Predictive Analytics In The World of Decision Management: Interview With James Taylor
James Taylor, CEO and Principle Consultant of Decision Management Solutions, is a leading expert on using business rules and analytics technology for decision management. James has authored several books, including Decision Management Systems: A Practical Guide to Using Business Rules and Predictive Analytics, and runs his blog “JT On EDM.” We caught up with James about the role of predictive analytics in decision management.
Tell us about yourself. What are your passions? What are you most intrigued by today? What’s something people would be surprised to hear about you?
I run a specialist consulting firm, Decision Management Solutions, providing consulting to companies adopting and using decision management technologies, such as business rules and predictive analytics, as well as advice to vendors in the space. I’m passionate about decision management and about decision modeling. I’m particularly intrigued by the potential for decision modeling to improve the way we apply predictive analytics to business problems. Personally I’m a keen reader of military and alternative history and people who have not heard me speak might be surprised to hear I grew up in England, though I have lived in the US for 25 years.
Give us a landscape overview, please: How is analytics and predictive analytics changing decision management?
Predictive analytics has always been a part of decision management – the original idea back in 2002 was as a way to describe the use of business rules, predictive analytics and optimization technology in combination to develop powerful systems that automated and managed operational decision-making. The last few years have seen two major shifts, however.
First, the acceptance of predictive analytics as a key component of a business strategy means that many people are approaching decision management first and foremost as a way to deploy and use predictive analytics. Previously the emphasis was probably slightly more toward using it as a way to successfully deploy business rules with predictive analytics to improve the quality of decision-making.
Second, the explosion of interest in big data means that more organizations are thinking about analytics on streaming data, about capturing event streams and using them for analytics, and about using non-traditional data in analytics. All of these add complexity to a decision management solution while also reinforcing the value of decision management in operationalizing analytics.
What business area is predictive analytics affecting the most? What benefits do businesses stand to gain from adopting analytic-focused methodologies?
Decision Management grew up in risk and fraud, especially credit risk and credit fraud. When the technologies were harder to use and more expensive, these were the decision areas where the ROI was greatest. As the cost of developing decision management systems has dropped, and the interest in predictive analytics has broadened, use cases have widened. Now marketing and sales, especially personalizing/targeting a “market of 1,” is one of the biggest areas across the board, with more fine-grained supply chain management and automated operations coming up fast. Analytics and decision management have value to offer any time an organization must make a decision many times in a repeatable fashion where that decision is non-trivial and where historical and other data can be used to make a more precise, more valuable decision.
Can you talk a little bit about data-driven micro-risk management, specifically for sales and marketing functions? (Referencing your article here.) What’s in it for sales and marketing?
The central tenant here is that each decision (about a single customer, a single transaction) should be made just for that customer or transaction – this was what Neil Raden and I call Micro Decisions in our book (Smart (Enough) Systems, Prentice Hall 2007). Predictive analytics and data mining let an organization divide customers and prospects into increasingly fine segments. Additional analytics allow the risk (say, retention risk) and opportunity of each customer to be predicted. These analytics can be wrapped with business rules to make a decision just for this customer – what offer should I make and how should I present it to maximize the likelihood that this specific customer will decide against cancelling their subscription? How should I describe this product to make it appealing to this customer?
Evidence shows that this kind of focus is practical and very effective. After all, people respond to your decisions about them as though you were only talking to them – they don’t care that you made the same decision for lots of other people. So, if they are going to respond personally, why not decide equally personally?
Where —either specific functions, companies, or industries—do you see “big data” and analytics having the most impact?
Right now it’s in customer treatment – how to make micro decisions about a specific customer or prospect as discussed above. I think with the internet of things and more streaming, event-based data is going to drive decision management and big data analytics into operations and supply chains ever more deeply. Plus, of course, it still really matters in risk management and fraud prevention. Historically, financial services, especially retail banking, was the core area. Insurance, primarily property and casualty, as well as consumer facing organizations like telcos and retailers, are gaining ground quickly.
How has your book Decision Management Systems: A Practical Guide to Using Business Rules and Predictive Analytics changed the way people are thinking about predictive analytics and the purposes it can serve? What do you hope to accomplish with this book? What changes in thinking or behaviors would you like it to lead to?
I think the books main value is to establish that these kinds of “smart” systems are practical now and practical for large enterprises – there’s no need to wait for the labs at MIT to come up with some new AI package. Nor is this something that only innovators and start-ups can do. Decision Management Systems are built with reliable, proven technology (data mining, predictive analytics, solvers, business rules management systems) using a straightforward methodology (discover and model decisions, build decision services, establish ongoing decision analysis for continuous improvement). They work, they generate a great ROI and they can be added to a legacy application portfolio.
As previously mentioned, I also think the recent increasing focus on analytics has shifted to why people buy and use both the book and the technology it describes. Now, the focus is often on deploying predictive analytics – people have the analytics, they just don’t see how they can use them to make a day-to-day difference. Decision management systems are the key to solving this dilemma.
What are some of the emerging tools and technologies that have you the most excited this year?
I am really excited about decision modeling. We and others have been using decision modeling to manage business rules development and design decision management systems for a while. Recently we were part of the development of a new standard – the Decision Model and Notation standard – that will bring a common notation for these models. We have found that decision models built using this notation are a fantastic way to specify requirements and frame the problem for predictive analytics – to develop what analytic methodologies like CRISP-DM call business understanding. Having a model of the decision you need to improve, understanding which business processes need those decisions made and which key performance indicators will be affected by improving it puts analytics into a clear business focus. I believe this is really going to change the way people do analysis for predictive analytics and eventually all analytics efforts. I recently published a book on decision modeling – working with Tom Debevoise we published The MicroGuide to Process and Decision Modeling in BPMN/DMN: Building More Effective Processes by Integrating Process Modeling with Decision Modeling and we are seeing significant interest in our decision modeling software, DecisionsFirst Modeler, in analytic teams.
What tools are “must haves” for enabling modern decision management?
1: A willingness to focus explicitly on decisions, as well as processes and data. Identify them, describe them, and model them (ideally using the new Decision Model and Notation standard).
2: Data mining and predictive analytic technology that allows an organization to turn its data (and the data it can buy) into insight that will help it make these decisions more accurately.
3: Business rules technology that ensures the logic that wraps around analytics, as well as the logic required by policy and regulation, and can be managed and updated so it is up to date and effective.
4: Performance management technology to track decision outcomes and match these outcomes to the business results achieved as part of a continuous feedback loop.
You can also buy packages that embed all this into a coherent solution for a particular business decision or set of business decisions too.
What do you think is required for wider company adoption rate of analytics/data-driven processes? How can we make the shift (or are you already seeing this shift?) from business leaders viewing analytics as something that’s “nice to have” vs. “necessary to my business?”
I think wider company adoption rate of analytics calls for two things . The first is that organizations need to see the value of analytics and data-driven processes in their day-to-day operations. High volume, transactional, operational processes are what generate all the data – and, these are by far the most effective place to apply analytics. We have to break the focus on strategic and knowledge worker analytics and focus instead on operationalizing analytics.
The other came up when I spoke to Tom Davenport for this Wall Street Journal column – Smart Machines and the Decisions They Support. The challenge for most organizations is that “They don’t have a set of decisions from which to choose in thinking about [analytic] technologies, and they haven’t thought carefully about each decision to which they want to apply technology.” This is why I am so excited about decision modeling. Process modeling helped move organizations to a place where they understood their business processes and could begin to systematically improve them. Decision modeling can and I believe will do the same for decision making and analytics.