Thought Leader Interview James Taylor: Automating Decision Making VP of Marketing at Fair Isaac Preface Dan Power, Editor of DSSResources.com, conducted this email interview in late July 2006. Q1: How did you get interested in decision support and especially decision automation? Taylor's Response: I have always been interested in how to make developing applications more intuitive and less dependent on large amounts of code. I worked with CASE and other modeling tools, with various 4GLs and on a cool meta-model-driven development environment in PeopleSoft’s R&D group. All of these things are interesting, but fundamentally, they assume that programmers should define how systems work. This is all well and good for many things in a system, but not really for business logic – the decisions within a system. These typically require more business know-how, and that makes it hard for programmers to manage them; programmers, after all, know how to program – not how to run the business. When I saw how business rules make it possible for business people to drive their systems by controlling the decision points within their systems, I was sold. Q2: How do you define the term decision support system? What is the overlap between your concept of enterprise decision management (EDM) and DSS? Taylor's Response: Companies face a wide range of decisions on a daily basis. These range from strategic decisions that have broad business scope but occur less frequently, to operational decisions that deal with individual transactions and occur with the highest volume. Strategic decisions might be those such as merger and acquisition decisions, whether or not to enter a new geographic region or add a store location. Operational decisions are going to be decisions like approve/decline this application, identify that this transaction is fraudulent, or decide how much of this claim to pay. In between, you are going to have a range of what you might call tactical decisions that determine how you will manage processes and customers. These might include determining which segments of a customer base will receive which offer, or figuring out what level of risk an enterprise will be allowed in accepting applicants. In my mind a Decision Support System, or DSS, is something designed to help people make a better decision – something to support (and hopefully improve) an individual’s decision-making capability. A DSS might support any kind of decision – strategic, tactical or operational. In reality, a DSS will most likely be focused on strategic and tactical decisions, as operational ones will need to be automated. With an automated decision, there is no person that can be supported and, hence, no decision support system. Enterprise Decision Management, in contrast, is focused on automating and improving operational decisions. In a sense, is it is about helping systems make better decisions. It is focused on high-volume operational decisions that must or should be automated. The overlap between DSS and EDM comes in the area of tactical decisions. If I am building a system to help someone make a decision quickly and effectively in a reasonably high-volume situation, then the kind of system I build (for example, one that combines generated suggestions with supporting information) blurs the line between a DSS and an EDM system. Q3: In general, what should managers know about business rules and decision automation? How can such tools be used to improve organization performance? Taylor's Response: The key thing managers should know about business rules and decision automation is that the technology works. Companies all over the world and in many industries are using business rules and other decision technologies to make a difference in their business. There are many ways you can use decision automation and business rules to improve organizational performance. These include: The separation of decision logic from mechanical implementation gives you more flexibility to make changes with minimal or zero impact on basic systems operation. Because the rules are managed and executed as an independent process called upon by your application, they can be updated at any time without requiring recompiles, links, or processing interruptions to your production application. Rule logic can be tested off-line and then introduced into a running application according to your strategy (scheduled, automatic or manual updates). Business rules are more understandable to business-level people, leading to better business/technical cooperation, reduced implementation times, and fewer opportunities for interpretation errors. Business rules syntax allows for a great deal of flexibility that you cannot replicate in traditional programming languages. Rules may be written with English phrases such as "If customer's age exceeds 65, then set customer's discount to .05". This type of familiarity lets business policymakers work side by side with the implementation team. Indeed, several of our customers have put business analysts through the rule writing course and had them writing rules, with no previous programming experience. Automating business decisions is often the key to delivering on the promise of straight-through processing. Many processes bog down at the point where an approval is required or where some other kind of risk/reward trade-off must be made. These steps are often those where decision support systems are developed. Automating these decisions so that 75, 85, even 95 percent of the transactions don’t need manual review can improve consistency and auditability, reduce costs, and speed processing. Focusing your staff’s expertise on the small percentage that cannot be automated also means that expert staff are applying their skills to the more complex, subtle transactions and being used more effectively. Decision automation allows you to move from performance monitoring to performance management. Most corporate performance management is based on dashboards, reports, key performance indicators (KPIs) and so on. This is really performance monitoring. What if the systems being so carefully instrumented to provide these great reports and dashboards were also easy to change? What if the people who see the dashboard and understand the importance of the KPIs could respond to them by logging on to a system and changing the way systems, and thus the company, behave? What if the reports that predict forthcoming problems had matching algorithms running automatically to drive change in systems? What if your alerts were able to tell you how your systems had responded rather than asking you to respond? This is performance management, and this takes decision automation. Improving performance often means making better decisions. Automating decisions with an EDM approach makes it easier and quicker to change decisions and enables better analytic refinement of those decisions. Q4: In general, what computerized decision support do you think managers need and want? Taylor's Response: I am not sure these are necessarily the same! Often what they want is more information about what is going on, tools to let them play with it, and Excel to make it easy to show others what they have found out or concluded. What they need are tools to do their job more effectively: They need to have routine decisions automated Not only will this mean that they don’t have to spend time doing mindless rubber-stamping, it also means that self-service systems can be developed that leverage these automated decisions, allowing customers to do more for themselves and perhaps empowering lower-lever staff to keep working with a customer as they can get the approval they need automatically. These are the kinds of systems developed using an EDM approach. They need to have an understanding of patterns The ebb-and-flow of normal operations needs to be analyzed and variations from that norm presented quickly and effectively to managers. They don’t need lots of reports telling them that nothing is unusual or that nothing has changed. They do need effective alerting and reporting when something unusual is going on. This is sometimes called Business Activity Monitoring and is clearly a form of decision support system. They need to be able to investigate exceptions effectively The percentage of decisions that cannot be automated must be understood and acted on. This means being able to drill into exceptions, find out why they are an exception (i.e. which rule fired to move them into the exception queue), analyze the situation, and make a decision. This is a kind of workflow/case management with the addition of good decision support tools to handle the decision-making for an individual transaction. They need to be able to productize patterns As managers discover different kinds of exceptions that can be handled in a repeatable fashion, they need to be able to put this handling into production. This requires both decision support systems to help identify and describe the patterns, as well as a decision automation environment designed for agility, so that it can be changed and evolved. What they need, in other words, is both EDM and DSS. Q5: Why should managers implement business rules and other decision automation technologies? Taylor's Response: Business rules deliver a unique benefit at a time when companies need to be increasingly agile in business while continuing to maintain and ensure compliance. They allow rapid and accurate operational implementation of new business strategies (i.e. new customer targets, policy changes, risk thresholds, new solution offerings) and external environmental factors (i.e. business climate, competition, interest rates, regulation). Business rules empower business users to automate and manage complex business rules with unparalleled, real-time control and agility. Predictive analytics – the other key decision automation technology – enables companies to turn their data assets into operational insight. There are really three main reasons to focus on decision automation technologies: Precision Precision is a measure of the effectiveness of a decision. Different decisions will require different ways to assess precision, but whatever is used should be focused on effectiveness, not efficiency, or on targeting. You may need to consider various financial outcomes such as profit, Customer Lifetime Value, revenue or losses, as well as the accuracy of predictions, comprehensiveness of factors involved, and the level of granularity achieved. Improving precision is a key reason for adopting decision automation technologies – especially those related to predictive analytics. Consistency Consistency measures how well integrated and coordinated your decisions are across your enterprise. Do you make the same decision, the same way, unless you mean not to? You can measure consistency over time: is today’s price the same as yesterday’s, across channels? Is the offer on the website the same as the offer made by the call center, and within and across product lines? Do I offer the same interest rate for different unsecured credit products? Highly consistent decisions need not be the same for all customers, all channels or over time, but the variations should be deliberate and designed, not incidental. Focusing on decisions as points of automation is vital. Agility Agility is a measure of how quickly, cheaply, and easily you can change the way you take a decision within your systems and organizational infrastructure. For example, if you want to introduce a new cross-sell strategy or a new pricing structure, how easy is it to change the systems specifications that support those decision strategies? How quickly would someone interacting with your organization notice that you had changed the way you wanted to make a decision? To measure agility, consider the total time and cost involved in moving from the point where you have the data that means you should change your decision process to the point where you have actually effected such a change. The higher this time and cost, the worse your agility. The use of business rules technology is largely driven by the need for agility, especially agility with consistency. It is also the case that speed–how quickly you can execute a decision – can drive the adoption of decision automation, as can cost reduction. By and large, automated decisions are cheaper and faster. If the volume of decisions is such that they are too costly to make by hand, or if the response time required is too hard to meet with manual decisions, then automation will add real value. Q6: How can automating and managing decisions make enterprise applications more effective? Taylor's Response: To quote the Butler Group, "Enterprise Applications tend to be pretty dumb. They collect data, store it and produce reports on it.” So, if your enterprise application(s) is/are dumb, what can you do about it? Look at the reports you generate. Talk to the people who read them (assuming someone does). Find out what prompts action on their part, and see if you can figure out the rules for taking the action. Have the system use the data to take the action for them; automate the decision that is taken when the report is reviewed. Look at the data you have. Talk to someone who understands data mining or predictive analytics. See what they could predict based on your data. Decide if any of this would be useful in running the business. Better yet, talk to your business users and see if they think it would be useful. See if you could improve a decision being taken by leveraging your data. Gather input from business users. While you are talking to your business users, ask them what they wish they knew. Maybe you can find a way to derive it from the data you have. Ask them what they would do if they knew that, and see if you can automate the action too (see above). Look at the way you collect data. Talk to someone who uses the data. Do they get all the data they need the first time, or do they have to go back and ask for more? What other data do they want and why/when do they want it (different circumstances are likely to make them want different data). See if you can figure out the rules that would let you collect the data they need (but no other data) the first time. Read the procedures or cheat sheets your users have. Do they work around something in the system? If so, can you figure out which rules are wrong in the system, because some are (otherwise they wouldn't need to cheat). How easy is it to change these rules? If it is not easy, could (should) you externalize those rules so you could manage and change them more easily? Look at the change logs. Are there pieces of the system that are always being changed? How much time and effort does this take? Is IT behind on these changes, or do they take so long that the business has stopped asking? Perhaps you could externalize this part of the application and make it possible for the business to change these rules themselves. There's more, but you get the gist. Automating and managing decisions can make your Enterprise Applications much less dumb. Q7: What is the potential for decision services, business rules and predictive analytics? Also, please explain these concepts. Taylor's Response: Decision services, built with business rules and predictive analytics, have the potential to bring the power of business intelligence to bear on operational processes. They can greatly reduce the maintenance tail and, hence, the Total Cost of Ownership (TCO) of complex information systems by making the business logic easy to modify. They can empower customers to self-serve and enhance the effectiveness of customer-facing staff by dramatically reducing the need to refer decisions to managers. They can make managers more effective by letting them focus on the overall business rather than the mechanics of day-to-day decisions. Decision Services Decision services are services within your Service Oriented Architecture (SOA) that automate and manage highly targeted operational decisions. A Decision Service isolates the logic behind operational business decisions, separating it from procedural application code. Decision Services also eliminate the time, cost, and technical risk of trying to simultaneously reprogram multiple individual systems to keep up with changing business requirements and make it much easier to improve a decision using analytics. By isolating the changes and improvements to decision-making inside a single component, you can increase your return on investment by leveraging the best decision you know how to make everywhere. Business Rules Business rules management systems enable the design, deployment, and maintenance of business rules and policies. These systems—also known as business rule engines and decision engines—put control into the hands of business users, allowing them to build and revise rules without IT support. Typically, a business rules management system is invoked as part of the total processing involved in an interaction with a customer. The system accesses and processes relevant transactional and historical data, uses predictive models and other analytics to segment customer populations for targeted action, executes business rules appropriate for the specifics of the customer and transaction at hand, and returns decisions to the production system or business staff. Using these systems gives business managers increased control and visibility over the factors used in each business decision. A business rule is defined, reviewed, modified, and reused as a corporate asset. The ability to manage these control points independent of the computer programming code that runs the automated applications allows for faster and more accurate review of business operations and implementation of changes as quickly and as often as necessary. Business rules management systems enable businesses to make highly consistent, strategy-driven decisions, and to change them with greater agility. Predictive Analytics Predictive analytics is a proven technology that’s been used successfully for decades. It encompasses a variety of mathematical techniques that derive insight from data with one clear-cut goal: find the best action for a given situation. Predictive analytics simplifies data to amplify its value, finding patterns that can guide decisions. When considering hundreds – or even thousands – of factors, and a universe of thousands or millions of customers, people just can't "connect the dots" to make the ideal decision. Predictive analytics connects the dots scientifically, guiding each decision to greater success. Simply put, predictive analytics is the science that makes decisions smarter. Predictive analytics applies diverse disciplines such as probability and statistics, machine learning, artificial intelligence, and computer science to business problems. Q8: How can decision automation and decision support systems improve agility in organizations? You've noted in your blog that Gartner defines agility as "the ability of an organization to sense environmental change and to respond efficiently and effectively to that change". Taylor's Response: Gartner’s perspective shows exactly why both decision automation and decision support systems are required. In order to “sense” environmental change you need decision support systems. Ideally, you would automate some of the sensing, with Complex Event Processing or Business Activity Monitoring systems. But even then, you are likely to have a set of situations that will require a person to do some analysis. That’s where the DSS comes in – helping them understand what change has happened. In order to respond effectively, you will often need decision automation, as decisions are often the key element of change. For instance, especially in your core business, your process is unlikely to change in response to anything but the most tectonic shifts in your market. The more typical environmental change is not going to change your business, but it is going to make you want to refine pricing, re-target offers, segment customers slightly differently, etc. These decisions might change constantly in response to variations in response rates, competitors, market conditions, the weather, and so on. Decision automation and management are key to being able to change these decisions quickly: The separation of decisions from processes and systems makes changing and managing them easier and allows them to be reused across multiple channels, systems or processes without any decrease in agility. The use of a business rules management system (BRMS) makes it easier to change these decisions rapidly by taking a declarative approach to the rules and by enabling them to be changed without a massive recoding effort for the main system. A BRMS empowers business users to make some or all of these changes themselves and so removes IT from the critical path increasing responsiveness. Having a single point of automation makes it easier to apply analytic insight in the shape of new segmentation or predictive models. Agility will require DSS and EDM – the one to sense the need for change, the other to make the change effective across your applications in a timely manner. Q9: What do you see as major trends in computerized decision support? Where are we headed? Taylor's Response:I think we will see more and more automation of decisions and more sophisticated support of decision-making. I think we will move away from thinking that decision support means reports or eye-candy, and towards thinking about the decisions being taken and what information technology can do to improve them. In particular: I believe we will see more of a focus on predicting the future, and less and less on reporting the past. Better prediction and pattern identification will make it easier to automate decisions and make it possible for knowledge-workers to apply their skills in the context of what the data tells them is likely to happen. More trends, more what-ifs, less reporting. I firmly believe there will be more automation of response, both to customers and to business events. The growth of the Internet, multiple mobile channels, and tech-savvy consumers means that more of the things consumers do will have to be responded to at once, and done so intelligently. This will drive additional automation of decision-making, as well as a need for more sophistication in decision-making. This represents a change from focusing on how data can help managers, to focusing on how data can help customers directly and those who interact with them. I think the ongoing investments in integrated information (CDI/MDM or Electronic Health Records etc.) will mean that better decisions become possible. While presenting a 360 degree of the customer to someone is useful, what companies really want is for everyone and everything with which the customer interacts to use the full view to make better decisions. The trend towards better management of the complete information picture will reinforce and overlap with the trend to automate more decisions. Simulation and what-if analysis of changes in the way a decision is made will become something that individual managers can control. They will have the ability to see what the impact of a change is likely to be before making it and the power to simulate various potential scenarios to see what strategies should be in place to cope well in advance. Knowledge workers will really be able to use their knowledge to manage the business. More and better analysis of both structured and unstructured text, new analytic techniques, and so on will mean we can draw more effective and more insightful conclusions from our data than ever before. About James Taylor James Taylor is vice president of product marketing for Fair Isaac’s Enterprise Decision Management Technologies, where he is responsible for working with clients to identify and bring to market advanced decision management solutions that will better solve their business needs. Taylor is widely considered as one of the leading experts and visionaries in enterprise decision management and he authors a blog on related subjects at edmblog.com. Citation Power, D., "James Taylor Interview: Automating Decision Making", DSSResources.COM, 10/6/2006.