Open Insurance Achieves 70% Data Self-Service with the Snowflake AI Data Cloud

Insurance has always been a data-driven industry. However, companies in the sector have been slow to move away from time-consuming, difficult and siloed data management practices. Open Insurance (Open)’s mission is to solve this problem, using a range of technology- and data-powered products and services. These include white-label insurance products; marketplaces and aggregator platforms; a direct-to-consumer brand; and access to leading-edge technology and data science capabilities. Accessing these products and services through a partner-led model enables leading brands in health insurance, telecommunications and other sectors to provide insurance to customers where and when they need it.

Open provides an on-brand look, feel and experience to its customers, underpinned by a data model that provides real-time insights to assist short- and long-term decision-making.

All 150 team members review data captured by the business — including customer transaction and product information — that is relevant to their roles, with senior leadership expected to run ad hoc queries themselves to obtain actionable insights. To support growing data volumes and organization-wide, role-based access to data through self-service, Open decided to go to market for a new data platform.         

This platform had to meet five key requirements:

• Built-in security controls and standards, along with the ability to mask personally identifiable information that only selected users were authorized to access 

• Automated scaling across multiple working data warehouses that could support app development with Streamlit 

• Extensive coverage of extract, transform, load (ETL) tools to manage the integration of multiple data sources 

• A multicluster architecture that enhanced performance

• Features that enabled Open to share data with teams in partner businesses for analysis and to optimize campaigns

Snowflake was the best fit. The organization implemented the Snowflake AI Data Cloud and integrated the product with internal analytics tools and SDKs, alongside external products, such as a Customer.io communication system, a CRM product and a Stripe billing system. For its UK operations, about 10 third-party systems integrate with the Snowflake AI Data Cloud. 

Open experienced the benefits of the Snowflake platform immediately. Its team also realized Snowflake could do much more than it initially thought, including surfacing trained GPT models and running these outputs in the Snowflake AI Data Cloud to support ad hoc exploration.    

With the Snowflake AI Data Cloud and Streamlit, Open is exposing and visualizing data to team members, at all levels of seniority and technical expertise, to facilitate self-service. “Through Snowflake, we have centralized and presented transactional and analytical data from a single source of truth,” says Sarith Fernando, General Manager, Data Product at Open. “We have also enabled multiple tools through GPT models and Streamlit so our people can visualize data for decision-making.”

Through the products and positive engagement with the business by the data product team, Open is achieving a 70% data self-service rate across its workforce.  

To move to self-service, the data product team modeled the most common data requests, including customer conversions and cancellations, streamlining the process for less-technical team members. Applying consistent data definitions and simplifying the use of data tables was integral to this process. The team also educated the heaviest data users, in functions such as finance and commercial, on the ease with which they could be upskilled to query data themselves. Now, most Open employees can access what they need within Snowflake and are confident enough in their output to request only light reviews from a product manager, engineer or via peer review, rather than the data team.  

By using Streamlit, Open spends 50% less time creating BI dashboards compared to its previous custom environment. It took just two engineers to move 30 dashboards from Open’s legacy provider to Streamlit in one month — a process that even included remodeling all data to use consistent definitions, designing Streamlit applications to visualize and interact with data more effectively, and documenting underlying data models that can be used by GPTs for various tasks. 

The number of BI requests at Open has also fallen by nearly a third since the deployment of self-service. After modeling all its key data through Snowflake and dbt Labs and improving its documentation, the company trained GPT models on the updated documentation, created a lightweight GPT model that could convert queries into natural language on SQL, ran SQL in the application, and enabled users to provide feedback to unlock chain-of-response benefits from the GPT model.

Coverage of ETL tools within the Snowflake AI Data Cloud has enabled Open to make integrating a recently acquired UK-based insurance technology company a seamless process. This entailed integrating 16 data sources, organizing it all and building a new platform — which ultimately took only a month to complete.  

With all data consolidated in the Snowflake AI Data Cloud, the stability and performance of data feeds to Open’s partner businesses has improved considerably over the legacy environment. According to Fernando, “With Snowflake, the accuracy and reliability of the data itself has gone up because we can identify and understand any issues far more quickly than we could previously.”