Blogs

Next Best Action in Unprecedented Circumstances: When No One Truly Knows the Next Best Action

Big Data Evangelist, IBM

Optimality is the new nirvana. The promise of "next best action" is that, somehow, we can program the optimal automated response into every business scenario. Of course, this dream presupposes that someone in your organization can specify the optimal response for any scenario that your personnel are likely to confront.

Good luck with that. The real world constantly presents you with unprecedented situations. These, by their very nature, suggest no clear course of action and fall between the cracks in your standard operating procedures.

What's an unprecedented circumstance? Most real-world organizations encounter those constantly, especially if you operate in a dynamic industry. Has a high-priority customer ever fallen into a gray area between different segments, rather than into one or another specific category around which you've built your account management procedures? Has a customer ever run into a showstopping product issue that you've never seen before? Have you ever had customers who gripe about anomalies they experience when using one or more of your products in combination with one or more third-party products? And so on and so forth.

Next-best-action business processes are simply codified precedent. You can write future-proof business logic - segmentation models, predictive analytics, business rules, orchestrations, etc. - only for those processes that you are 100 percent sure will never run into unprecedented circumstances. And that, as any business process professional will surely tell you, will never happen. Sometimes you're simply stuck until the requisite manager or internal expert renders judgment on what's best to do. That's why business processes are always designed with exception-handling procedures - alerts, notifications, escalations, decision support interfaces, and the like - as a key element.

Human judgment will always be needed when you're pushing next-best-action platforms, such as IBM SPSS Decision Management, into uncharted circumstances. Advanced analytics tools can help process participants, individually and collectively, identify the next best steps.

Some of the key enablers of a human-assisted next-best-action process include:

  • Machine-learning algorithms: These automate the process of finding key variables and dependencies in fresh empirical data. Available in tools such as IBM SPSS Modeler, machine learning algorithms automatically learn to recognize complex patterns and make intelligent decisions based on new data, above and beyond the training data upon which traditional statistical models are built. Machine-learning technologies can help your data scientists to quickly identify optimal response behaviors to incorporate into the embedded models, rules, and other process logic that drives next best action.
  • Deep Question & Answer (DeepQA) technology: This capability, which powers IBM Watson, can help decision makers rapidly identify the next best actions in unprecedented situations with a high degree of confidence. One of the hallmarks of unprecedented situations is that the problem domain is often unstructured, as is the massive body of information relevant to the decision at hand. DeepQA--which combines machine learning, natural language processing, semantic analysis, and other advanced knowledge management technologies--can provide a real-time lifeline for somebody who's on the front line and needs a good-enough answer at this very moment.
  • Self-service business intelligence (BI) tools: These can help business analysts to rapidly sort through all the metrics, trends, and scenarios relevant to the decision at hand. When decision-automation machinery grinds to a halt due to exceptional circumstances, next-generation decision-support tools are indispensable. If, at a moment's notice, subject-matter experts can use tools such as IBM Cognos to flexibly drill into and visualize the underlying data, they can identify the root issue more rapidly and formulate a recommendation for what to do next.
  • Crowdsourcing: Quite often, when we're stuck, the next-best-action intelligence we need to tap into is in the heads of our colleagues, or others in our organization who may have faced this situation in the past. Crowdsourcing is the approach in which we harvest the ad-hoc guidance of this valuable human resource. Social collaboration tools, such as IBM Connections, become indispensable resources for teams to scrounge up useful guidance in a pinch.

No matter where you source that operational guidance from, make sure you keep tabs on what action was finally taken. Did it address the issue effectively? Is it now the precedent that you should embed in the models and rules inside your automated next-best-action business process? Or did it create unforeseen problems?

It's not the next best action if it failed to make things better. You can't be a continually self-optimizing organization if you don't learn from both your successes and your mistakes.

For More Information:

Follow Jim On: