CSP Magazine

Big Data Mining

It’s not scary: Simply developing actionable reports means more profit for convenience retailers

Back before a certain brand of protein milkshake proved its power in c-store cooler sets nationwide, Donna Perkins couldn’t figure out why anyone would want to drink it.

Still, her chain’s merchandising team gave it a shot anyway.

To prove her taste buds right, Perkins, pricebook manager for Maryville, Tenn.-based Calloway Oil and its 23 E-Z Stop Food Marts, ran a report, expecting results to validate her palate.

It turns out, as many retailers know, that the shake, Muscle Milk, really shook things up in the cooler, and Perkins took her own lesson to heart: Data doesn’t lie.

“You have to be willing to listen to the data, to be in school,” she says. “Maybe the data will tell you something new.”

In an era of “big data” and the idea of churning billions of bytes to spot the most subtle buying trends and tics in consumer behavior, many retailers may actually be missing the bigger picture.

“By definition, there are very few retailers who come close in our space to big data,” says Drew Mize, COO of The Pinnacle Corp., Arlington, Texas. “But that doesn’t mean there’s not a huge opportunity to change their businesses with data that’s available.”

Pulling transaction data from POS registers and developing meaningful reports is standard for many store-level technology solutions, Mize says. However, obstacles still exist:

▶ Creating meaningful reports: While a software solution can come with dozens of ready-to-run analytics, charts and graphs, operators often want to customize reports based on the characteristics most meaningful to them.

▶ Training: Showing personnel how to create their own reports and act on those findings can be difficult. From knowing what the numbers mean to running reports that can inspire action, key individuals must actively participate in the process, say retailers such as Perkins.

▶ Integration: Adopting new soft- ware solutions may mean someone on staff must integrate a new stream of data into the company’s existing system, a process that could take a skilled IT person days or sometimes weeks to do.

▶ Communication: Often, essential information arrives weeks after the fact. Generating real-time reports enables retailers to be more proactive vs. reactive.

Having attended many technology events, Perkins believes retailers face many of the same challenges. The frustrating part for her is knowing that benefits exist at the other end, especially in areas of loss prevention, promotions and merchandising.

“A big issue is getting people trained to use the data that’s there,” she says. “Making useful tools for operations people and ... quickly knowing what happened yesterday, not three weeks ago, is very important. Especially being [a smaller chain], we have to be nimble and quick to respond.”

CONTINUED: Defining Big Data

Defining Big Data

Understanding what big data means may help retailers see what’s most relevant to their businesses. Stamford, Conn.-based Gartner Inc., a technology research and advisory group, describes big data as high-speed computers combining large amounts of information from any number of sources. The three V’s that typically describe big data are volume, velocity and variety.

Some define big data more specifically as working with historic data sets, while others see it as analyzing a stream of new information. Either way, the idea is to quickly process vast amounts of data using algorithms and correlation analytics to spot unique trends and gain strategic insights.

But the story starts with data—the sheer amount that exists today and is growing exponentially. It’s a relatively recent phenomenon, starting about a decade ago, when the pace of information going from analog to digital picked up. The evolution led to expansive data ware - houses and so-called server farms, with companies such as Amazon, Oracle and Microsoft offering low-cost solutions.

Earlier this year, Minneapolis-based General Mills shared statistics it gleaned from analyzing social media posts from c-store customers, tracking keywords, developing buckets and assessing behavior. Officials said people put up a lot of “craving posts” for slushy drinks or baked goods and that conversations spiked whenever c-stores introduced a special deal.

“For the most part, [c-store] implementations are modest compared to what the world is talking about,” says Greg Gilkerson, president of PDI, Temple, Texas. “But while we may not generate big data, what becomes relevant is possibly accessing big databases to enhance what we have.”

Gilkerson suggests merging a feed from the National Weather Service to a database of business transactions from c-stores. “There’s so much more data available today than previously and it’s only possible [because] you have relatively inexpensive storage and systems implemented to the level where you can now collect information,” he says.

But even a little data can be overwhelming. “While I don’t think data for a significant majority of our retailers is ‘big data,’ ” says Mize of Pinnacle, “it’s still an overload of data.”

Crawling Out of Chaos

No one can deny that the industry has made progress. “Fifteen years ago, the problem was, ‘I need data, but I can’t get the data [from my POS].’ That’s no longer a problem,” Mize says. “One of the biggest struggles retailers are faced with is getting through the data to understand what they’re actually after.”

Often, the problem comes before the question. A loss-prevention manager will start noticing something’s not right in the numbers he or she is seeing. The hunt for what’s wrong will lead to an investigation and eventually an automated report that tracks the specific problem.

An example, Mize says, would be shortages in fuel inventory. Pinpointing the cause is a common exercise. Some - times it’s as simple as drive-offs, but are these “true” drive-offs or an employee filling up a friend’s car without paying? Sometimes the problem is an incomplete fuel delivery. The answers lie in piecing together data from different devices and departments and hoping that existing systems can flag the problem.

Of course, the trick is developing that automated report. This is a focus for many retailers, both large and small, says Rich Hathaway, director of business intelligence for ADD Systems, Flanders, N.J. With the company’s solution, retailers can develop exception reports that spot costs or sales figures that fall beyond set parameters; the proper person receives an email when numbers exceed those limits.

For Hathaway, the term “business intelligence” is not a highly technical solution, but it does involve the ability to draw data from multiple sources. “It’s an evolution,” he says.

Creating reports can be a user-friendly process, says Eros Fleming, global account executive for DataMax Group Inc., Round Rock, Texas. Solutions today allow retailers to gather different performance metrics and market-basket analysis models into a single location, often a personalized dashboard screen on someone’s computer.

