Skip
Current Issue
This Month's Print Issue

Follow Fast Company

We’ll come to you.

Fastcoworks Created for and commissioned by IBM
Fastcoworks Created for and commissioned by IBM
Instant Insights: The Game-Changing Power of Real-Time Analytics

In the age of big data, no company can thrive without finding a way to analyze vast and often hugely varied reams of information both quickly and intelligently. Across industries, in fact, making use of insights derived from that data, in real time—even as transactions occur—has emerged as a singularly dynamic business tool.

Today, any company running next-generation mainframes to power secure, real-time, in-transaction analytics—with the ability to cross-sell and up-sell products and services; defend against fraud; and run cognitive, predictive algorithms—has given itself a meaningful advantage over its competition. Which is why IBM, which produces systems that do precisely that, is at the very center of the new paradigm. "Most of the world's transactional systems already run on z Systems," notes Donna Dillenberger, an IBM Fellow, engineer, and mainframe expert. "From airline reservation systems to major credit card and banking transactions to supply-chain logistics—IBM mainframes handle the majority of that data, globally."

The analytics associated with, say, a credit card swipe within a z Systems mainframe environment have long been mind-bogglingly fast, Dillenberger says, for two reasons. First, the action—the swipe itself—and the analytics occur within the same box, or platform. In the not-so-distant past, most companies took data from transactional systems and copied or offloaded that data to another box to be crunched. This copying and off-premises analysis of data could take anywhere from hours to, in some cases, weeks.

"In short, organizations were always doing analytics on stale data," says Dillenberger. "Now, with the analytics running in the instant the transaction itself happens, the system doesn't have to make any network calls or activate extraneous operations. At the time of a credit card swipe you can see whether a transaction is fraudulent (and block it), or even predict with cognitive algorithms the next best action or recommendation for that specific customer, right then and there."

This sort of instantaneous, personalized, in-transaction data crunching is a critical advancement over traditional affinity groupings ("You watched Casablanca on Netflix last week; you might also like these movies."). As Mike Gualtieri, principal analyst at Forrester Research, puts it, "The latest z Systems' ability to run predictive and in-transaction analytics without affecting transaction time is extraordinary. We're talking about sophisticated analytics taking place in single-digit milliseconds."

The second reason for z Systems' unprecedented speed is not simply that the data crunching occurs "on-site," but the analytics themselves run two to three times faster on the z platform than on other platforms. "We have the fastest microprocessor and the most scalable I/O [input/output] subsystem in the industry," Dillenberger says. With the new z13s mainframe's increased memory, larger cache, stronger cryptography, and, perhaps most critically, IBM's partnership with the powerhouse predictive-analytics software company Zementis, the landscape—and the cloudscape—has suddenly become even more intriguing.

What does this mean for businesses? Barclays Bank, for instance, which processes millions of customer transactions every day from ATMs and mobile devices, partnered with IBM on a cloud-based analytics strategy—and was able to analyze data a full 60% faster than it had previously.

IBM's Mythili Venkatakrishnan, the z Systems architecture and technology lead for analytics, explains that in-transaction analytics bring benefits to consumers as well. "In the banking industry," she says, "we've had the ability to detect for credit card fraud, but until now, instantaneous, in-transaction, predictive analytics have not been among the capabilities." Consider the scenario in which a customer lives in New York but regularly travels to Florida. On each trip south, her first credit card purchase goes through, but the second is held up. Why doesn't the card issuer know that she regularly travels to Florida?

If the banks analyzed the history of a customer's travel patterns, notes Venkatakrishnan, they could easily tell whether a particular transaction was in line with an ongoing pattern. The problem is that fraud prevention has long relied on binary rules-based detection: The answer to any query is either yes or no. Should they deny the charge outright? Should they hold it up and text you? Should they approve it? Life is an ever-changing series of patterns, and analytics that rely on binary decisions are hard-pressed to deliver the best outcomes.

"With new in-transaction predictive analytics—like those we've developed with Zementis—we can analyze information that is far more complex and nuanced," Venkatakrishnan says. "For example, we can tell whether the merchant that you're trying to buy from while on vacation is similar to merchants you've regularly done business with in the past. We’re innovating in this space faster than we ever have."


To watch a webcast about driving faster and more accurate predictive insights for your business, click here.


Other Related Topics:

This article was created and commissioned by IBM, and the views expressed are their own.

loading