For years analytics has been a field of descriptions and predictions. Analysts ask “What insights can come from this data?” Or “What is most likely to happen next?” The next frontier of analytics is prescriptive: “What’s the best course of action?”
The first two tasks can be addressed through statistics and machine learning. With enough time and resources an accomplished data scientist can tune and fit a model that describes a data set and can adequately predict an outcome in new data.
The Value of Prescriptive
What if your doctor told you that you were in very poor health and you weren’t likely to survive the next six months? Would that be enough information? Probably not. You would want to know why the prognosis was so bad. And you would want to know everything you could do to change course.
With any complex task, from self-driving cars to predictive maintenance, it’s not enough to see it coming. The critical component isn’t the impending accident or a machine overheating, but what can be done to avoid it. And that’s the real magic of prescriptive analytics. Not just knowing the likelihood of an event, but also the factors behind that prediction.
But the hurdle many are facing today is that prescriptive outputs aren’t part of the common statistical and machine-learning toolkits. Just as algorithmic transparency isn’t part of these tools, those same results are critical to knowing why a prediction was made, and what can be done about it.
Alternatively, predictions can be transparent and prescriptive, like we get from EmcienPatterns. Here the impact of each variable is built directly into the prediction itself, and even attached to the outcome in real time.
With these prescriptive markers, along with the prediction APIs, have made it possible for Emcien customers to plug intelligence directly into their existing infrastructure for end to end automation. From intelligent CRMs to automated maintenance alerts, the key to prescriptive analytics is knowing why.