Practitioners in our field have long been evangelizing on the critical link between decision management and predictive analytics. As James Taylor accurately and succinctly stated “Decision Management operationalizes predictive analytics. Traditional approaches to analytics are hard to scale and hard to use in the real-time environment required in modern enterprise architectures.”
On cue I noted with great interest several writers predicting analytics trends for 2016. These included:
- Data streaming. Large amounts of data coming in can’t just be gathered and analyzed; instead it must be gathered, analyzed, and acted upon in real time.
- Lifestyle analytics. Predictive analytics integrated into our everyday life by understanding decisions we make every day.
- Data governance. We’re collecting so much data without discretion that we must employ intelligence on what is stored and how we care for it to avoid possible breaches. It’s very easy to steal a needle from a haystack but not so easy to take a spoon from an organized utensil drawer.
- Black box algorithms. Overcomplicated, hidden algorithms peddled upon unsuspecting organizations with no one qualified to evaluate, understand or use them properly can lead to a plethora of problems. The reliance on “the one data expert guy” is just too dangerous.
- Connecting insight to action. This is perhaps most important of all for our discussion here. It’s wonderful to collect and view all these insights from vast amounts of data. But how do we make it truly actionable?
All of these predicted trends point very clearly to a need for all the tools we discuss in decision management – both in terms of analysis and implementation. That said, our experience has repeatedly shown that most enterprises either don’t agree with or, more likely, don’t understand the connection. I regularly hear predictive analytics in one discussion and decision management in another with no apparent consideration for linking the two. This is partly due to some organizations slowly realizing decision management is more than “just another name for rules”. It’s been estimated that 2 out of every 3 companies fail to fully take advantage of their data and what it might provide them. They simply do not operationalize their decisions.
At the same time, these same organizations have realized the limits of dashboards and now seek to do something with that data besides look at different views. Once decision management is introduced as both a methodology and technology – one that can create a common understanding between all parties – we have traction to present new and powerful approaches to predictive analytics. This in turn will allow the newly agile enterprise to drive their decision logic.
The purpose of this missive is not to introduce or explain technical approaches and detailed studies of platforms and integration. Instead I only want to suggest ways to get these key ideas of decision management with predictive analytics out to facilitate their introduction to these willing corporations.
The first and most obvious benefit provided by the combination of decision management and predictive analytics is the collaboration inherently resulting from the former. No longer are we dependent on the mad data scientist in a white lab coat! Now every concerned group – business analyst, IT, business owner and the data scientists can all view, understand and work with the same information. This removes the ubiquitous black box and enables truly effective data governance.
But let’s face it – what we really want most of all is to create something actionable from all that data. We know more and more of it will be coming in real time. By beginning to incorporate additional elements of decision management while collaborating across all stakeholders, we can create (or significantly enhance) the ability to:
- Understand what the data represents • Ensure the quality of the data
- Develop models from the data
- Visualize the data in more powerful and understandable ways
- Extract the insights
- Model the decisions
With these tools, the opportunities for making data actionable are plentiful. Predictive models can be validated and moved to production quickly. Historical data can be used by business analysts to create business rule use cases. Systems can be integrated with other predictive models to provide real time analysis on constant data streams. New models, decisions and rules can be created, tested and deployed very quickly.
The potential applications areas for these innovative hybrid approaches are very exciting. In the financial services and mortgage world we can foresee powerful tools to assess portfolio risk, rapidly validate compliance and understand the impact of potential changes, provide default prediction, consistently correlate risk versus price and have real insights into strong or weak areas of the business. Lifestyle analytics may provide even more accurate models to evaluate the behavior of mortgage applicants. Poring through the volumes of historical loan data may yield new valuable insights to mortgage fraud that can quickly be deployed in an actionable manner.
The tools are there and ready to use. While each is powerful and effective alone, the combination of predictive analytics with decision management can help us reach new levels of automation.
