UPDATED 2/19/2015 for SAS® Decision Manager v 2.2 reflecting ongoing integration of capabilities into the core product as well as new features (in italics).
SAS has announced its new Decision Management product – SAS® Decision Manager – back in 2013 and it is now in version 2.2. Over recent years SAS sees an increasing focus from clients on using analytics both for strategic decision-making and operational decision-making. Organizations seeking to develop analytic solutions for operational decision-making have historically struggled as there is a high degree of dysfunction resulting from silos in the analytic lifecycle. As you move from discovery to modeling, deployment and monitoring there are hand offs and inefficiencies. For instance a business manager or analyst specifies a need, data scientists try to develop analytic models to meet this need and then have to work with IT to recode this model into something that can be deployed. Testing to confirm that this code matches the model, and that it delivers the business intent, follows. All this means a long time to value for the business – up to 9 months at times. What clients need is an integrated lifecycle supported by a common platform that drives collaboration between business, IT and analytics teams – what I see as the three legged stool of success. And this needs to deliver continuous improvement of the decision-making.
SAS Decision Manager is designed to deliver this environment. A consistent user experience and management environment supports analytics, IT and business people across the whole lifecycle of preparing data, modeling, optimizing, deployment and monitoring. SAS Decision Manager can be deployed in batch, in-database, in-memory, real-time and streaming environments. Specifically SAS Decision Manager:
- Integrates business rules, predictive analytic models and optimization with business processes
- Provides a common decision authoring environment
- Centralizes decision design that supports discovery, analysis, simulation, detailed testing and reporting
- Leverages the existing SAS high-performance computing and big data infrastructure
SAS Decision Manager provides a new, project-based collaboration and governance environment and pulls together SAS ® Business Rules Manager with existing capabilities including SAS® Visual Data Builder, SAS® Model Manager (review here) and SAS® Data Management. Because it runs on the shared SAS platform it can also take advantage of lineage and metadata capabilities, security, monitoring, workflow and scheduling.
A typical user experience with SAS Decision Manager brings all these capabilities into a single browser-based UI. SAS Decision Manager can connect to and access a wide range of existing data sources, leveraging any work already done to bring data into the SAS environment. This data can be visualized and explored in the SAS Decision Manager interface New models can be developed against this data using tools such as SAS® Enterprise Miner (review here) or SAS® Rapid Predictive Modeler (review here). These models, as well as R or PMML models being managed in Model Manager can be analyzed and compared using the Model Manager functionality.
The decision editor is a key element of the product and provides a single environment to author and test decisions and detail when a predictive model is relevant to a defined business scenario. Users can design decisions to include branches, sub-flows, A/B splits and more to the decision specification. Each model has a signature showing the attributes it needs so the data it will need to score a transaction is clear when it is added to the flow. And challenger models can be identified to automatically be enacted when threshold tolerance of production models is reached.
As well as models, a task in the decision can allow business or IT users to specify business rules using the decision table editor (part of SAS Business Rules Manager). Users can specify rules against the same properties used by the model, the model outputs or other data in the project. These rules can represent policies like maximum values allowed, brands we don’t want to buy or whatever. These rules are then combined with the analytic models in the decision flow to produce decisions that are both analytically accurate and compliant with policies and regulations.
Business rules can be imported from other tables stored in SAS metadata and can be generated using advanced analytical algorithms, enabling users to discover potentially unknown business rules that may be buried in operational data from their business rules environment.
From the Decision definition a set of data can be run through and the results analyzed and used to refine the rules. Rule log information is also available, important both for regulatory compliance and when analyzing decision behavior for continuous improvement. The decision flow can also be used to walk through the decision one step at a time for a test case. Once everything is ready the decision flow can be deployed and accessed from any other system, from a web page to a legacy application or mobile app.
Lineage is important for the impact analysis of published and deployed decisions, including models and business rules and the data that these objects use. SAS Decision Manager has been extended to include the ability to trace object relationships from data to model to decision flows for instance, or showing the lineage of data. Data governance and other monitoring/control tools can be applied as the output of these tools is integrated with the SAS platform.
More information on SAS’ Decision Management products can be found here. SAS is one of the vendors in the Decision Management Systems Platform Technologies report.