When milliseconds matter: Injecting real-time analytics into operational systems
I have participated in quite a few client engagements—particularly with banks, healthcare providers and insurers—to discuss how injecting analytics into their real-time operational systems can help them both cut costs (through, for example, pre-payment detection and prevention of fraud) and drive top-line growth (by enhancing their offers to customers). With operations backed by transactions and data on z Systems, the technology is in place to do just this.
Injecting analytics into existing rules-based systems
Usually the first response I receive from clients when I suggest injecting analytics into real-time systems is “We already do that.” And so I peek under the covers, look at the architecture and find—in virtually all cases—a client that is wholly reliant on a rules-based system to drive decisions in its transactional systems.
That’s not to diminish the importance of rules: Rules are great, and I am very happy when I find a client using decision management software on z/OS. But rules are inadequate to drive truly intelligent decisions within the scope of transactional systems. Rather, incorporating predictive analytics into the transactions—creating an architecture that uses rules and models together, bound by a process orchestrator—can deliver an ideal business result within the context of stringent response times and CPU time SLAs.
My team has produced a reference architecture and assembled a set of “blinded” use cases based on ongoing engagements with clients who are taking the next step: incorporating predictive technology into existing rules-based systems. We have put out the word in several conferences this year, even though it’s a lot of ground to cover in only an hour. I’ll be delivering our presentation, “When milliseconds matter: Architecting real-time analytics into operational systems,” at the IBM Insight conference. To attend, head on over to Room 2280A, on Monday, October 26, at 10:30 AM.
Taking the broad view on real-time analytics
Insight this year will see a new type of session, the “super session,” which will be much larger than a breakout but will stop short of being a keynote. I have aligned the z Systems super session to our theme of real-time analytics.
But I’m taking a bit of a broader view, because I find “real-time” to be a very subjective description. For one thing, in many cases “real-time” is not used in the context I have just described—analytics within the context of an in-flight transaction. As I will describe in this session, that’s because delivering real-time analytics in operations is possible only by injecting the analytic engine directly into the transaction itself. But, unfortunately, many vendors aren’t able to do so.
Thus, I instead consider “real-time” analytics from the point of view of the business analyst or the data scientist who is kicking off an analysis. And even then, the time required to move data into position for analysis—which is usually considerable and comes at great expense—does not figure into the real-time equation.
z Systems maintains a broad, deep technology portfolio allowing z data to be analyzed in place, whether in the context of a transaction or of a human being kicking off a batch job. Our team believes that we can deliver superior results, and our super session will relate several never-before-heard client success stories, as well as featuring guest speakers and announcements and taking a deep look into our Apache Spark strategy. To attend “Distinguish your business: Compete by driving real-time high-quality insights,” be in Room 4048A on Tuesday, October 27, 10:30 AM.
I hope to see you at both of these sessions. If you can’t make one or both, catch me at our station in the Solution EXPO, where we’ll have pods focused on real-time, exploratory and IT analytics.