6. The Path to “Next Best Offers” for Retail Customers
Thomas H. Davenport, John Lucker, and Leandro DalleMule
Retailers hunger for new, effective ways to drive sales, traffic, and growth for their stores, sites, catalogs, and other channels. In the past, they relied on local salespeople to match the right products to the right customers and to suggest the perfect offer to motivate a sale. Today, multiple customer channels, shorthanded staff, and busy consumers are driving innovative mechanisms for next best offers, using data analysis and technology to enable scalable precision. Next best offers also have relevance for any other industry with consumers as customers, including consumer financial services, travel and transportation, and telecommunications.
The “next best offer” (NBO) is a targeted offer or proposed action for customers based on the following:
• Analyses of their past shopping history and behavior
• Other customer preferences, attributes, and life stages
• Purchasing context
• Attributes of the products or services from which they can choose
NBOs should result in a high likelihood of purchase, but the best programs go beyond the sales transaction. They reward the customer for past loyalty, deepen an existing customer relationship, and appeal in a highly relevant way.
NBOs tend to apply particularly to companies providing services directly to consumers; business-to-business firms may not have enough data to draw on. Offers can consist of products and service discounts (diaper or spa treatment coupons), information (Google ads to click), or even relationships (LinkedIn and Facebook recommendations). They may be delivered through in-store salespeople, call centers, direct mail, kiosks, register receipts, and mobile devices.
Clearly, well-designed NBOs are the future of retailing; presently, however, NBOs are either poorly executed or not done at all. Most offers are indiscriminate, ill-targeted, and too numerous—the new junk mail. One major retail bank concluded that its offers were more likely to create ill will than increases in sales.
Analytics and the Path to Effective Next Best Offers
The world of customer analytics is a complex and fast-changing one with incredible potential. This is the process by which data from customer behavior is applied to key business decisions via market segmentation and predictive analytics. NBO programs are a worthy target for any company wanting to develop or improve its use of data and analytics to serve customers, because they require knowledge of customers, products, offers, and the rules and algorithms for combining them.
No organization today has “mastered” NBOs, but some have made dramatic progress toward creating offers that do the following:
• Meet the company’s objectives
• Are targeted to a customer segment of one
• Arrive via the customer’s preferred channel
• Are delivered when the customer is in the mood and location to buy
• Have a high conversion rate that can be achieved and measured
• Take into account the customer’s life stage, previous buying behavior, current location, and all responses to previous offers
• Incorporate the discussions and behaviors of friends in social media
Short of the perfect offer, there is still substantial opportunity to create and improve NBOs. In our research, we’ve created a framework for effective NBO initiatives, as shown in Figure 6.1. Some companies may not be ready to undertake all these steps at once, but eventual progress in each phase will be necessary to improve offers.
Figure 6.1. The path to next best offers.
Offer Strategy Design
As with any strategy, an organization should begin by reflecting on what it wants to accomplish with its offers and how those goals can best be achieved. Offer strategy design should include topics such as these:
• How you want offers to affect your customer relationship
• What channels you plan to use and under what circumstances
• What data to gather and analyze
• What you plan to offer
• How an offer may impact the market and competition
• Collaboration with manufacturers that supply products and finance offers
The U.K.-based retailer Tesco has been very successful with its targeted coupon offers in its loyalty program, Clubcard. But critical to the design of the offers program was Tesco’s unrelenting desire to both know more about its customers’ preferences than anyone else and to reward customer loyalty and other desired behaviors with coupons that they will welcome and try. The offers generated by Tesco and its in-house consultant, dunnhumby, achieve redemption rates averaging between 8% and 14%. This is far higher than the rest of the grocery industry, which averages between 1% and 2%. dunnhumby’s research suggests that offers targeting loyal customers lead to higher revenue lift as well.
Microsoft’s recent consumer offer campaign for Bing, its new search engine, focused on getting users to try Bing or use it more frequently. The company’s marketers wanted to make the offers via email, but because the company sells email applications and is an avowed foe of spam, the offers needed to be perceived as highly relevant rather than invasive. So Microsoft employed a new technology, Infor’s Interaction Advisor, for real-time targeted emails in the very successful campaign.
