In previous columns I have discussed the importance of focusing on operational decisions when applying advanced analytics. Most organizations have a huge opportunity if they can use advanced analytics to improve their day-to-day operational decision-making, not just strategic and management decisions. Improving these types of decisions can reduce fraud, improve risk management and increase customer satisfaction. Decision management has another key benefit, though: It’s an effective tool for organizations that need to respond in real time.

Operational decisions are made by call center staff, branch and office staff and by automated systems. These decisions must increasingly be made in real time. Customers don’t want to wait or move to a different channel to get a response; they expect you to respond in real time in the channel they have selected. With just-in-time supply chains, decisions about restocking and shipping need to be made instantaneously – not at the close of business or end of the week. As fraud attacks become more sophisticated, the potential for loss rises exponentially when you can’t detect and reject the fraud as it is perpetrated. Add to this the growth in automated channels and mobile devices, and it is clear that real-time responses are becoming the norm.

If we need real-time responses, we can’t really leave humans in the loop; humans are relatively poor at responding in real time. We don’t like to just wait and watch things, we require time to think and do analysis, and we like to take lunch breaks and go home. Our systems need to make more decisions so that they can respond for us.

Historically we have been reluctant to automate decisions. We restrict automated decision-making to simple, low value decisions. At the same time, most organizations have identified a set of complex, high value decisions that won’t be made in real time because we prefer to refer such decisions to experts. This leaves a gap where we expect operational decisions to be made very quickly by front-line workers. We expect these workers to make good decisions, but we can’t afford to use experts (and wouldn’t find enough experts even if we could afford it). If the staff does have the skills and experience to make good decisions, we increasingly can’t give them the time they require to do so. To respond in real time we need to address this gap.

We could use traditional technologies and approaches to automate these decisions. The risk, however, is that doing so will sacrifice accuracy of decision-making for speed. We must make sure our decisions are informed by our data; we must apply analytics to our decision-making, even as we automate our decision-making. This means moving beyond thinking of analytics as a way to present data to humans to improve their decision-making. It means thinking of analytics as a way to improve a decision, whether those decisions are made by people or embedded in systems.

This is where decision management comes in. Effectively applying technologies such as business rules management systems, data mining and predictive analytic models allows us to push the automation boundary. We can automate more complex, more valuable decisions. If we marry these technologies to the variety of new big data sources available to us we can handle even more sources of data. Additionally, when we automate these decisions we can better handle the volume and velocity of big data. Automating these decisions means we can respond in real time when these decisions are required for appropriate response.

Using decision management to respond in real time is about building Decision Management Systems: agile, analytic, adaptive systems that automate and improve our operational decisions. Developing these systems includes three elements:

  • Decision discovery. Decision discovery is all about beginning with the decision in mind. To automate our operational decisions we need to be clear exactly how those decisions should be made. This means identifying them and developing a model of how we expect to make them using proven decision modeling approaches such as the new Decision Model and Notation standard. We need to be clear what the decision-making elements are, what information is required and how our analytics are going to be applied.
  • Decision services. Decision services are service-oriented components that combine business rules and advanced analytics built in order to automate the decisions. These services can be invoked by new or existing processes and systems, delivering consistent (and consistently accurate) answers. In-database analytics, analytic model deployment and standards such as PMML are all critical to getting advanced analytics into these services.
  • Decision analysis. Finally, we need to close the loop. We need to be able to monitor our decision-making, see how it is working for us, and put in place continuous improvement and test-and-learn processes. This means collecting data from all the systems involved about how well decisions worked and analyzing it in a decision-centric way. It’s not enough to collect weblog data and analyze it to improve the website and collect customer service data to improve the CRM system. We need to integrate all the different data sources in a big data architecture in order to assess the true business impact of the decision-making.

Consumers and partners increasingly want us to respond in real time. To respond effectively, we need to make good decisions – data-driven decisions. Decision management lets us respond in real time by automating operational decisions while still putting our data, and our analytics, to work.