The essence of risk management | Modeling approaches

In this second interview about Risk Labs, Horiens’ risk analysis laboratory, Márcio dos Santos details the possible modeling approaches and the differentials for clients.

If you haven’t read the first part of the interview, click here

So there are two approaches to modeling: data-based or model-based. Are they both accurate?

When it comes to precision, the same question applies to the modeling or frequentist approach. It doesn’t change anything. When calculating that the frequency of a particular accident is 1%, care must be taken when interpreting a degree of certainty for it. Statistics and engineering do not aim to calculate probability assertively, because there is no such thing. Probability is an element in making a decision, even before the event takes place. You can talk about what degree of confidence interval or degree of risk I have in this estimate, but there is uncertainty. What will count is the appetite and propensity to make a decision with that degree of uncertainty.

Do customers see value?

Quite a lot. They are increasingly learning to value the incorporation of uncertainty into deterministic models. For example, the stability of buildings. To analyze risk, we take the deterministic model and incorporate uncertainties into it, such as variations in wind, temperature, and concrete and soil resistance. In this way, we were able to rehearse in which scenarios there would be a risk of problems for the building. It’s a much more sophisticated analysis that requires greater involvement from the client, but just like data-based analysis, it allows you to make decisions without surprises, focusing on the main risks. It is very common in geological and geotechnical risks, which are the highest risks in civil engineering, such as tunnels, dams and excavations in general. It brings benefits for the drafting of contracts, the leasing of risks between the parties. It’s very important.

How long does the analysis take?

If it’s a problem we’ve analyzed in the past and we have workflows – code already prepared – it’s very quick. It only takes a few weeks. Risk analysis based on engineering models, on the other hand, needs to look more closely at aspects of design, operation and interviews with the client’s engineering team. The journey is much longer and deeper.

Is this more in-depth journey a differentiator for Horiens?

Today, risk analysis is a highly developed area in which companies have invested heavily. But we have noticed a concentration on the data side. We are in a scenario of sparse data, which does not allow us to build models from the data alone. This makes it impossible for most players to carry out risk analysis, as they restrict risk analysis to data analysis (data science). If you don’t have data, you can’t analyze risks. At Risk Labs, even because of our origins, the team has incorporated the possibility of simulating the models from a probabilistic point of view.

I’m a geotechnical engineer. So I was already working with geotechnical uncertainty before working with insurance and, for me, the question of testing deterministic models is more familiar than doing data analytics. There are far fewer companies offering this on the market. You need to be in “shoulder-to-shoulder” contact with engineers and designers, and understand engineering models such as slope stability. Although many engineers work with analytics, few go into the business of simulating deterministic models.

Risk Labs technology is available to support all companies in their risk decision-making. Visit /who-we-are/risk-labs/ and consult our experts to find out more.

To read the rest of this interview, click on the link below:

The essence of risk management | The importance of the human eye