How we leveraged pet insurance data to predict diseases
For a client in the insurance industry, data is essential. This is where SPRYFOX’s machine learning models come in. Using our engine, the client can get a better sense of pets’ health trajectories. Our models allow insurance experts to predict their furry customers’ likelihood of contracting specific conditions, providing vital insights that can help pets live longer, healthier lives.
Our client is a pet health insurance company with over 2 million insurance claims from more than 750,000 animals. In line with their digital transformation, they wanted to provide more data-driven customer insights, reduce insurance costs, and elevate insured pets’ overall health status.
We started with a proof-of-concept phase: evaluating our existing data for disease prediction and testing basic machine learning methods. From there, we worked with the customer to uncover the massive business value of our joint efforts.
It quickly became clear there were vast product opportunities we could leverage. So to kick off the production phase, we evaluated complex machine learning methods and their ability to predict future diseases—seamlessly integrating external and internal data sources to develop our solution.
We overcame challenges such as the scope of diseases we set out to predict (we’re talking thousands of conditions!) and the correct fusion with external data sources. On completion of this phase, we delved into the actual software development. Our goal was to provide a robust, high-performance interface the customer could use to interact with our algorithms.
We created a powerful prediction engine trained on over 15 years of data—delivering valuable insights into the factors and preconditions that increase the risk of disease in pets. To achieve the highest possible accuracy, we built the engine using multiple machine learning methods in parallel.
A crucial feature of the predictive engine? Its connectivity to the customer’s own database, which allows for querying based on individual pets or groups of pets. The client can filter their search based on geography, breed, age, and other characteristics.
Now, we recognize the importance of accessibility. We knew we needed to prioritize the needs of our non-technical consumers, and focus on the interpretability of our prediction results. So in addition to a full-fledged API—used flexibly within existing customer products—we provided a web-based application with the ability to query and visualize all analytics.
The response has exceeded our expectations. Today our client can predict their furry customers’ possible health journeys, providing key insights that’ll help pets live longer, healthier lives.