Taming the chaos of specialty insurance

Ditching the rigid structures of the past

As Development Manager and Principal Developer at Price Forbes, the wholesale, specialty, and reinsurance arm of the Ardonagh Group, Alin Neagu and Robert Meyer have firsthand insights into the technical evolution that has supported its evolution into London’s leading independent specialty broker.

Price Forbes’ growth has involved acquiring brokers and unifying them under a single brand. It’s an effective model, but from a data perspective, it presents a unique challenge: as an insurance broker that arranges cover for everything from residential houses and satellites to individual body parts, how does Price Forbes manage the sheer diversity of risks it works with?

Neagu and Meyer quickly realized that the answer didn’t lie in the rigid structures of the past. To build a platform of the future, they had to move away from legacy databases and embrace MongoDB Atlas.

Here’s the story of why they made the switch and how it revolutionized Price Forbes’ development culture.

The problem with rigid schemas in specialty insurance

The driving force behind Price Forbes’ move to MongoDB Atlas was a project to build a Compare-the-Market-style quote-and-bind system.

In the world of specialty insurance, Price Forbes deals with incredibly niche areas. When a broker enters details about a risk, they might be dealing with construction, cargo, aviation, or even outer space. From a technical standpoint, it’s a polymorphic nightmare.

“While every product has a top-level identity, each has its own unique set of fields and properties,” said Meyer. “When we looked at our existing MySQL implementation, we realized we were facing a series of impossible trade-offs. We could use attribute tables that are slow to query, we could use a single massive table with thousands of mostly empty columns for every possible product field, or we could create a different table for every single product.”

None of these SQL-based options were sustainable for a fast-growing business. Price Forbes needed a solution that allowed users to simply throw a product into the database and have it stored in a single collection while remaining fully searchable.

MongoDB’s JavaScript Object Notation (JSON) model was the natural fit. It allowed Price Forbes to identify each product by a class so that when it was retrieved, it was automatically deserialized into the correct format for the application.

Doubling performance: Moving from Cosmos DB to MongoDB Atlas

Price Forbes’ journey didn't lead straight to MongoDB Atlas; it started with a MongoDB-compatible instance of Cosmos DB on Azure. It worked as a proof of concept, but Price Forbes quickly hit a ceiling: Cosmos DB was using an outdated version of MongoDB, which limited document size and prevented Price Forbes from using the latest aggregation frameworks. It also turned out to be quite slow, leading to increased complaints.

“So we engaged directly with MongoDB and ran another proof of concept,” said Neagu. “The results were staggering: our performance more than doubled.”

The migration process itself, supported by MongoDB Professional Services, was incredibly smooth. Price Forbes was given five days of professional service time, but it only took two days to write the migration scripts and migrate all data remotely. The team didn’t encounter a single issue with any data type; it was a completely painless experience.

Powering the Placement Hub with reactive code

The most significant success of Price Forbes’ MongoDB adoption has been the Placement Hub. This is a next-generation platform that enables brokers to acquire risk information, process it into policy administration systems, and access vital documentation.

A key reason why this platform is so fast is that Price Forbes decided to use reactive code. In a standard, non-blocking model, a server thread handles a request and remains occupied until it completes. This can bog down the system when handling thousands of requests.

“We use the analogy of walking 10 dogs,” said Neagu. “In a non-reactive model, you throw a ball and wait for the first dog to return it before throwing the next. In a reactive model, you throw all 10 balls in quick succession; the ‘threads’ are released to do other work while the ‘dogs’ are busy fetching the data.”

Because MongoDB Atlas has drivers specifically designed for reactive programming, Price Forbes’ application is lightning-fast. It’s achieved a 90% adoption rate for the Placement Hub—a rate that’s unheard of in the insurance industry.

A developer-first philosophy

Ask Meyer and Neagu whether they would recommend MongoDB to another organization, and their immediate response is simple: “Do you love your developers?”

“In modern software development, we work with objects—what we call POJOs (Plain Old Java Objects),” said Meyer. “These objects are naturally represented as JSON, which is exactly how MongoDB stores data. This means there is a natural flow between our code and the database. Our developers no longer have to write crazy joins or complex stored procedures just to retrieve data. It just works out of the box.”

When Meyer and Neagu say that developers are lazy at heart, they mean it in the best possible way: developers look for the path of least resistance. With MongoDB, Ardonagh can create data objects without the headache of setting up rigid schemas or complex migration processes.

This ease of use is why all of its developers have now moved to MongoDB for new projects. Price Forbes maintains some legacy SQL systems for mandated policy administration, as required by the UK’s Financial Conduct Authority (FCA), but all active software development is now based on MongoDB.

The AI frontier: Vector Search and document intelligence

Price Forbes isn’t just using MongoDB for what it can do today; it also keeps its people at the cusp of breaking industry developments.

In the coming financial year, for example, Price Forbes is launching projects that will put it at the forefront of artificial intelligence in insurance. It is also currently working on using Vector Search within MongoDB Atlas to transform two key areas:

• Automated claims estimates: Price Forbes plans to vectorize historical claims data, including pictures and text, so claims handlers can automatically see real-world cost examples for new claims based on similar past patterns.
• Document intelligence: Specialty insurance policies can be 20 pages long. Instead of a client or broker manually browsing those pages to see what’s covered, Price Forbes wants to allow them to simply ask a question. By embedding these documents as vectors in MongoDB, the system can understand the entire document and provide instant answers.

“In the competitive world of brokerage, being able to provide that kind of instant service can be the difference between a client choosing us or a competitor,” noted Neagu.

The stuff of champions

Being MongoDB Champions has allowed Price Forbes to mentor colleagues and ensure that it’s always using the most optimized queries and aggregations. Its people have also seen firsthand that when you provide developers with a platform that’s as easy to use as their own code frameworks, they innovate effortlessly.

“That’s what we need to continue our journey from a one-pound acquisition to a global leader,” Meyer concluded. “And MongoDB Atlas gives us the scalability, availability, and performance to do it.”