From Decision Management to Prescriptive Analytics
A number of organizations have adopted the idea of making use of the Decision Management approach and technologies to problems such as risk, fraud, eligibility, maximizing and more. If you read this blog, you probably already know what Decision Management brings to the table.
Decision Management is all about automating repeatable decisions in a maintainable way so that they can be optimized in a continuous fashion.
Decision systems can use Business Rules Management Systems (BRMS), but they do not need to restrict themselves to just that: they can also be built on Predictive Analytics technology; or they can even consist of a combination of both. The increasing availability of data that can be used to test, optimize decisions, or extract insights from, makes it possible for decision-centric applications to combine expertise and data to levels not seen in previous generations of applications.
In this post, we’ll outline the evolution from pure Business Rules Systems to Prescriptive Analytics platforms for decision-centric applications.
Business Rules Management Systems
Business Rules Management Systems manage decision logic: how a decision should be made given the data available for the decision-making. A BRMS strives to make it clear how a decision will be made, and why a given decision was made. The management aspects revolves around the ability to store decision artifacts in a repository that can be queried, and around maintainability, so that it is easy for business users to quickly react by changing the decisions and quickly deploying them.
Decision logic is provided in the form of business rules, each being made of a condition and an action. When a condition is true, the rule is said to fire (which means that its action is executed). Conditions work on the data that is currently available, while actions modify the data. When the conditions of rules use the data that is modified in the actions of other rules, you end up with a network of condition-actions through which a decision engine navigates to get to the conclusion of a decision.
Designing the decisions usually involves using existing policies, regulations or knowledge in the organization, and formalizing that information one way or another (for example, using a Decision Model and Notation tool).
Then, these decisions can be encoded in the target BRMS, ready for execution.
When a lot of historical data is available, or becomes available after the use of a BRMS for some time, Predictive Analytics systems can be used on that data to assess the possibility that something will be true in the future.
Predictive Analytics systems use mathematical techniques to correlate data and build a model to determine when a given outcome will result depending on new data values.
The output of a Predictive Analytics system may then be used to build some decision logic (in a BRMS) corresponding to the predictive model. Then, by fine-tuning the decision logic, one can end up with a small number of maintainable business rules providing conclusions to data that depend on the original historical data.
Predictive Analytics therefore add a new dimension to BRMS: experience (and time).
When automatable decisions are designed, it is crucial to create KPIs (Key Performance Indicators) which will provide a measure of the business performance of the automatable decisions.
With these KPIs, it becomes possible to tell whether a decision provides the expected business value; but also, if the decision is changed, whether it provides better or worse outcomes (using these same measurements).
We call “decision analytics” the inclusion of business KPIs as an integral part of the implementation of automated decisions, combined with the ability to run simulations to measure their value, and track it at run-time.
Continuously measuring the quality of decisions implemented through predictive models and business rules, taking the outcomes obtained in real life, champion-challenger experiments, or simulations, leads to a constant improvement of the decisions with a high reactivity to change.
Little by little, the quality of the data sets will improve, making them suitable for more fine-grained simulation scenarios on the existing decisions. By combining these data sets with different decisions and different versions of the same decisions, it becomes possible to obtain a prescription of which decisions to make given the data available at hand.
In short, Prescriptive Analytics platforms provide:
- The ability to include insight gained from data through machine learning and predictive analytics, and from the expertise extracted from subject matter experts
- The ability to measure the business quality of the decisions by means of simulations and experiments both against historical data and in real time deployment
- The ability to optimize the decisions based on the results of the decision analytics
- The ability to dynamically deploy the resulting decisions in production systems