Increasingly, people in business intelligence and analytics circles are talking about prescriptive analytics. So what are they talking about? Simply put, prescriptive analytics provides users with the best options for dealing with given business situations based on the concept of optimizing the process of choosing between the available options. It lies at the high end of Gartner's analytics maturity model, which starts with descriptive analytics and progresses to diagnostic analytics and predictive analytics before finishing at the prescriptive level.
Both predictive and prescriptive analytics support proactive optimization of what is best in the future, based on a variety of scenarios. The problems businesses face are often quite complex and potentially can be addressed by taking multiple courses of action. The difference between the two approaches is that predictive analytics helps model future events, while prescriptive analysis aims to show users how different actions will affect business performance and point them toward the optimal choice. As data-driven organizations continue to recognize that information can provide strategic competitive advantages, more will strive toward the prescriptive end of the analytics spectrum.
But care is required: Although prescriptive analytics has exceptionally high business-impact potential, it can become overwhelming and complex rather quickly. Partly as a result of that, it remains an untapped opportunity in the vast majority of organizations. According to a recent Gartner report, only 3% of surveyed companies are currently using prescriptive analytics software, compared to 30% that are active users of predictive analytics tools. But with the continued explosion of data combined with vast improvements in technology, prescriptive analytics adoption is expected to grow substantially in the coming years.
Prescriptive approach throws a lot into the mix
Prescriptive analytics tools formulate optimizations of business outcomes by combining historical data, business rules, mathematical models, variables, constraints and machine-learning algorithms. Prescriptive analytics, much like its predictive cousin, is used in scenarios where there are too many options, variables, constraints and data points for the human mind to efficiently evaluate without assistance from technology. It is also used when experimenting in the real world would be prohibitively expensive or overly risky, or take too much time. Sophisticated analytical models and Monte Carlo simulations are run with known and randomized variables to recommend next steps, display if/then scenarios and gain a better understanding of the range of possible outcomes.
Some examples of business processes that prescriptive analytics is being applied to include pricing, inventory management, operational resource allocation, production planning, supply chain optimization, transportation and distribution planning, utility management, sales lead assignment, marketing mix optimization and financial planning. For example, airline ticket pricing systems use prescriptive analytics to sort through complex combinations of travel factors, demand levels and purchase timing to present potential passengers with prices designed to optimize profits but also not deter sales. Another highly visible case study example is UPS's application of prescriptive analytics in optimizing package delivery routes.
Prescriptive analytics applications have actually been around for quite some time. Business-school operations management courses usually cover one or more prescriptive analytics tools and techniques. The software options available today include Excel and Excel add-ins from vendors such as Frontline Systems as well as products from the likes of SAS, IBM, SAP, Tibco Software, MathWorks, Ayata, River Logic and KXEN (which is being acquired by SAP). In my experience, most prescriptive analytics professionals start with Excel; if the optimization problem that needs to be solved exceeds the base capabilities of Excel and add-in options, more advanced prescriptive analytics tools are sought out.
What's needed, and not needed, on prescriptive analytics
There are some common misconceptions that prescriptive analytics, and advanced analytics in general, require a data warehouse and a data scientist (or multiple data scientists). Although both are useful assets, there are other ways to tackle prescriptive analytics and the process of applying its simulation and optimization techniques to address key business challenges. To get started on learning more about what's involved, two good books to read are Spreadsheet Modeling & Decision Analysis by Cliff Ragsdale and Management Science: The Art of Modeling with Spreadsheets by Stephen Powell and Kenneth Baker. And while you don't have to be a data scientist, it is helpful to have an understanding of statistics.
The prescriptive analytics process is similar to the Cross Industry Standard Process for Data Mining (CRISP-DM). It begins with establishing a comprehensive description of the business process to be modeled, including defining the business objective and the variables, control factors and constraints to be analyzed. Often the model will evolve from an initial conceptual mental model to a visual, logical or mathematical model. In many cases, the definition phase alone introduces new questions and advantageous insights. Much like in predictive modeling, coming up with a complete and accurate definition of the business process to model and the prescriptive objective is critical for valid and actionable results.
The entire process is highly iterative in nature and requires close cooperation between analytics professionals and subject-matter experts in business units. Once a prescriptive analytics model is finalized, it can be used for manual decision making or embedded into operational systems to support automated, real-time decisions.
In future articles, I'll showcase and dive into specific use cases, tools, best practices and techniques for taking advantage of this new component in the analytics maturity model.
ABOUT THE AUTHOR
Jen Underwood is founder and principal consultant at Impact Analytix, a boutique business intelligence and predictive analytics consulting firm. She has more than 15 years of experience in the data warehousing, BI, reporting and predictive analytics industry. Email her at email@example.com.
This was first published in September 2013