What do we consider a good model? A model that runs fast and can run via cloud? A model with a lot of features and few (known?) bugs? Do particular programming languages or software packages make a good solution?
We will try to find our answers during a two day event full of software trainings, collaboratively organized by actuaries and software developers from Triple A Netherlands, Triple A Poland and Triple A Spain. What better venue to pick than the brand new Barcelona office?
This international team, in which languages do not only differ per country but also per profession and expertise, shows that language barriers can be overcome. It makes me realize that a choice of programming language cannot be the answer to the question what makes a good model. In order to work together efficiently, it is important to understand each other and to connect with each other. Communication is key, a language is just a means to an end. Triple A embraces this philosophy by staying as software independent as possible to advise and serve clients in their specific needs and to solve the right problem. The right tool is often case dependent and might probably even vary over time.
The trainings held introduce us to several programming languages and software packages to understand more about what tool could be optimal for what particular case.
Listening, reading and viewing makes you understand concepts, it is experiencing and explaining to others what most contributes to reproducing what you have learnt. Interactive training sessions are therefore provided both to and by the team and modern technology even supports remote teaching for people unable to attend physically. The latter makes me realize that it is not just the amount of (known/unknown) bugs and issues that defines model quality. Especially in a field in which expectations rarely turn out to match realizations, I realize it is more about how well one is able to anticipate on what is faced.
The event is structured using an agile framework, which is all about being flexible and being adaptive. The first training day is closed with a review and retrospective session, inspecting both model and people performance. A discussion on what went well and what can be improved supporting personal and collective growth, effects already visible during the parallel training sessions of the second day.
To work together efficiently, it is important to understand and connect with each other and it makes me realize that a high quality model should meet expectations rather than have a lot of very fast and pretty though redundant features.
So then what does comprise a good model? The elements discussed can be combined and summarized in my personal definition of a good model. To me, a good model is a model which fits its purpose and is adaptable, as technological and regulatory dynamics will continuously require changes to whatever you have.
« It is neither the smartest nor strongest that will survive, it is the ones that are most adaptable to the changing environment in which one finds itself. » Darwin
© 2020 AAA-Riskfinance – Belgium. Tous droits réservés.