Don't leave me hanging. Tell me what to do,
The Path to Prescriptive Analytics
EDW and BI concepts and topics have gravitated toward a more process-oriented approach and closed-loop analytics. Analytics must be driven back into other planning or execution systems. For example, market insights can power a statistical forecast used in demand planning that, in turn, drives supply chain execution.
How the insights themselves are derived is typically through age-old statistical and mathematical models applied to new sources of data (such as consumer behavior patterns captured by loyalty cards or Internet clickstreams). Simple statistic models can not only be applied to financials (such as revenue forecasting), but also to marketing research (such as product preference probabilities). Old artificial intelligence and expert system concepts such as fuzzy logic and neural networks have found traction reborn as part of data mining. These concepts have been refashioned for finding patterns in large data sets, using mountains of facts to support decision-making rather than expert judgment calls. New-to-business-planning concepts (such as genetic algorithms, cellular automata and chaos theory) promise to help model complex patterns (such as consumer or market behavior) that, in turn, can feed corporate planning or war-gaming tools.
Such applications promise to make planning more leading than lagging, and more scientific than guesstimate. In parallel, business contributions such as balance score-carding, performance-based budgeting, activity-based costing, and value-based management influences the nature of how enterprise planning and budgeting will be done.
Annual budgeting cycles will move to agile and adaptive plans to better react to the ever-changing business environments. Continuous planning cycles would call for the need for dynamic and adaptive planning models that could either be re-parameterized based on new business assumptions or re-modeled based on new understandings of trends or behaviors.
As BI and EDW applications have matured, evolutionary stages have developed that have given rise to the importance of integrating data mining and planning processes into BI. Various BI pundits have different takes on BI maturity models, but all have elements that differentiate between traditional reporting, descriptive modeling, predictive modeling, and prescriptive modeling. Traditional BI reporting answers the question of, “What happened?”, but offers no causes or relationships. Descriptive modeling addresses the question of, “Why did it happen?”, and is the basis for predictive modeling, which asks the question, ”What will happen?.”
Here is where SAP Integrated Planning leverages the EDW and BI infrastructure: BI models and tools can be re-parameterized and re-adjusted to predict potential outcomes and reveal causal insights. For example, a value driver model may predict the impact on profitability if customer service call center hours are cut. A prescriptive model is a heuristic or business rule that dictates how to “Make it happen.” A prescriptive model executes on (or institutionalizes) insights like putting diapers closer to beer on Friday nights, because data mining association analysis discovered a buying correlation. SAP-IP can also be fashioned into an optimizer itself, such as cutting projects within a program portfolio that do not deliver the best expected value.
