Shaping the Future with "What If" Analytics
Prescriptive analytics using simulation techniques can increase our knowledge and level of confidence in making informed decisions.
By Mark Peco, CBIP
[Editor's Note: Mark Peco is leading a session entitled Harness the Power of "What-If" Analytics: Shaping Your Future with Simulation at the TDWI World Conference in Boston (July 20-25, 2014).]
Prescriptive Information
Information that is prescriptive provides rules, instructions, and procedures to follow. It supports management's need to know what if. It clearly states what to do, how to do it, who should do it, and when it should be done. Prescriptive information connects decisions and activities with expected outcomes. Outcomes are associated with targets, goals, and objectives. Well-defined goals are proxies for a desired set of future conditions. Prescriptive information helps us shape our future by telling us what actions should be taken that will lead to desired future business conditions. This article outlines how prescriptive information is produced and what role analytics can play in the process.
The Rise of Analytics
Analytics is a common term used frequently in business discussions and has become embedded in modern business culture and vocabulary. Many business leaders believe it to be a foundation of their future competitive advantage. However, the term can be confusing. Analytics has many perspectives and capabilities depending on the context of the discussion and the users' experience. In its basic definition, analytics helps business professionals "connect the dots" in their environments and generate insights to drive decisions and actions with increased clarity and confidence. The ultimate goal of analytics is to improve business performance.
Business impact is sometimes elusive due to confusion, poor alignment, and unclear expectations. The term analytics is sometimes not specific enough when business opportunities and problems are discussed. There are several categories of analytics that address different types of information requirements.
Because the term is vague, it can lead to confused expectations in the minds of stakeholders. If business problems are poorly framed or the wrong approach is used, then business impact will likely be minimal. This leads to unhappy stakeholders.
Categories of Analytic Capabilities
Analytics can be segmented into five levels according to their capabilities.
| Level |
Analytic Category |
Business Information |
Analytic Capability |
| 1 |
Descriptive |
What is happening? |
Monitor and detect faults |
| 2 |
Diagnostic |
Why did it happen? |
Explain and troubleshoot |
| 3 |
Predictive |
What will likely happen? |
Estimate future conditions |
| 4 |
Prescriptive |
What should be done? |
Estimate results of actions |
| 5 |
Optimized |
What is the best way? |
Recommend the best action |
Each category is useful and provides different classes of information to the organization. If expectations are not well managed or the application area is poorly understood, it is possible to implement the wrong category and not produce any meaningful impact.
The first category provides descriptive information. This category depends on existing data to provide business monitoring capability. The basic components include available data, integration techniques, and visualization capability. Descriptive analytics is intended to monitor, analyze and control existing processes based on current operating conditions and resource allocation decisions.
The second level, diagnostic analytics, builds on the previous components and requires detailed data that can be analyzed using multi-dimensional methods such as OLAP, statistical methods and cause-and-effect rules. This technique supports investigation into the factors contributing to observed symptoms.
The third level is Predictive Analytics. This technique identifies patterns in existing data that are used to draw inferences about expected future behavior. Predictive analytics examines historical data to identify patterns and trends that are embedded in empirical models used to create predictions. Because these models are empirical, their scope is limited to existing data availability. These techniques predict future conditions assuming that the modeled patterns remain valid.
All three levels provide value and capability to an organization. However, they cannot advise a manager how to move the organization forward through a series of decisions that will shape the future. They are not connected to the management levers that create future conditions. Guiding the organization into the future is the role of prescriptive analytics.
Prescriptive Analytics
Prescriptive information provides action-oriented instructions. It is produced by prescriptive or "what if" analytics and it is enabled by simulation techniques. Simulation techniques have existed for many years in a variety of disciplines, including engineering, science, management, finance, and economics. Many terms are used to describe prescriptive analytics, including:
- Business simulation
- Process simulation
- Simulation analytics
- Prescriptive analytics
- Simulation modeling
- Prescriptive modeling
These approaches produce prescriptive information using simulation techniques. Their common features include system models, mathematical representations, decision variables, external input variables, constraints, and output variables.
Prescriptive analytics is based on the fundamental principle of developing and executing a system model implemented in software. System models are created using a variety of methods depending on the requirements. A system model is a representation of reality as defined by a business problem. System models have scope boundaries and they specify what input variables are transformed into output variables based on assumed rules of behavior.
System models are created by translating important features from reality into structural and functional perspectives. Structural models describe physical and logical connectivity of system components. Functional models describe how input variables are transformed into output variables based on existing domain knowledge. Functional models require rules of behavior to be expressed as equations or rules of thumb. The equations or rules of thumb are then solved by an algorithm that translates input variables into output variables.
To reasonably model real-world conditions, it is important to have a variety of modeling techniques available to deal with different types of conditions and behaviors found in the real business world. The following types of modeling techniques are useful for developing prescriptive analytics models:
- Continuous physical models based on laws of science
- Business process models based on policies, desired capabilities, and outcomes
- Stock and flow models based on accumulator and flow variables
- Monte Carlo models based on estimating uncertain variables using probability distributions
- Discrete Event models based on queuing concepts, resource allocations, and service levels
- Empirical models based on experiments, observations, and existing measurements
Providing Prescriptive Information Using Simulation
With simulation, analysts can operate system models within a software platform to generate prescriptive information. Virtual laboratories are created that enable business professionals to evaluate different planning and operating scenarios. The output from these simulations -- prescriptive information -- advises managers how to allocate resources and schedule workflows to meet their objectives.
Analysts manipulate decision variables and observe system performance. System models can work with a variety of interesting areas including business processes, physical plant, industrial processes, resource allocations, investment plans, organizational structures, market behavior, and supply chains. Experimental design techniques and research methods enable analysts to determine "cause and effect" patterns that can be shared with other stakeholders to generate improved knowledge and performance.
Such simulation lets analysts examine strategies, plans, and assets that do not already exist. Future-state conditions can be modeled that reflect changes to process design, capital investments, corporate re-structuring, new products. and other innovative ideas. Furthermore, simulation has the power to analyze and provide prescriptive information related to what if and what might be without being constrained by the past. It enables business professionals to be creative and test their ideas in a virtual world before committing major investments and changes in the real world.
Conclusion
The future is uncertain and its results cannot be guaranteed. There are too many factors that are unknown in terms of their magnitude and timing. However, the use of prescriptive analytics using simulation techniques provides an approach that increases our knowledge and level of confidence in making informed decisions that guide us towards our business objectives.
Mark Peco, CBIP, is a consultant and educator. He holds undergraduate and graduate degrees in engineering from the University of Waterloo. He is based near Toronto and is a faculty member of TDWI. He can be contacted at [email protected] .