The objective for offers may well change over time. For example, the DVD rental firm Redbox initially made email- and kiosk-based offers with the objective of having customers try its rentals. In the process, customers got used to renting through the Redbox channel in a familiar, convenient location that the customer had to visit anyway (often grocery stores). The Redbox technical process required a learning curve for checkout and check-in. As the business grew, executives realized that more revenue and profit growth would result from offers encouraging customers to rent more than one DVD per visit.
It’s just as important to declare what not to pursue in offers. One retail chain concluded that social media was not an important factor in determining offer content. The company’s marketing analysts monitored social media content about the company and observed that the products it sells are not a major focus of discussion. For Ticket master, however, there is little doubt that social media play an important role in young customers’ decisions about what concerts and events to attend. Therefore, analyzing social media and using it as a delivery channel will be increasingly important to its offers.
Know Your Customer
Targeted offers are based on the detailed analysis of information about the customer, product offering, and purchase context. Customer information can include basic attributes such as demographics, residence, previous purchases, income, and assets. From these raw information sources, a vast trove of synthetic data can be created by combining like and disparate fields in meaningful ways—ratios, statistically derived fields, and derivations of averages and probabilities. Some of these fields are readily available, but others can be difficult to obtain and integrate with other customer data. In addition, new possibilities for customer information are opening up through “SoMoLo” (social, mobile, location) data:
• Where is the customer at this moment—anywhere near one of my stores?
• What is the customer saying about my company or brands in social media, and how influential is he?
• What are my customers’ friends buying and discussing online?
Walmart acquired the start-up Kosmix to begin employing SoMoLo data in its offers. The apparel retailer H&M created a partnership with the online gaming firm MyTown to gather and use information on customer location. If a potential customer is playing the game on a mobile device near an H&M store, H&M makes offers of items to be used in the game. Customers are encouraged to go into the store and scan the item for a discount. Early results suggested that, out of 700,000 online check-ins by customers, 300,000 went into the store and scanned an item.
Know Your Offers
Many companies overlook the fact that they also need accurate product information and attributes to succeed with NBOs. There must be a sound basis for matching a customer and a product based on customer-specific, appealing product attributes.
For some products, product attributes can easily be obtained from third-party databases. For example, firms making movie offers (including Netflix, AT&T, and Comcast) can surmise that if you liked one movie with a particular actor or plot type, you will probably like another. But for other retail industries, such as apparel and grocery retailing, compiling product attributes is much more difficult. Manufacturers don’t have official classifications of whether a sweater is “fashion-forward” or “traditional.” Grocery retailers can’t easily determine what food products appeal to customers with adventurous, healthy, or penny-pinching tastes.
It’s also important to know what products manufacturers want to promote and what their objectives are for the customer’s product use. Do they want customers to try it, acquire more of it, or perhaps buy it in combination with another product?
Tesco has aggressively pursued the classification of product attributes to ensure that customers receive offers related to their tastes. Attributes, such as whether a product is frozen or not, or the cost per kilogram, are sourced from its product databases. But for those involving taste and lifestyle, which are more difficult to classify, Tesco employs a “rolling snowball” approach to identifying taste-related product attributes. For example, to identify products that appeal to adventurous palates, it takes a product that is widely agreed to be adventurous in a country context. In the U.K., Tesco chose Thai green curry paste and identified other adventurous products by analyzing relatedness coefficients. If customers who bought curry paste also bought squid and wild rocket (arugula) pesto, these products probably appeal to adventurous customers.
Know the Purchase Context
Offers should also be based on a variety of purchase context factors, such as the inbound channel for customer contact. Did it occur by walk-in, telephone, email, web browser, receiving mass media messages? Online offers can be based on a variety of immediately preceding behaviors, including the previous site visited and click-streams on the company’s own site. The customer’s reason for contact is another important variable. Is he or she shopping for someone else, seeking service, carrying out another transaction, seeking offers, or simply minding his or her own business?
