Prescriptive analytics is touted as being a form of ‘uber-analytics’ – the final frontier. It can loosely be defined as a workflow that takes a business from data to insight, to hypothesis, to action. Well, what the hell does all that mean? In a nutshell it means prescriptive analytics not only tells us what might happen in the future, as predictive analytics does, but it tells us how things should happen for the best outcome.
Everything starts with data, and depending on what we want to achieve, we use data to get a view on how things have turned out in the past, how they might turn out in the future, and how we should best deploy resources to maximize future returns. These correspond with the three major forms of analysis:
- Business Intelligence (BI) gives us a look in the rear view mirror. Past and present activities can be analyzed ad-nauseam and displayed on charts, graphs, maps, tables and dashboards. This activity is generally known as descriptive analytics and currently forms the major part of the analytic activity in most businesses.
- Predictive analytics is primarily concerned with the future. It uses data mining and machine learning technologies to find patterns in data which might predict how events will turn out in the future. A good example is predicting the outcome of a bank loan. The analysis of historical data might show that someone who has two mortgages and is over 50 years of age is a bad risk.
- Discovery analytics has been recently introduced as a name to describe the act of throwing a pile of data at an algorithm and asking it to show us something we didn’t already know. A classic example is using cluster analysis to segment customers in ways are new and novel. Discovery analytics is not really separate from predictive analytics in some ways, since we are looking for insights that might guide future activities.
- Prescriptive analytics is closely identified with optimization – but has been extended to mean much more than this. Optimization means we can deploy resources to the best effect given the constraints under which a business operates. If predictive analytics has given us an insight into future activities, prescriptive analytics can use these predictions to optimize future operations.
Prescriptive analytics is increasingly meant as the complete workflow from data to optimized operations – this is what we want after all. All the forms of analysis mentioned above contribute to this, but we also have to include business process management (BPM), since for most purposes this is the end point. So it should be clear that this broad definition of prescriptive analytics is something to aim at, and not a living reality in most businesses.
The word ‘prescriptive’ implies prescribed actions – what to do, what not to do, when to take action, who should take action, and how to deal with variations and create contingencies. Prescriptive analytics uses the facts that are derived from BI, and the hypotheses derived from predictive analytics. We need hypotheses to deal with the future, and need to ensure that the hypotheses derived from predictive analytics are accurate. This is a major undertaking in itself, and new territory for most businesses. Prescriptive analytics can be undertaken at the enterprise level, or at the departmental level – or even more finely grained. Either way a couple of basic workflows need to be executed:
- Prescriptive analytics design – collect data, analyze data, create hypotheses, optimize operations, create business processes.
- Prescriptive analytics operations – capture data, detect events, initiate actions.
Obviously this is much simplified, but it does convey the gist. The prescriptive analytics design needs to be revisited regularly, and this will involve revisiting predictive analysis to ensure hypotheses are still relevant, re-optimizing to make sure resources are being used in the best way possible, and modifying business processes when needed.
It’s a big ask, and for most businesses a future. However it does give something to pitch at, and gives an overall shape for the analytical activities in a business. Right now we are infatuated with all things visual. But seeing something on a dashboard or chart doesn’t mean that we understand why things are as they are, or that we can do something about it. Prescriptive analytics provides the framework to address both these issues.
Some of the players in this market include:
FICO – broad, integrated suite of technologies for prescriptive analytics.
IBM – pretty much a complete solution.
River Logic -mostly focused on optimization.
SAS – addresses most areas of prescriptive analytics.