top of page
ClsoeIT Logo

Workflows in Model Governance – Farewell, Dear Excel

Models used by financial institutions (FIs) are not defined only by their source code, datasets, and documentation, but also by the entire model lifecycle, which can typically be broken down into a set of workflows. 

Therefore, FIs often acquire workflow management systems, typically standalone, or the processes are "defined and maintained" in good old Excel sheets. While this might have seemed like a sufficient solution, there are much more advanced approaches, with an incomparably higher value for the business.

Excel vs. a real workflow management system

Let's start with the basics. The main motivation behind having a powerful workflow management system in place (and sorry my dear Excel, I'm not talking about you) is the following :  

  • Reliable, real-time overview of the current state of work on models (development, validation, internal audit...).

  • Each step in the workflow containing a precise description of the work to be done (work item). The expert performing the task knows exactly what to do and what result she/he is expected to provide. This way, even when a team member leaves, his knowledge is not lost forever.

  • Workflows spread the methodology consistently across teams in the organization. The result is better alignment, for instance between model developers and validators. 

  • Reference this case: Model developers must upload the data in sufficient quality. The workflow management system will allow checking the provided data and decide if the model can progress further in its lifecycle, or if it is necessary to first provide data in better quality.

  • An information base for ad-hoc reporting, process optimization, and measuring teams' efficiency.

  • A platform for steering a number of teams, even working remotely (no matter if just next-door or offshore). 

The workflow management system used for model governance should also reflect user roles, so team members work only on tasks they've expertise & role for.  After defining the workflow, experts can work on tasks even in parallel where suitable, so the workload is spread efficiently, in an agile manner (if the team and management is ready for it, of course). Additionally, the endless "What should I do now?" and "What are you working on now?” emails are eliminated through better management. 

Model Governance suite workflow

Example of a standard validation workflow taken from Model Governance Suite. Each rectangle represents a human work item. The horizontal swimlanes determine the users' role in the team. In the first step, the developer has to provide the data. The next step is performed by a model owner, he should approve the collected data provided in the previous step. This could even be done automatically in the workflow (e.g. by triggering a Python calculation, which will check the data provided). This way, the workflow follows a number of decision gateways and loopbacks which are triggered when, for example, there is something more to be provided. All the events performed are tracked and stored, for reporting or audit purposes.

The second step to glory

Until now, we’ve focused more on a workflow management system alone. However, the real power is hidden in having the workflow management system connected or even embedded within the model inventory. This way, work is always tracked without unnecessary burdens like model ID copypasting. Workflows then become really interactive, for example allowing users to update qualitative model data directly in the workflow step without context-switching. Furthermore, the system can pre-select workflows based on factors like model type (IRB credit risk model vs. experimental ML pricing model) or importance, as different models naturally require a different approach. A good example is model validation - models covering a significant part of the portfolio will have to be validated using a thorough workflow, meanwhile, a wild experimental model used for an internal decision will either require a very simplified validation or maybe not even that.


All these improvements, combined together, bring the usually rather static model inventory to life, upgrading it from (a simple yet must-have) catalogue to an effective foundation of daily operations of model validation, development, model risk management or audit departments.

Related Posts

See All

MGS integration with antivirus

One of the MGS features is to manage model-related files and documents. Of course, common and other model non-related files can be uploaded also to the public folder, users' private folder, or shared

MGS to Databricks connector

Analytics teams in banks and insurance companies use platforms like Databricks for data science, ML/AI model development, model testing, and more. These same companies also have teams such as Senior R


bottom of page