Strategy is the meeting point of capability and need. Capability without need is simply wasted resources, and needs without capability will remain unresolved. The need for model management is being reinforced by the demands of the regulators. However it would be a wasted opportunity indeed if it was not seized to improve business performance, reduce risk and provide more information to managers and others involved in the production and use of predictive models.
The needs associated with model management can be summarised as:
- Regulatory needs – for transparency, understanding and clear documentation.
- Business needs – for optimal models, good integration with other information systems, greater model creation productivity, good quality documentation, closer collaboration between analysts and business management, management information and reduced business risk.
All of these needs can, to a greater or lesser extent, be met by a model management initiative. It is fairly obvious that while it addresses regulatory needs, the business benefits are far more substantial and numerous. However, as with all initiatives that integrate people and systems there will almost certainly be resistance in some quarters. Research has consistently shown (e.g. Tom DeMarco) that 25 per cent of projects concerned with information systems fail for political reasons. It makes sense to raise the profile of model management to the most senior level possible, simply because it is so important to the business.
To implement an effective model management strategy a number of capabilities need to be put in place. These include:
- High level sponsorship and awareness of the implications of failing to manage predictive models in an effective manner.
- Education and training of analysts and data scientists so models are produced that meet regulatory requirements.
- A model management platform that makes the task of documenting, validating, developing, and monitoring large numbers of complex predictive models feasible.
- Integration with both production systems and management information systems such as business intelligence and enterprise performance management systems.
- Reporting and escalation mechanisms to make sure model management remains effective.
Model management is really just part of the broader task of managing the decision automation process. And just as process automation culminated in Enterprise Resource Planning systems, so decision automation (the role for predictive models) will require an extensive model management platform that integrates and manages models which use techniques including predictive models. The alternative would be the equivalent of trying to run the transactional activity of a large corporation using spreadsheets.
Regulatory requirements provide a much needed incentive for businesses using predictive models to get their house in order. Done effectively the benefits go way beyond satisfying regulatory requirements, and provide a foundation for the broader management of the decision automation process, already well underway in many firms operating within financial services.
Previous article in this series is Predictive Model management Business Benefits
Download full report: