For Interview Wednesday, today we hear from James Taylor, CEO of Decision Management Solutions in Palo Alto, California. Email him firstname.lastname@example.org, or follow on Twitter @jamet123. James' work epitomizes the mature use of evidence: Developing decision processes, figuring out ahead of time what evidence is required for a particular type of decision, then continually refining that to improve outcomes. I'm fond of saying "create a decision culture, not a data culture" and decision management is the fundamental step toward that. One of the interesting things he does is show people how to apply decision modeling. Of course we can't always do this, because our decisions aren't routine/repeatable enough, and we lack evidence – although I believe we could achieve something more meaningful in the middle ground, somewhere between establishing hard business rules and handling every strategic decision as a one-off process. But enough about me, let's hear from James.
#1. How did you get into the decision-making field, and what types of decisions do you help people make?
I started off in software as a product manager. Having worked on several products that needed to embed decision-making using business rules, I decided to join a company whose product was a business rules management system or BRMS. While there are many things you can do with business rules and with a BRMS, automating decision-making is where they really shine. That got me started but then we were acquired by HNC and then FICO – companies with a long history of using advanced predictive analytics as well as business rules to automate and manage credit risk decisions.
That brought me squarely into the analytics space and led me to the realization that Decision Management – the identification, automation, and management of high volume operational decisions – was a specific and very high-value way to apply analytics. That was about 11 or 12 years ago and I have been working in Decision Management and helping people build Decision Management Systems ever since. The specific kinds of decisions that are my primary focus are repeatable, operational decisions made at the front line of an organization in very high volume. These decisions, often about a single customer or a single transaction, are generally wholly or partly automated as you would expect. They range from decisions about credit or delivery risk to fraud detection, from approvals and eligibility decisions to next best action and pricing decisions. Often our focus is not so much on helping a person MAKE these decisions as helping them manage the SYSTEM that makes them.
#2. There's no shortage of barriers to better decision-making (problems with data, process, technology, critical thinking, feedback, and so on). Where does your work contribute most to breaking down these barriers?
I think there are really three areas – the focus on operational decisions, the use of business rules and predictive analytics as a pair, and the use of decision modeling. The first is simply identifying that these decisions matter and that analytics and other technology can be applied to automate, manage, and improve them. Many organizations think that only executives or managers make decisions and so neglect to improve the decision-making of their call center staff, their retail staff, their website, their mobile application etc.
The ROI on improving these decisions is high because although each decision is small, the cumulative effect is large because these decisions are made so often. The second is in recognizing business rules and predictive analytics as a pair of technologies. Business rules allow policies, regulations, and best practices to be applied rigorously while maintaining the agility to change them when necessary. They also act as a great platform for using predictive analytics, allowing you to define what to DO based on the prediction. Decision Management focuses on using them together to be analytically prescriptive. The third is in using decision modeling as a way to specify decision requirements. This helps identify the analytics, business rules and data required to make and improve the decision. It allows organizations to be explicit about the decision making they need and to create a framework for continuous improvement and innovation.
#3. One challenge with delivering findings is that people don't always see how or where they might be applied. In your experience, what are some effective ways to encourage the use of data in decision making?
Two things really seem to help. The first is mapping decisions to metrics, showing executives and managers that these low-level decisions contribute to hitting key metrics and performance indicators. If, for instance, I care about my level of customer retention, then all sorts of decisions made at the front line (what retention offer to make, what renewal price to offer, service configuration recommendations) make a difference. If I don't manage those decisions then I am not managing that metric. Once they make this connection they are very keen to improve the quality of these decisions and that leads naturally to using analytics to improve them.
Unlike executive or management decisions there is a sense that the "decision makers" for these decisions don't have much experience and it is easier therefore to present analytics as a critical tool for better decisions. The second is to model out the decision. Working with teams to define how the decision is made, and how they would like it to be made, often makes them realize how poorly defined it has been historically and how little control they have over it. Understanding the decision and then asking "if only" questions – "we could make a better decision if only we knew…" – make the role of analytics clear and the value of data flows naturally from this. Metrics are influenced by decisions, decisions are improved by analytics, those analytics require data. This creates a chain of business value.
#4. On a scale of 1 to 10, how would you rate the "state of the evidence" in your field? Where 1 = weak (data aren't available to show what works and what doesn't), and 10 = mature (people have comprehensive evidence to inform their decision-making, in the right place at the right time).
This is a tricky question to answer. In some areas, like credit risk, the approach is very mature. Few decisions about approving a loan or extending a credit card limit are made without analytics and evidence. Elsewhere it is pretty nascent. I would say 10 in credit risk and fraud, 5-6 in marketing and customer treatment, and mostly 1-2 elsewhere.
#5. What do you want your legacy to be?
From a work perspective I would like people to take their operational decisions as seriously as they take their strategic, tactical, executive and managerial decisions. Thanks, James. Great explanation of creating value with evidence by stepping back and designing a mature decision process.