Businesses know good customer service means more than a friendly operator on the other end of a phone call. Today’s retail environment includes the freedom of shopping, buying and firing off complaints online, and that means businesses need to find new ways to connect with buyers.
“The way we interact with customers has changed in the past five years,” said Scott Groenendal, program director of the predictive analytics market management team at IBM. “The customer experience framework has changed. The touch points have increased. And we should be able to do a better job of cross-selling.”
Last week, Groenendal was a presenter at the Henry Stewart Marketing Analytics 2011 conference, a two-day summit held in Boston. Speakers talked about ways to improve marketing analytics programs in pursuit of a
Conference attendees offered several tips on how to build and maintain a successful marketing analytics program, including the following:
Keep your eye on the customer metrics ball. Rex Davis, vice president of custom insights for Cincinnati-based analytics firm DunnhumbyUSA, said businesses should strive to make personal connections with customers. One way of doing that is through a marketing analytics program that aligns important business metrics goals.
“The key thing we like to look at is keeping the customer central in terms of how the retail organization is looking at their business,” said Davis, whose company works with the likes of Kroger, Tesco and Panera Bread.
When Dunnhumby is brought in by a client, it makes sure senior executives have access to customer-related metrics. But, Davis stresses, metrics alone are not enough; empirical evidence showing how sales are affected is vital to a marketing analytics program’s success.
“It has to tie back to something they see,” he said. “Otherwise you’re just having a conversation with yourselves.”
Davis said many of the top metrics will be centered on loyal customers, but other metrics will take a front seat when problems arise. Those metrics should be determined on a case-by-case basis.
Data is good; analyzing data is better. The ways to interact with retail organizations have skyrocketed, but additional channels or touch points also provide businesses an opportunity to more intimately understand who their customers are.
“In the U.S., 65% of the top 50 retailers have a loyalty program,” said Amit Maheshwari, practice director of the industry advisory group for Wipro IT Business in the Greater Seattle area. “They can track what each customer is doing and give a personalized offer.”
A loyalty program can provide access to customer data, which should be analyzed to find out what kinds of products customers bought, Maheshwari said.
He shared a case study of a large grocery retailer using an email marketing campaign, which had a 0.2% redemption rate. Applying analytics against a segmented list of customers ranging from frequent shoppers to bargain hunters and their purchases showed that the most extreme customer segments had the highest response rates. Armed with the information, the retailer retooled its campaign to target specific types of people and the attributes those segments were looking for, which increased its redemption rate to 1.6%.
Unstructured seriousness. Businesses looking for a comprehensive understanding of their customers will need to become adept at analyzing different kinds of data.
“Customers may look the same with traditional data,” said David Hastings, director of Dayton, Ohio-based Teradata Corp.’s advanced analytics center. “But if you look further to Web data, for example, differentiation can occur.”
Web data can include how frequently customers browsed a business’ site, what items they viewed and how often, and if they placed a product aside for purchase consideration only to abandon it later.
The kind of data Hastings is referring to can enter the realm of “big data,” or explosive volumes of data in forms that can be hard to store and analyze with traditional tools. Managing big data requires new tools, such as MapReduce. Originally developed by Google, MapReduce is a programming framework that breaks jobs into pieces, performs the work for those pieces in parallel and brings them back together, Hastings said.
Text mining, for example, is a more complex approach that may benefit from MapReduce and help with customer relations. Hastings described the practice as a three-level process with the ultimate goal of not only uncovering sentiment, but intent as well. Because raw text data can be so massive, tools like MapReduce can help break the material into pieces for a more meaningful analysis.
Marry predictive analytics to business rules. Businesses delving into more advanced analytics are branching out from a looking-back to a looking-forward approach.
“Predictive analytics helps with data understanding,” said IBM’s Groenendal. “Data mining should pull disparate data sources together for one version of the truth. That should be your data mining workbench. It should be data agnostic [and] understand unstructured and structured data.”
Predictive analytics can help with strategic and tactical decision making, such as determining future staffing requirements, but it should also inform operational decision making, Groenendal said. This can be accomplished by embedding predictive analytics into business rules.
“Without business rules, you could make the wrong decision,” he said. “Just relying on gut, you could miss out on the nuggets data mining can help uncover.”
The two provide complimentary forms of knowledge and can produce results in hours, minutes, sometimes seconds, Groenendal said. That translates into a better workflow and better customer service.