This article on prescriptive analytics is the fifth in a series of guest posts written by Dan Vesset, Group Vice President of the Analytics and Information Management market research and advisory practice at IDC.
Analytics solutions ultimately aim to provide better decision support — so that humans can make better decisions augmented by relevant information. Decision support capabilities can be segmented into five related categories, each of which is deployed to answer different types of questions:
- Planning analytics: What is our plan?
- Descriptive analytics: What happened?
- Diagnostic analytics: Why did it happen?
- Predictive analytics: What will happen next?
- Prescriptive analytics: What should be done about it?
In this series of blog posts, we’ll address each of these analytics capabilities. For a fuller introduction to the topic as a whole, see the first post in the series. This fifth post focuses on prescriptive analytics.
Prescriptive analytics: What should we do about it?
IDC research shows that as we move around the analytics cycle from planning, to descriptive, to diagnostic, and then to predictive analytics, the number of enterprises employing each analytics category decreases. However, this trend line won’t hold for long as more enterprises are discovering the power and value of predictive and prescriptive analytics.
Prescriptive analytics take the inputs from prediction and — combined with rules and constraint-based optimization — enable better decisions about what to do. The decision might be to send an automated task to a human decision maker along with a set of next action recommendations, or to send a precise next action command to another system.
Prescriptive analytics is therefore best suited for situations where constraints are precise. This usually happens with tactical choices, where many decisions need to be made within a given period. Examples of this include programmatic advertising buys, stock trading, and fraud detection. However, the universe of situations where prescriptive analytics are being applied continues to expand and will eventually permeate many types of decision making processes.
Broader adoption of prescriptive analytics is often hindered not by the functionality of prescriptive analytics solutions, but rather by external factors such as government regulation, market risk, or organizational behavior. . This is the situation in healthcare, for example, where some early successes exist but broad adoption of prescriptive analytics will still take years. Regardless of the timeframe for pervasive adoption of prescriptive analytics, every enterprise should begin to assess the applicability of this type of analytics for their own operations.
The functions that these tools provide fall broadly into three categories:
- Build a business case: Prescriptive analytics are best used when data-driven decision making goes beyond human capabilities, such as when there are too many input variables, or data volumes are high. A business case will help identify whether machine-generated recommendations are appropriate and trustworthy.
- Define rules: Prescriptive analytics require rules to be codified that can be applied to generate recommendations. Business rules thus need to be identified and actions defined for each possible outcome. Rules are decisions that are programmatically implemented in software. The system receives and analyzes data, then prescribes the next best course of action based on pre-determined parameters. Prescriptive models can be very complex to implement. Appropriate analytic techniques need to be applied to ensure that all possible outcomes are considered to prevent missteps. This includes application of optimization and other analytic techniques in conjunction with rules management.
- Test, Test, Test: Since the intent of prescriptive analytics is to automate the decision-making process, testing the models to ensure that they are providing meaningful recommendations is imperative to prevent costly mistakes.
While prescriptive analytics are dependent on rules, the most transformative power of prescriptive analytics along with the other four types of analytics is to “break the rules” – to use machine learning to identify new rules, monitor them over time, and adjust the rules as needed.
Once you’ve done that, you feed the rules back into your planning process so that you can begin the analytics cycle all over again, typically for a new budget period. It’s an iterative process to ensure you operationalize your analytics better and better over time.
In the foreseeable future, the combination of the five types of analytics will bring about a fundamental change in processes for software development, BI, analytics, data integration, and data management. The movement will be away from the deterministic, rules-dominated paradigm that has existed for decades, toward a paradigm that is constantly and automatically adapting based on continuous detection, analytics, decision, and action loops.
For IBM’s view on the Analytics Cycle, check out our smartpaper, “How Can You Trust Your Data Without the Big Picture?”