Making insurance more transparent through advanced data management
In a nutshell
- Managing data in insurance is complex because insurance policies have dynamic lifecycles.
- In order to make insurance less complex and more transparent for non-insurance businesses, Qover’s data team prioritises having a robust data model and a versatile tech stack.
- Qover’s data team provides data extracts and dashboards for insurance programs to make sure that partners have access to the right data at the right time, in their preferred format.
Data has always played a pivotal role in insurance. After all, it's what allows insurers to price risks accurately.
While the complex relationship between data and insurance goes back centuries, properly managing that data remains a challenge.
For those (fellow data nerds) who've read the popular book Data Warehouse Toolkit, you might recall that data in insurance is only found in its final chapters. That’s because, in order to handle insurance data correctly you've got to understand and set up everything that comes before it.
As the person in charge of data at a leading insurtech, Grégoire is up for the challenge. We recently sat down with Qover’s Head of Data to ask how he, along with our data team, managed to come up with a data framework that allows us to ultimately make insurance more transparent for our partners.
Why is managing insurance data so challenging?
Data in insurance is complex because insurance policies change over time.
To better understand how this works, let's consider an ice cream shop. A customer pays for their ice cream and eats it, which means the transaction is immediate and final. It's unlikely they're going to come back and ask to switch their vanilla scoop for chocolate after having a few bites.
With insurance on the other hand, data entities like contracts and claims have complex lifecycles that can change over time. For example policyholders may need to extend their coverage or cancel their policy. In the first case, the contract price increases and boosts insurer revenue. In the latter, the insurer may have to refund a portion of the premium.
That's why it's essential to organise insurance data accurately to reflect how dynamic insurance policies are and manage any edge cases that might come up.
What is Qover’s approach to data management as an embedded insurance orchestrator?
As an embedded insurance orchestrator, Qover acts as a central hub that connects multiple players in the insurance ecosystem, each with their own data formats and requirements (because why not add another layer of complexity to the mix?).
So even though we aim to make insurance less complex for our partners, we also need to ensure that the complexity doesn’t compound on our end. We have to balance being flexible enough to create insurance products that meet each partner's needs with keeping all of that data streamlined on our side.
This means it’s important to leverage the synergies between insurance programs that seem very different at first glance (travel insurance and motor insurance, for example).
Ultimately, our mission as a data team is to make insurance more transparent for our partners, i.e. the right data is available in the right format at all times. In order to do that, we prioritise having a robust data model and a versatile tech stack. Going into the details of our data model is enough for another blog post, so here we'll focus on the tech stack.
Can you talk a bit more about Qover’s data stack?
Unlike many companies where data departments primarily provide analysis and insights, our primary focus is operational. We support the launch and regular maintenance of insurance programs by providing data resources, such as extracts and dashboards, to various ecosystem players.
As Qover continues to grow – and therefore requires more data sources and deliverables – we need to be able to launch new insurance programs at scale.
To that end, we adopt the modern data stack approach, which decouples components while making sure everything is integrated seamlessly – largely through code and automation.
We orchestrate our data pipelines with dbt, following the extract, load, transform (ELT) approach.
Qover's data warehouse is organised using the medallion architecture popularised by Databricks. It's divided into layers based on how mature the data is, which helps us transform raw data into business-ready tables for our partners.
What is Qover's insurance data team working on next?
Part of Qover's ethos is that insurance solutions should be delivered in days, not months. That's why we've built a versatile platform where we easily configure any insurance product, whether it's motor, home, travel, etc.
Reporting is no exception, so we're currently developing an extra layer on top of our data warehouse called the semantic layer, which will allow partners to self-serve all the data they need.
This could be for data extracts – such as getting a list of all the claims that were processed over the past few months and the amounts paid – or for key performance indicators, like knowing the average handling time for claims in a specific country.
Once we can prove that our data infrastructure scales easily (i.e. being able to smoothly launch new programs without any bespoke developments), we'll start investing more resources in machine learning and AI-based initiatives (which our claims and customer care teams are already doing actually).
So in short, there's never a dull moment on Qover's data team.
.webp)