Other contextual factors might include the time of day, current weather, and whether the customer is alone or accompanied. One Chinese shoe retailer has developed offers that target companion shoppers. When a woman walks in the store with her husband, this retailer offers him a relatively inexpensive item. The decision of which item to offer the husband is heavily based on his higher price sensitivity as a companion, versus his lower sensitivity when shopping for himself.
Some of the most valuable purchase context information today comes in the form of SoMoLo data. With proper usage, retailers can develop a ubiquitous capability to offer products and enhance the customer experience. Social and mobile data channels the voice of the consumer and many aspects of his or her preferences and behaviors, telling retailers what offers are more likely to succeed and when.
An interesting application of social data to develop highly customized offers comes from Sony, which has been experimenting with Facebook Connect, a tool that allows Facebook members to take their social networks with them around the Internet. Sony plans to use Connect to enable its developers to create personalized video game offers on the PlayStation 3 console. Game developers can pull information out of Facebook and push information to it. The next generation of video game offers could have pictures of your friends or your tastes and interests built right in.
If appropriately analyzed, mobile data also can help you better understand customer preferences, needs, and desires, and significantly enhance retailers’ ability to design their NBOs. Many retailers are focused on immediate location, which is valuable in targeting customers who have a strong propensity to buy. But location history can reveal a lot about customers as well. A company called Sense Networks has developed an application to help infer a person’s lifestyle based on his or her location history. Sense Networks claims it can estimate customer attributes such as age, probability of being a business traveler, wealth, and next likely location. By comparing where targeted customers go against data points on the movements of other customers, the company can create granular segments and allow retailers to offer targeted, timely NBOs.
Analytics and Execution: Deciding on and Making the Offer
NBOs are created by a predictive model or test, based on a series of variables or attributes. The goal is to identify the attributes most related to specific, desired customer propensities, actions, and outcomes. Simple predictive NBOs, such as those offered initially by Amazon.com, with “people who bought this may also buy that” cross-purchase correlations, don’t employ substantial knowledge of the customer or product attributes. In addition, Amazon makes email-based NBOs based on past purchase behavior. Unfortunately, if a customer buys something for a friend, he might be stuck with irrelevant offers for years.
Personalized offers normally are based on a combination of algorithms predicting a customer’s probabilistic propensity to purchase, customer lifetime value, cross-sell and up-sell probabilities, and business rules governing what offers are made under what circumstances. For example, a business rule might determine what offer is made when several products have equal propensity scores or might limit the overall contact frequency for a customer.
A key aspect of offer execution is to decide how and by whom the offer is to be delivered. The outbound mode of delivery of the offer is usually the same as the inbound channel, but not always. It can include
• Face-to-face outreach by a human
• In-store kiosk
• Mobile device
• Online: email or banner ad
• Register receipt
• Mass media
Many companies are attempting to address offers through multiple channels. “Our customers never met a channel they didn’t like,” said a retail banking executive. At CVS, the company’s ExtraCare loyalty program offers are delivered through register receipts, in-store kiosks, email, and even targeted circulars, and the company is experimenting with mobile coupons. Qdoba Mexican Grill, a quick-serve franchise, is using mobile coupons to expand its card-based loyalty program. It can deliver offers at certain times to increase traffic, while smoothing demand during peak times. Late-night campaigns near universities have seen a 40% redemption rate, while the average redemption rate is 16% for the whole program.
Starbucks uses over 11 online channels to develop targeted offers, gauge customer satisfaction and reaction, develop products, and enhance brand advocacy. Today more than 30 million Facebook users “like” Starbucks, more than 2 million follow the retailer on Twitter, and more than 300,000 images with Starbucks tags were uploaded to Flickr. Facebook fans spend on average $235 per year at Starbucks—more than twice the amount that nonfans spend. These fans comprise a loyal affinity group with strong purchase propensities. The company also uses location-based services such as Foursquare to offer rewards to customers for brand advocacy. Its smartphone app allows customers to opt in to messages based on age, gender, interests, and location, which enables Starbucks to tailor promotions to specific audiences.
