An effective predictive model management strategy involves relevant processes, tools, reporting and management. With these pieces in place the business benefits associated with model management are substantial. The most significant of these is risk reduction, less regulatory overhead and improved business performance. An unmanaged predictive model environment inevitably creates blind spots and associated risk. This will be quickly identified by regulators, creating unscheduled workloads to address regulatory issues, and possibly worse. And of course business processes will suffer. The benefits associated with effective model management can be broadly outlined as:
- Reduced regulatory overhead with less time spent on audits and more time spent developing much-needed new models. The creation of workflows and documentation means that regulatory requirements are more easily addressed.
- Avoiding sub-optimal decisions because of model degradation. The automation of validation, tracking and monitoring of existing models provides early identification of model degradation.
- Faster deployment of models simply because the environment is better managed and more resources can be dedicated to the task, since the regulatory overhead is reduced. Tools are also available to port models from the development environment, which might use a different language to the one needed in production. A lack of such tools can create significant delays and extra costs.
- More sophisticated modelling and simulation. Since model management is a prerequisite for additional optimisation and reporting, models can be tested in ‘what-if’ scenarios and analysed in the context of the constraints and objectives of the business (often called prescriptive analytics).
- Faster development of models when development activities can be monitored and activities are coordinated.
Model management is not some passive activity aimed at simply at satisfying regulators, although it does address this need. Ultimately model management should form the hub around which model development, testing, validation, deployment and reporting all revolve. Going beyond this it allows models to be seen in the context of business performance, and provides the links to associated activities such as optimisation, corporate performance monitoring, business intelligence and other activities that are used to determine the status and performance of the business.
In many ways model management moves the use of predictive technologies from the realm of the esoteric into mainstream business operations, with all the necessary controls, processes, reporting and links to the rest of the business. And of course we should not forget that it will lighten the regulatory burden considerably.
A broader consideration of decision automation, of which the use of predictive models is just one part, implies the need for an integrated, centralised mechanism for managing the technologies and processes involved. A model management platform should facilitate such integration, and be able to encompass activities such as optimisation and business rules management – or any other decision automation methods that might be used. Ultimately we will need an integrated environment in exactly the same way Enterprise Resource Planning systems integrated the transactional environment.
Previous article in this series: Model Management Tools
Next article in this series is Predictive Model management Strategy