to choose hx came down to fit with how DUAL actually operates.
Not just a vendor, but a partner. That decision came out of a deliberately cross-functional process. DUAL first built an end-to-end prototype with hx, then ran the formal RFP, and treated the choice as one the whole business had to own rather than just the technology team. Cook stood up a steering committee with senior representatives from across the business to take the major decisions and resolve issues as they arose.
For Farnworth, the project’s executive sponsor, that same togetherness was what made the build work: “The real key for us was making sure that actuarial, underwriting and operations all worked seamlessly together as one team. That was the key for success for us.”
Specialty insurance experience. hx works with a number of specialty insurers and MGAs, and the team understands how to model the complexity of the lines DUAL writes. “hyperexponential have a number of clients in this space, but very critically, they understand how to model and manage the complexity that’s key to many lines of business in this space,” Cook says.
A Python-based stack that fits DUAL’s architecture. DUAL runs an API-based digital backbone connecting portals on the front end to policy admin systems and workflow on the back. hx integrates directly into that backbone, and being Python end-to-end makes the platform straightforward to staff and to fold into DUAL’s existing development pipelines.
Inheritance, reuse, and a structured environment. Together with Farnworth, Cook set a clear architectural goal: common foundation models with region-specific elements layered on top. hx’s library model, where code sits outside individual models and can be called by many, made that vision work in practice. The structured separation between development, test, and live is one of the things the pricing team would now find hardest to give up.
A roadmap DUAL is already growing into. Portfolio Intelligence is next. Once it is embedded, Farnworth expects two things to change: actuaries will interrogate live portfolio data without rebuilding pipelines, and underwriters will review and improve their own portfolios in the same place they price. “Rather than having to go into reserving dashboards or look at individual pricing models, they’re going have it all in front of them to be able to really make the most for our carriers.”
Looking back at the decision to sponsor the programme, Farnworth is direct: “It’s worked really well for us. It’s led to us not just changing our pricing, but also the whole framework that goes around it, the underwriting workflow, the governance, the data capture.”
