Organizations rely on analytical models to guide strategic and tactical decision making, such as estimating risk, predicting customer behavior, estimating fair value and analyzing business strategies. Because models play an increasingly important role in business processes, it is critical to reduce the likelihood of erroneous model output or incorrect interpretation of results.

In the Spotlight Best Practices for Managing Predictive Models in a Production Environment This white paper provides an overview of the model management lifecycle process and gives an introduction to SAS Model Manager. Ultimately, the paper recommends best practices for the management of predictive models in a production environment.

But what happens when the number of models in an organization reaches into the thousands, with new models coming online each day for deployment into a wide range of critical operational systems? This challenge is commonplace in the financial services industry and is becoming more common in communications, retail and other industries as well.

Managing the life cycle of analytics today is largely a manual process. Moving data and models from creation into production often entails tedious translations from one programming language or operating system format to another as the analytics are “pushed” across various IT platforms. Mistakes can be made as code is cut, pasted or rewritten by different people, and rarely is there enough time to add comments to describe why a particular algorithm was used or what reasoning went into choosing particular variables to input into a model.

As organizations implement analytical models into more and more of their operational systems, new risks are incurred – namely the profits lost by not updating models frequently enough. If models decay, projections are inaccurate and that leads to poor business decisions.

In addition, the lack of automated continual management of analytical models through their entire life cycle is becoming painfully evident as organizations struggle to meet increasing regulatory mandates. Only after a detailed modeling approach is in place can compliance reports generated tell a consistent, accurate story.

New software streamlines the process
SAS Model Manager streamlines the model implementation process and makes it easy to deploy analytics into operational systems by automating the tedious and often error-prone steps involved in the workflow of creating, managing and deploying analytical models. It provides patented technology with a secure, centralized repository for storing and organizing models, backed by extensive documentation templates for each model.

SAS Model Manager works with any predictive model created by various analytical modeling tools – SAS Enterprise Miner, SAS/STAT, Base SAS, and SAS code-based models. It provides a framework for:

• Organizing and tracking the tasks of model development.
• Model verification and testing.
• Comparative model performance benchmarking.
• Model deployment into production environments.
• Publishing and sharing model performance data through established reporting channels.
• Retiring models from production status.

SAS enables the easy transfer of analytics from the environment where they were originally created to the production operational environment with a consistent, repeatable process. By making it easier to implement and operationalize analytics findings, SAS helps organizations get the most value from their technology, data and analytic investments.