nib Group provides health and medical insurance to over 1.6 million Australian and New Zealand residents. One of nib’s key business strategies is “Racing the Red Queen”. This strategy identifies that “the status quo is death” and that nib’s future lies in creating competitive advantage across the Group through constant innovation, technology, agility, having world-class talent, cultural alignment, and prudent risk management.
As part of the Racing the Red Queen strategic framework, nib brought in Intellify to see whether they could assist with using Machine Learning to better solve one of the most challenging, and consequential problems within their business – Outstanding Claims (OSC) Forecasting for Private Hospital. Each month, many thousands of operations are performed in Private Hospitals around Australia. Even though an operation may occur in one month it may take up to 2 years before the hospital provider claims the expense from nib. To adequately provision enough reserves to meet these obligations, nib needs to be able to make a forecast of how many of these “outstanding claims” are coming through the pipeline. The forecasts are also used to guide product pricing, marketing and business strategy. While standard actuarial practices are the current typical approach to OSC forecasting, nib saw significant opportunity in working with Intellify to “Race the Red Queen” and demonstrate whether machine learning based methods could yield a more accurate estimation.
Intellify’s Data Scientists worked closely with the nib actuarial team to quickly understand and examine the problem space thoroughly. It became clear that there was an alternative “statistical time-series” based approach that could be tried that offered an alternative way of approaching and solving the problem. Using an ensemble of machine learning models, the team was able to develop a solution that showed a substantial reduction in the error of nib’s 2-year OSC forecast. External, publicly available datasets were sourced and incorporated into the model and an additional performance boost was observed. Finally, Intellify also developed a robust, novel method of estimating confidence intervals or the OSC forecasts using Bayesian statistics and probabilistic bootstrapping techniques to develop an alternative method for estimating risk margins. All model work was developed using Databricks and MLflow on AWS – allowing rapid iteration with experiment tracking and model management. On the back of the project’s success, nib is currently working with Intellify to scale out the solution across all modalities across its claims forecasting function.
How we helped
The entire engagement was a strong, collaborative endeavour between Intellify’s Machine Learning consultants and nib’s actuarial team. The project was run by Intellify in an Agile, iterative way, featuring daily stand-ups and weekly results presentations that were used to inform and plan for the next week’s tasks and objectives. Intellify helped to introduce, socialize and demonstrate the value of a range of machine learning techniques and productionisation approaches to a range of nib stakeholders including the actuarial team, nib’s existing data science team, and the nib executive leadership team that resulted in substantial business impact.