Overall, retailers have been pushing the creation of such reports out of IT and into marketing and operations, says Mize of Pinnacle: “Retailers are getting more versed, but it’s still a struggle and there’s still a fair amount of relying on IT to build reports.”

CONTINUED: Analyzing the Data

Data Analysis

C-store retailers aren’t alone in their struggle to turn numbers into meaningful information. Newton, Mass.-based TheServerSide.com, an online forum for program architects and developers, says the big-data market—the business of storing and warehousing information—is maturing, while the analytics piece is trailing behind.

TheServerSide group described a large “skills gap” in this area, with ongoing debate on the type of mathematician needed to master this kind of creative analysis. The problem is the variety of data sets involved. Some are defined or “structured,” such as POS transaction logs; others are “unstructured,” such as social media.

For the c-store channel, one solution may be bypassing the heavy number crunching altogether.

What retailers really want from their data is something to act on, says Jim Manzi, chairman of Applied Predictive Technologies (APT), Arlington, Va. “Did my action—for example, stores staying open for two more hours—improve key performance metrics?” he says.

Finding correlations among data sets can spark interesting questions, but the real work comes with conducting physical in-field tests, Manzi says.

For example, picture a retailer implementing a 10% discount on beverages across stores in June. Sales go up. Using only correlational analysis, the retailer might think the discount caused that increase. “However, if that retailer instead conducted a test of the discount program by implementing it in a small subset of stores and compared the results to the rest of the chain,” he says, “he or she may instead discover that sales were up everywhere, with or without the promotion.”

So the cause of the sales increase may have been consumers traveling during the summer, which led to more frequent fuel purchases and ultimately more store purchases. “In this example, the promotion might simply be giving away margin to consumers who would have purchased anyway, a profit-negative program that could be avoided by testing it prior to any rollout,” he says.

Retailers considering any big- data analysis will most likely choose a cloud-based provider, Manzi says. Such solutions allow for the integration of a variety of data streams, including sales, transaction-level data, customer loyalty, weather, traffic and demographics.

“Automatically updated cloud-based solutions [let] executives ... adjust and optimize each business decision based on the most recent data available,” Manzi says, “without ongoing IT involvement.”

The Integration Question

The use of cloud-based solutions alludes to the larger dilemma of expanding data sets. With more business-focused apps coming out, new information might bring insight, but it can also add new concerns.

For instance, Brentwood, Tenn.-based Mapco recently rolled out a new mobile app that allows its field managers to validate that store personnel did certain tasks. Traveling managers log in on their cells, do online checklists and see uploaded photos that prove work was done.

“The accountability within our organization has increased drastically,” says Brian Veasman, director of operations for the 472-store chain, about the solution, supplied by San Francisco-based Zenput. “And we’ve become proactive instead of reactive in our retail execution.”

In lieu of such a solution, Vladik Rikhter, CEO of Zenput, says companies such as Mapco often use a jumble of emails, texts and photos to verify work was done. He and his co-founding partner created the app in 2012 to address the issue. Its appeal has risen with the increasing ubiquity of people being comfortable with uploading photos to social media, along with falling data-storage costs.

The rise of business apps begs the question of integration. While a chain can take the time to integrate the new data flow into its host solution, time, energy and even issues of data storage can put the entire exercise to question.

“We’re willing to play nice when it comes to integration, but the product is so straightforward,” Rikhter says. “You can get a new checklist up and running in a couple of minutes. Then you could spend a couple of weeks to get it integrated, but where in Oracle or SAP is it going to sit? And where are the photos going to go? We’re more than happy to see it happen, but why?”

“In some cases, integration is worth - while, but some make sense as a stand- alone,” says Gilkerson of PDI. “[For PDI’s solutions,] we have a standardized way to accept data from third parties, but retailers may be as happy with a certain solution being a stand-alone app.”

CONTINUED: A Retailer's View

A Retailer’s View

For Calloway Oil, data can lead to important insights. Perkins, who uses a system from PDI, cites greater improvement in how the chain assesses the success of a promotion via analytics. Understanding redemption rates is critical in fashioning a successful promotion, she says. If a store runs a special that gives cents off a gallon of gas, then it’s good to understand that redemption rates jump depending on if it’s 3 cents a gallon or 10 cents.

While that’s a more obvious example, she says redemption rates can go from an acceptable 70% to a sky-high 95%, calling to question the promotion altogether, because many retailers count on a lower redemption rate to actually make money on a promotion.

“Sometimes you might not be making a profit on a given promotion,” she says. “But was it successful? What happened to other sales? You’ve got to look at the big picture.”

Top Reasons to Use ‘Big Data’

For most c-store retailers, data analysis comes nowhere near the larger concept of so-called “big data.” But that doesn’t mean retailers can’t benefit greatly by mining two or three years of transactional data for any number of purposes. Here’s a quick list of important reasons to consider business analytics:

Inventory management: Knowing what’s selling helps the ordering process and cuts down on out-of-stocks.

Loyalty: Retailers managing rewards programs can better understand their favored customers and track the strategy’s overall effectiveness.

Merchandising and product assortment: Retailers can better understand what sells, giving category managers leverage to cut back on some products and stock more of others. It also helps in understanding market-basket trends, sometimes making the case for slower-moving items favored by loyal customers.

Promotion analysis: Such information can help identify promotion effectiveness, as well as help retailers design better strategies.

Loss prevention: Often doing predictive analysis can help spot inconsistencies that may be signs of internal theft.

Want breaking news at your fingertips?

Get today’s need-to-know convenience industry intelligence. Sign up to receive texts from CSP on news and insights that matter to your brand.

Related Content


More from our partners