alphaspirit - Fotolia
Published: 06 Feb 2017
In the beginning, there was descriptive analytics -- data-parsing methodologies that cleverly analyzed large amounts of data about customers, products, financials or most anything else and yielded insightful new categories for those items. Predictive analytics then followed as an even more dazzling practice that could fine-tune our understanding of "what comes next" with great accuracy and granularity so we could maximize our time and investment in planning for the outcome we want.
Next in line is prescriptive analytics: the science of outcomes. It's less intuitive and much harder to embrace, yet it feeds the enterprise the kind of news we don't necessarily want to hear. Descriptive and predictive results simply provide better data for making decisions -- always a good thing -- and an important refinement of what is already happening. But prescriptive results take it a step further: They tell us what to do. That makes prescriptive at least as important as its siblings in moving the enterprise forward.
Prescriptive models don't just inform those involved in the decision-making process, they are the decision-making process. They articulate the best outcome, which can create friction among those who aren't comfortable relinquishing their decision-making responsibilities to a machine.
Playing by the (changing) rules
Prescriptive models also require careful framing, or rules, to produce outcomes according to the best interests of the business. When prescriptive analytics is applied, the process itself needs to include as much information as possible about the enterprise by creating a framework for interpreting the prescriptive results. That framework is built on business rules.
Business rules defining the enterprise's operations serve to gauge the impact of prescriptive recommendations on operations, efficiency and the bottom line. Projected outcomes are brought in line with institutional priorities, values and goals. The rules are based on established policy, best practices, internal and external constraints, local and global objectives, and so on. They determine to what degree prescriptive recommendations and anticipated outcomes truly work.
Constructing those rules may be an exhaustive, time-consuming and meticulous undertaking, requiring participation from all areas of an organization. Yet, the toughest work is still to come.
The rules must be dynamic; organic; and, to some degree, fluid. The entire point of an analytics-based institutional culture is acquiescence to the objective reality of real-world data. A corporate self-image based on that data will necessarily evolve. It follows that the business rules driving prescriptive analytics must also evolve. Therefore, the prescriptive process and the successful outcomes it delivers will feed back into the rules and steadily refine them.
An electronics manufacturer in southern Indiana put this idea to work in selecting its optimum long-term customer contracts. Though its headquarters are in the U.S., most of its actual manufacturing facilities are located on other continents. Capacity to manufacture and deliver in those other countries is governed by a number of risk factors involving fluctuating availability of raw materials, economic conditions affecting logistics and employee turnover. So the business rules applied to the company's contract evaluation process are critical to the accuracy of analysis and must be adjusted frequently.
Data inside and out
Another daunting challenge is hybridization of inputs into the prescriptive process. Descriptive and predictive processes use data that's carefully preformatted and well-thought-out. Prescriptive processes must model diverse facts, features and events from inside and outside the enterprise. That's called environmental data, and it can be messy because it's composed of unstructured and multi-sourced data that potentially includes everything from internet posts and video to free-form text based on speeches and white papers.
Codifying and classifying this diverse amount of data is cumbersome and expensive; perhaps the most off-putting of all are the prescriptive analytics components. Building processes to capture and format this kind of data can be viewed as a serious impediment to implementation. Yet, the task is essential. It can mean the difference between adequate and perfect modeling solutions.
The healthcare industry has been a leader in modeling prescriptive solutions with the environment. The service provider's need for efficiency is greater than ever because of the massive changes in healthcare economics in recent years. Capacity planning is a key factor in optimizing logistics and resources for service delivery. The models incorporate vast amounts of environmental data, including highly granular demographics, trends in health by region, and economic conditions at both national and regional levels. By using these models, many healthcare providers are adjusting near-term and long-term investment plans for optimal service delivery.
Prescriptive analytics closes the big data loop. It's a natural endpoint for the descriptive and predictive processes that precede it. Whatever the hype and hoopla surrounding prescriptive models, its success depends on a combination of mathematical innovation, mastery of data and old-fashioned hard work.
Predictive modeling from both business and technology perspectives
Finding the right skill set to build predictive analytics models
Forward-looking advice on prescriptive analytics management