PREDiCT model refresh

Following the release of the updated performance analysis of PREDiCT Large Loss last month, we have now completed the latest upgrade to the model. The dataset which underpins PREDiCT has been updated, and the model has been refreshed.

The benefits of the refreshed model include:

• Further improved PREDiCT accuracy when compared to paid outcomes on a portfolio of cases - we tested the performance of the new dataset against the existing dataset on a sample portfolio. The variance between the model’s damages prediction and the ultimate paid outcome across the portfolio improved by 22%.
• The addition of over 12 months’ worth of settled cases - these comprise cases from all injury cohorts, enabling enhanced modelling predictions based on a larger pool of data. 
• Updated inflationary information and data - the issue of claims inflation has been a significant source of additional leakage in large loss claims in recent years. Updated data including more recent settlements provides greater fidelity to inflationary aspects of the predictions.
• The addition of two new model predictors – these new predictors specifically target improved accuracy for future loss of earnings predictions.

Our recent performance analysis detailed the significant benefits of PREDiCT’s insights for our clients. In particular:

• Client approved reserving accuracy improved from +138% compared with ultimate paid outcome, to +55% following the market launch of PREDiCT in July 2021.
• When using PREDiCT for proactive case handling strategies, our clients experienced a 7% reduction in median claims lifecycles.
• Average damages paid in cases have seen settlements achieved at a 16% lower percentile compared to client approved reserves. Accelerated lifecycles have also helped avoid additional layers of costs in claims.

Whilst PREDiCT forms an integral part of our claims handling proposition, we have also seen exciting developments in the number of ways PREDiCT has been used by clients to meet particular challenges, solve specific issues, and deliver vital insights.