NVIDIA x Sia Partners Exclusive Event
Benchmark & Reverse Engineering
Sia Partners has developed skills in reverse engineering that facilitate the reconstruction of pricing models, using only a set of data from an initial “unknown model”. With internal or external databases, these methods create an accurate benchmark.
Sia Partners has developed skills in Reverse Engineering to reconstruct pricing models from a dataset of a previously unknown model. These datasets can come from various sources (data capture, actual loss experience of an insurer, etc.).
Our skills in Reverse Engineering also make it possible to highlight the influence of the explanatory variables, for example, we can extract pricing adjustments for the location variable.
Our approach is based on internally developed tools and is articulated according to the following four stages:
1 – Database
- Profiles (predictor variables)
- Rates (predictive variables)
- Insurer base / Open Data
- Data capture
2 – Analyzing the variables
- Uni-variate analysis: used to determine which variables have an impact on pricing.
- Bi-variate analysis: used to construct a zoning chart, for example.
3. Database Optimisation
Following the previous step, the variables impacting pricing are identified. The database is then cleaned to remove duplicates.
4 – Modeling
Simulation of several pricing models (cost-frequency, GLM, Machine Learning) to identify the various additive models and deduce the final pricing formula. This step enables us to determine the profiles attached to each model used, as well as the multiplicative variables (zoning, for example).
With the tightening of regulatory constraints and commercial competition, price positioning has become a major issue for social protection players. This is why Sia Partners has developed Reverse Engineering methods to support its customers in the creation of new products, ensuring coherence in their product offerings and in integrating market norms.