Finding New Meaning in DATA
If companies had a crystal ball to help them predict which consumer packaged goods would sell best where, their worries would be over. Retailers could plan their inventories precisely and marketers could zero in on the best prospects for new products.
No one has invented that crystal ball yet, but technology is getting closer to helping marketers predict the answers to important questions by analyzing the volume of purchase data now available. With billions of dollars worth of food products moving from food manufacturers' factories to retailers each year, any improvement in companies' ability to predict demand could save millions of dollars by allowing for better inventory allocation. It's no surprise, then, that giants like Battle Creek, Mich.-based Kellogg Co., which sells $13 billion worth of food to retailers each year, are implementing so-called predictive analytics to forecast which stores need which products when.
"The use of these kinds of analytics is picking up, definitely over the last couple of years among manufacturers who want to improve on-shelf availability," says Robert Byrne, chief executive officer at Norwalk, Conn.-based Terra Technology, which provides a predictive analytics solution to Kellogg. Data show that improving product availability by 3 percentage points correlates with a 1 point gain in sales, Byrne says.
Terra Technology |
Predicting Future Activity
Predictive analytics, which involves using data to predict future activity, can give companies a competitive edge. "The beautiful thing is we are at a point in time where some of the companies that are just starting in [predictive analytics] can leapfrog what came before," says Marek Polonski, vice president of Applied Predictive Technologies Inc. in Arlington, Va. "They can be on the cutting edge in a matter of weeks."
When precise data drive decisions, better decision-making results, proponents say. Applications vary, from Kellogg's use of predictive analytics to improve its products' on-shelf availability while maximizing its supply chain efficiency, to other marketers predicting how a promotion will perform or determining the best in-store locations for a secondary display. Retailers can use predictive analytics to help allocate shelf space or determine how a new product introduction will affect overall category sales.
"We are starting to see this capability...expand across the functions," says Ramesh Murthy, head of the North American retail and CPG practice of Tata Consultancy Services in Mumbai, India. "For example, after success in the supply chain, it is being applied in pricing, etc." As the technology develops, more retailers will likely adopt it, Murthy says, especially as they come to understand the value of the new data sources.
Converting Data to Information
The amount of data available from retailer checkout terminals is voluminous, and when analyzed proficiently, it can provide companies with meaningful information to help manage operations. Retailers can learn basic purchase trends and which products are commonly bought together. When this data is combined with demographics from shopper loyalty programs and U.S. Census data, it becomes even more valuable.
Catalina Marketing |
"Historical purchase data is one of the best predictors of future purchase intent," says John Caron, vice president of marketing at Catalina Marketing in St. Petersburg, Fla. Marketers can glean highly specific information from purchase data, such as brand preferences, purchase frequency, payment type and more.
The popularity of loyalty programs at retailers is driving data accumulation. A customer who swipes her frequent-shopper card before checking out tells the store precisely what is in her basket that day. Many stores also track customers by credit card number, Polonski says. The resulting data might not be quite as valuable as that coming from loyalty card users, however, since customers use different credit cards or cash on different shopping trips, while loyalty data is collected more consistently.
Big picture data, which point out marketwide sales trends, is available from market research firms, such as SymphonyIRI Group and Nielsen Co., while demographics also can be relevant. For example, a retailer using predictive analytics to determine product mix may want to incorporate U.S. Census data showing average family size in the vicinity of the store.
Social media also can be a useful source of information about shoppers' preferences. Retailers are going beyond counting "likes" and mentions on Facebook or Twitter to correlating social media activity with specific customers. "They are starting to think about how to segment their customers based on their social media engagement – are they an influencer, are they an early adopter, etc. – to then integrate with predictive analytics to define appropriate promotional and communication models," Murthy says.
Providing Answers to Questions
So how can retailers and CPG companies best use all this data? Retailers can analyze it to examine just about any business idea, Polonsky says.
"For example, let's say I'm introducing a new product and I want to understand what impact that is having on my customers and my bottom line; predictive analytics can be used for that," Polonski says. Predictive analytics also can help executives make capital investment decisions by providing information to predict the return-on-investment of a store remodel or determine which stores should be remodeled. Human resources professionals are turning to predictive analytics to test new training programs by trying several different variations.
More commonly, retailers are using the applications to make product-related decisions about promotions, pricing and location in the stores. Retailers might use predictive analytics to determine a new cereal's effect on sales in one store before introducing the product to the entire chain, Polonski says. The analytics would allow the retailer to examine performance of the cereal in isolation as well as gauge its impact on the category. For example, did consumers switch from another brand of cereal to the new brand? Or did they buy more cereal altogether?
But the analytics can go a step further by providing information on the role the new cereal plays in the customer's overall basket. "Am I getting folks to buy more milk? In other words, is there a halo effect on the rest of the store or are folks just coming in and getting that cereal," Polonski asks. If customers with the new cereal in their basket have larger overall orders than they did previously, the new cereal is likely affecting purchase patterns.
The new cereal also could be driving loyalty, particularly if it's not available elsewhere. Are customers more pleased with the store's cereal selection, and thus becoming more loyal customers? "That's the framework," Polonski says. "Since this cereal introduction is a trial, I can see how it will do when I roll it out everywhere."
Naturally, the analytics can be used for existing products as well, says Michael Havens, director of retail marketing for 1010data, a provider of cloud-based analytics for big data. The data might show a retailer that buyers of Coke also buy M&Ms. "If you know that, you might not want to promote Coke and M&Ms at the same time, because customers are already buying them together," Havens suggests. "And sometimes you'll find an affinity one way but not the other. For example, you might learn that if they buy M&Ms, they buy Coke; but if they buy Coke, they don't necessarily buy M&Ms." That type of information can improve promotions' effectiveness, Havens says.
Collaborating for Mutual Benefit
When manufacturers and retailers work together, the effectiveness of predictive analytics on sales and margins often multiplies. "Where the real power comes in is when [retailers and manufacturers] are partnering on a new product introduction," Polonski says. "The beautiful thing is when they are collaborating on a product introduction, they use a common language and common platform, which means they have a common way to interpret the results. And there is really no ambiguity about who is right. There is just one version of the truth."
Applied Predictive Technologies |
But collaboration requires both parties to share data, which doesn't always happen. While a CPG manufacturer might interact with 10 to 20 major retail grocery chains, those retailers are likely to have hundreds or even thousands of suppliers. Thus, many retailers are less excited about sharing data than manufacturers are.
But CPGs can make the case for collaboration by demonstrating its value to retailers, Byrne says. Information-overload is a problem for many retailers who don't have time to decipher all the data their checkouts generate, so CPGs can provide them a service by turning the data into meaningful information. "The retailers are saying, 'If I make the data available to you, you need to understand it, and respond and give me advice,'" Byrne says.
Walmart has effectively collaborated with manufacturers this way, Byrne says. "Walmart has done a great job of outsourcing data analysis by having the manufacturers do it for free," he says. Ultimately, by sharing the data and the information gleaned from it, both parties stand to benefit from higher sales and earnings.