Follow Datanami:
June 28, 2012

Retailers Find Further Reason for Real-Time

Datanami Staff

Retail, both in physical stores and online outlets, has been one of the latest, greatest consumers of big data hardware and analytics tool. Many stores already have sophisticated systems in place to allow for near real-time responses to price fluctuation needs and in-store personalized marketing programs.

Retailers use streams of data from registers, social media, call centers, return and invoicing centers and various other places to piece together competitive elements of their business. But big data in retail doesn’t end there.

Another area that’s garnering more notice from large retail chains is in fraud detection and theft—both on the part of employees and thieves wandering the stores.

While loss prevention in stores might mean video monitoring and mirrors, many loss prevention specialists say these small-time crooks are no matter in the big picture; thieves are using technology themselves to run rings that span borders.

Other criminal activity, including wide-scale theft from organized retail crime units that tap stolen credit and gift cards, not to mention the retailers’ own employees that skim from the top (or worse) are among the most serious, pervasive ways retailers find themselves leaking profits.

Among the many tools available in retail’s loss prevention arsenal are exception-based reporting systems (ERBs) which deliver exception reports that examine everything from cash shortfalls and overages; suspicious returns, fake employee numbers, discounts and several other noteworthy aspects of point-of-sale (POS) and retail activity.

According to one loss prevention implementation company, LP Innovations, ERB systems have evolved dramatically and become more widely used due to lowered infrastructure costs. As the company’s manager of data analytics, Michael Elliott told us recently, developments that include the elimination of high storage costs have allowed retailers to look back in far greater detail to spot trends that indicate theft and fraudulent activity.

Elliott has been working on loss prevention analytics solutions since the 1990s and says he was one of the first “power users” of the ICL aspect application for data mining in retail before moving on to implement several different EBR and XBL systems at several retail organizations. Elliott claims that while there has been an evolution of the systems he’s worked with over the years, there is still a desire for more immediate detection and alerts of fraudulent activity while it is occurring, or at least in a short enough amount of time to make the move to stop it.

With these systems, the focus has traditionally been on POS data from keylogs that are loaded daily into the systems and analyzed at particular intervals based mostly on trends and patterns that analysts know to look for. As storage got cheaper and the platforms grew more advanced, the detailed data could be analyzed over a longer period, but these systems still lack the ability to turn around real-time data for retailers, which would allow them to spot fraud as it is happening and be alerted accordingly.

With EBR implementations, there is always a demand for triple-polling of data or, as Elliott explains, “bringing in the real-time, so if you’re say, The Limited, and there’s a customer that’s doing a large return, corporate is notified” immediately to nip the activity in the bud—or at least be left with more than just POS information to work from.

Elliott says that although we’re getting closer, there is a gap in terms of real-time detection of fraud and pinging of the both retailer’s security arm and the credit card company. Platforms are evolving to help meet these needs and integration technologies with a retailer’s other data (point of sale data, video data, etc.) are also developing.

According to the loss prevention systems expert, fraud continues to keep pace with technology innovations meant to detect and curb it. For instance, he says that gift cards alone have complicated retail fraud and theft detection systems and processes enormously. He pointed organized rings of criminals that find the numbers of gift cards that haven’t been activated yet. The criminals keep calling to check on the balances on these cards and as soon as they’re activated, they use these to buy merchandise, which then is taken to stores for cash redemption.

Battling the “gift card criminals” is not a simple process; it requires data from the balance-checking phone calls, returns at stores across the country, the use of cards that might be compromised or used suspiciously—all of this takes a unique, and preferably real-time approach to data management—one that will continue to evolve, lest gift cards cash out smaller retailers.

A number of companies with complex event processing and other high performance analytics offerings can handle these demands—but these systems for detecting fraud and theft come at a steep price and require complex integration processes with other sources of data.

While Elliott says they’re getting to that point where real-time is “real” most companies only review and act on flagged fraud anomalies up to 30 days after the fact.

What the retail industry needs are sophisticated fraud detection and alert systems that let retailers pinpoint fraud as it happens, deactivating gift or credit cards—or at least flagging transactions for same-day investigation.

Related Stories

How 8 Small Companies are Retooling Big Data

Partnership Targets BI Scalability

Six Super-Scale Hadoop Deployments