Версия для печати
Прогнозная аналитика: что важнее – прецедент, причина или результат?5 июля 2012
Многие согласятся с тем, что любому бизнесу, любой индустрии и любому рынку стоит быть более дальновидными, уделять внимание прогнозной аналитике, составлять планы и по возможности не допускать ошибок, основываясь на предыдущем опыте. Прогнозная аналитика – логичное расширение функционала стандартных BI систем – может помочь сделать то, что традиционная BI ранее сделать не могла. Итак, что имеет значение? Читайте в материале Майка Уотшке, специалиста компании SAP. (Новость опубликована на английском языке)
The Business Case
Typically, predictive analytics derives from a perceived performance problem in sales, inventory, supply optimization, marketing, etc. However, a reactive approach rarely garners the results that a proactive approach does. So, can’t we simply foresee that we need predictive analytics from past reactive failures?
Sounds ridiculous when posed as a question, but the point is that we should all take a hard look at processes and results that we can improve given a strategic baseline (where do we want to be?) and the ability to predict. The business case for using predictive capabilities is it allows us to really understand what performance gains are reasonable to expect and what it will take to realize those improvements.
This is an interesting topic, because companies typically try to understand what process and other optimization issues exist prior to embarking on a predictive project. In reality, predictive analytics can proactively help you uncover hidden issues by exposing the data. Once you have the data at your disposal, you can span multiple dimensions/attributes, dates, etc. where anomalies and patterns exist – but at a level difficult to get to and understand using common reporting tools.
You can then apply predictive algorithms (statistical calculations) to the harvested data to gain visibility into process deficiencies and other performance problems. The key is to agree that there are organizational, operational, and other issues that you can expect to see real and measurable improvements, given you had a historical-based window into the future. In effect, you use the problem statement to justify a predictive project you can use to discover the real cause.
The most important aspect of predictive analytics is the effect—the end result, the performance improvements, and the competitive advantage gained in the marketplace. After all, if predictions and resulting actions to achieve the gains aren’t the end result, what good is predictive analytics in the first place? You must have responsive actions in place when you discover issues that will correct them and provide positive results. For every use case for predictive analytics, it’s wise to define a closed-loop resolution that moves the performance needle.
Bottom line: case, cause, and effect are all important, but providing a reasonable plan of action for performance improvement should be the ultimate goal of any predictive analytics project.