Some upscale retailers, such as Nordstrom, and financial services firms serving wealthy customers believe that the best channel for delivering an offer is a human being. Many organizations provide multiple offers, usually ranked by the customer’s propensity to accept them. A salesperson can select an offer based on real-time perceived receptivity and comfort level with the client. When a salesperson delivers offers, a delicate interplay often occurs between the sales-person’s perceptions of the customer and the offers presented by the model. Insisting that a salesperson deliver an offer in all cases may create lower satisfaction and reduced offer compliance. The investment firm T. Rowe Price estimates that its targeted offers shouldn’t be delivered more than 50% of the time. Otherwise, the employee probably isn’t tuning into what the customer really wants.
Online offers are less personal but can be sophisticated. Traditionally, online marketers have created a few different email offers and sent them to selected customer segments, and the offer is designed before the customer opens the message. However, sophisticated companies such as Microsoft are approaching offers much more dynamically. For example, emailed offers for trying or becoming more engaged with the Bing search engine are customized at the time of opening. In 200 milliseconds, a lag time imperceptible to customers, the offer is assembled based on the most recent responses of other customers and the available real-time information about the current customer. The real-time targeted ads have lifted conversion rates between 20% and 70% under different circumstances.
Given the richness, diversity, and inherent personal nature of much of the data used for NBOs, a large array of issues naturally emerge that touch on legal, ethical, political, and public policy concepts. These issues have largely been in the background thus far, but they are becoming more prominent. They address not just identity protection or privacy, but also how pervasive and invasive the offers can be as perceived by customers—the offers’ mood, tone, and feel.
Here are some questions central to this topic:
• How might consumers be fairly or unfairly treated by NBOs?
• Are offers being made based on truthful and accurate information or erroneous and spurious data?
• Are consumers comfortable with offers that are derived from seemingly unrelated information and that make assumptions about propensities?
• Are consumers aware and accepting of evolving data usage capabilities, and can they opt into or out of future data usage techniques?
• How might a consumer react if an offer results in a “false positive” and the offer insults or somehow offends the person’s sensibilities?
This abbreviated list of concerns includes some critical topics that need specific and holistic consideration. Some touch on highly technical and controversial legal regulations articulated in laws such as the Fair Credit Reporting Act (FCRA). Outside of the U.S., particularly in the EU, regulations limiting the use of consumer information for NBOs can be much more restrictive.
Learning from and Adapting NBOs
Because offer creation is an inexact, but constantly improving, science, one of the most important components of a successful NBO process is to learn from and adapt to results. Some offers will meet customer needs better than others, so there must be a way to measure and improve success, both in the aggregate and for individual customers. The best way to view NBOs, as a CVS executive noted, is “every offer is a test.” If you don’t constantly try out new variables, algorithms, and business rules, your offers won’t get better.
One way to learn from offers is to articulate some rules of thumb that govern the creation of offers. These will differ for each company, and it’s important to articulate them explicitly so that they can guide offers. Here are some rules of thumb we derived from our discussions with companies:
• Up-sell happens only face-to-face (U.S. retail bank).
• Only fashion-forward shoes are discussed through social media (FootLocker).
• Our customers like offers that provide discounts on the same things they have bought previously (CVS).
• Offering a substantial discount on relevant items in categories where we would like to earn our member’s business creates incremental value for us and our suppliers (Sam’s Club).
• Our offers should generally be provided directly through our customer’s relationship to sales associates via face-to-face customer interactions supported by powerful predictive analytical tools at the point-of-sale (Nordstrom).
• Customers don’t seek to buy banking services often, so we need to partner with other providers to build the relationship (European bank).
Rules of thumb should be based on data-driven and fact-based analyses, not on convention or lore. And they should be tested occasionally to ensure that they still apply.
The key to NBOs is progress and innovation through action. It would be very difficult for a retailer today to incorporate all the possible variables into an NBO model. But it certainly makes sense to gather and incorporate key variables, such as basic demographics and customer purchase history. Most retailers, in fact, need to accelerate their work in this area because customers are not impressed by the quality and value of offers thus far. Channels and predictive variables will continue to grow in number, so if next best offers aren’t quickly improving and evolving, they will only fall further behind.