Context

Supporting the back testing of several technical frequency and severity models:

• Check the adequacy between the cost predictions provided by the technical model and the costs actually observed over the past year. In other words, comparing the theory, as predicted by the model, and reality.
• Based on the backtesting results, propose adjustments to the technical model.

Context

Custy has established itself as a major player in the development of digital solutions dedicated to the insurance industry.
There are present in France, Belgium and Luxembourg.

Approach

• Needs analysis to include group insurance in an existing sofware previously focused on the management of individual life insurance products.
• Actuarial analysis of the formulas associated with the various group insurance coverages.
• Link between back end developers, front end developers and management.

Context

Supported the performance of claims history analysis and of ultimate loss ratios estimations. Including:

• Management summary
• Description of the datasets
• Description of the actuarial methods
• Results of the analysis
• Sensitivities to the most relevant assumptions

Context

Supported pricing teams with

• Data transformation and data visualization of the input dataset : exposure graphs, marginal impact graphs, correlation analysis, outlier detection…
• Additional analyzes based on the input dataset
• Technical pricing of the new products using GAM models : implementation of an interative procedure whose goal is to determine the optimal model + autoimation of the pricing process
• Commercial pricing

Context

Supported pricing teams with

• Data transformation and data visualization of the input dataset : exposure graphs, marginal impact graphs, correlation analysis, outlier detection…
• Additional analyzes based on the input dataset
• Technical pricing of the new products using GAM models : implementation of an interative procedure whose goal is to determine the optimal model + autoimation of the pricing process
• Commercial pricing

Context

Custy has established itself as a major player in the development of digital solutions dedicated to the insurance industry.
There are present in France, Belgium and Luxembourg.

Approach

• Actuarial analyses related to the migration of a portfolio from insurance company A to insurance company B.
• Use and transformation of a VBA code designed to validate all actuarial calculations linked to the migration.

Context

A major European bank faced the need to overhaul its Internal Ratings Based (IRB) approach for calculating Probability of Default (PD) and Loss Given Default (LGD). Marked by a submission to the European Central Bank (ECB), the project aimed to enhance the bank’s risk assessment accuracy and regulatory compliance, emphasizing the importance of precision in financial risk management.

Approach

Altair’s mission in developing PD and LGD models within the IRB framework significantly enhanced risk assessment capabilities by automating the model modernization process, spanning from descriptive statistics to the actual creation of the models and their visual support tools. This streamlined the modeling process, making it more efficient and less prone to errors and enabled dynamic adjustments and updates to the models as new data became available. Importantly, automation enabled the exploration of the full range of possible models, ensuring comprehensive coverage of modeling approaches and scenarios. The incorporation of visual support tools further enhanced the interpretability and usability of the models, allowing stakeholders to gain insights more intuitively and make informed decisions more effectively.
These advancements added significant value by improving the accuracy of risk assessments, reducing the time and resources required for model development and maintenance, and enhancing the overall strategic decision-making process for financial institutions. Altair’s complied with regulatory demands but also elevated risk management practices through innovation and strategic efficiency.

Context

Researchers at ULB and UMons have asked us to take part in the “Portraits of Journalists” project. This major survey, conducted by LaPIJ (ULB/UMons), the Center for Journalism Studies (UGhent), AJP and VVJ, aims to present the “Belgian journalist”.

Approach

We carried out a series of descriptive and inferential analyses (Student-type tests for comparison of means, Chi2-type tests for comparison of proportions), with numerous cross-tabulations of variables, and took part in proofreading the FR section of the report.

The report has been published and is available at the following address:

L’enquête « Portrait des journalistes belges en 2023 » est en ligne !

Context

Ictyodev is a veterinary medicine research laboratory. As part of their research projects, they needed to carry out analyses to compare the efficacy of various methods, for example.

Approach

As a research laboratory, Ictyodev has precise specifications and specific biostatistical expectations. They carried out their measurements and created their dataset, which they sent to us for analysis with a listing of the comparisons they wanted to make and the hypotheses they wished to verify.

For our first assignment, we carried out comparisons of means on different indicators (Analysis of Variance – parametric and non-parametric ANOVA), as well as a survival analysis (Kaplan Meier).

The second mission consisted in building 3 mixed models (fixed effects and random effects).

Reports were delivered with software outputs/results (R) and their interpretation. These reports were created and delivered according to the customer’s expectations, so that they could then be evaluated and reformatted according to the writing and publication standards associated with the laboratory’s sector of activity.

Context

As part of a call for projects commissioned by ADEME (Impacts of incentive-based pricing on incivilities in waste management), Ecogeos (a consultancy specialized in waste treatment) asked us to intervene at various stages of the project.

Approach

The expertise required by ADEME was knowledge of analysis methods adapted to small samples. Ecogeos therefore had to carry out extensive work on sampling quality to ensure that the sample was as representative as possible, and to use non-parametric tests for the analysis where necessary to obtain the least biased results possible.

We were involved in the following areas:

• Participation in the definition of the data collection methodology
• Performing data analysis, both descriptive and inferential (measuring the impact of variables on each other, testing hypotheses), on small samples (non-parametric tests)
• Participation in follow-up meetings

A public report has been published by ADEME and is available for consultation here: https://librairie.ademe.fr/dechets-economie-circulaire/6774-tarification-incitative-et-incivilites.html