6 Best Practices for Real-Time Analytics

Real-time analytics enable faster, more precise and more effective decisions than conventional decisions made with stale data or no data. Real-time analytics require a structured decision process with predefined logic, and the data must be immediately available. Acquiring the data is often the limiting factor in the speed of making any kind of decision.

Speaking during the Gartner Business Intelligence & Analytics Summit this week in Grapevine, Texas, W. Roy Schulte, vice president and distinguished analyst at Gartner, said the concept of real-time analytics was historically applied primarily to operational decisions. However, new low-latency analytic tools and techniques also can improve the quality of tactical and strategic decisions.

“Real-time analytics can allow data science teams to perform modelling, simulations and optimizations based on a complete set of transaction data and not just samples,” said Mr. Schulte. “End users can harness increasingly sophisticated analytic capabilities through packaged real-time analytics embedded into data discovery tools and applications without prohibitive processing wait times or the need for developers to intervene.”

Mr. Schulte shared six best practices for making fast, real-time decisions without giving up the quality of the decisions.

1. Make Slow Operational Decisions Real-Time

Operational decisions are mostly, or entirely, structured, and are typically repeated many times. Operational decisions that go from slow to near-real-time may require new software tools, new kinds of data, new business process designs, and other changes to the business. The point of real-time analytics is to respond to conditions as they are at the moment, not to process yesterday's data or data from last month.

2. Track the Results of Real-Time Decisions and Modify Rules and Analytics Frequently

Most real-time operational decisions are repeatable. For example, a scoring model used to approve credit card transactions may be developed once on historical data, and then used for evaluating real-time credit card transactions for days or weeks. It’s important to track the results to make sure the models are working correctly and if necessary, modify rules and analytics frequently to get the right results for quick decision-making.

3. Use System “Guard Rails” and Human Oversight to Prevent Real-Time Mistakes

“Computers have no common sense, so they will make mistakes — sometimes dramatic and consequential mistakes,” said Mr. Schulte. “System logic should be used to check other systems, and people should monitor systems periodically.” A “stop” button should be incorporated, so people can do something quickly when a problem is detected. A system guard rails should be in place, sometimes in the form of “circuit breakers” that stop processing when a problem arises.

Download eBook: 5 Key Actions for IT Leaders for Effective Decision Making

4. Use Continuous Intelligence for Situation Awareness

Continuous intelligence (monitoring) systems run all day, listening to events as they occur, until they detect a threat or opportunity that requires a response by a person or system. The system proactively "pushes" an alert or other notification to a person via email, screen pop or other mechanism; or it triggers an automated response.

5. Provide Multiple Personalized Views, But a Common Operating Picture

Use this continuous intelligence to provide a common operating picture across the enterprise. Each person involved in a situation may have a personalized view specific to their role within the organization, but providing the real-time analytics across the organization ensures all involved have the same understanding of a situation.

6. Pursue Decision Management as a Discipline Comparable to Data Management and Business Process Management

“Decision management” is the discipline for designing and building systems that make decisions, where "decision" means determining a course of action. Decision-making systems are implemented by using rule engines, analytic software tools, 3GL programs or even human decision-makers - people are decision-making "systems" when they make decisions by following fully-defined, structured sets of rules.