Progress might have been all right once but it has gone on too long.

—Ogden Nash

Experts define the term data mining largely in terms of its usefulness in uncovering hidden trends or yielding previously unknown insights about the nature of a firm’s customers. SAS Institute defines data mining as “the process of finding anomalies, patterns, and correlations within large data sets to predict outcomes.”¹ Michael J. A. Berry and Gordon S. Linoff, who have written several books on the subject, define data mining as “the process of exploration and analysis, by automatic or semiautomatic means, of large quantities of data in order to discover meaningful patterns and rules,” and, for the customer-centric company, founded on the belief “that business actions should be based on learning, that informed decisions are better than uninformed decisions, and that measuring results is beneficial to the business.”² And Jill Dyché, partner and cofounder of Baseline Consulting, says data mining is “a type of advanced analysis used to determine certain patterns within data . . . most often associated with predictive analysis based on historical detail, and the generation of models for further analysis and query.”³

Rather than limit ourselves to the term data mining, however, we prefer the term customer analytics. Although data mining and customer analytics are not really different things, the analogy to mining itself implies a batch process, with the enterprise searching out nuggets of information and then putting them to use. The reality, however, in the interactive age, is that businesses need to have continuously developing, real-time insights into the nature of their individual customers, not only so that the right marketing campaign can be created and launched, but also so the customer can be given the appropriate offer in real time (real-time analytics), while she is on the phone, reading a newspaper online, shopping at the Web site, or standing at the checkout counter.

Customer analytics, therefore, offers the missing link to understanding customers: prediction.⁴ Prediction helps enterprises use the value of customer information to optimize each interaction with each customer. Today, leading companies integrate the most relevant elements of their customer-analytics algorithms into their actual touch point applications. If a customer behaves a certain way, then the mathematical algorithm can analyze that behavior and instantly access the most relevant offer for that customer, taking into account everything the enterprise knows or is able to predict about each customer, in real time. Customer analytics enables the enterprise to classify, estimate, predict, cluster, and more accurately describe data about customers, using mathematical models and algorithms that ultimately simplify how it views its customer base and how it behaves toward individual customers.

The dilemma facing many companies that amass huge customer databases today is simply how to make sense of the data. Analytical software has become a critical component of the customer-strategy enterprise, and the data scientists who can operate such software are in great demand. The mathematical data models that analytical software can produce are inherently simplifications of the “real world”—they represent how customers have behaved before and will likely behave again. They enable a company to view correlations within large sets of customer data and within and among various parts of its business. By analyzing historic information and applying it to current customer data, these mathematical models and algorithms can predict future events, with varying degrees of accuracy, based not just on the amount of data collected but also on the power of the analysis applied to the data. Using customer analytics, an enterprise can sometimes predict whether a customer will buy a certain product or will defect to a competitor.

Companies produce large amounts of data through a wide array of customer-related business processes, including order entry, billing, reservations, complaint handling, product specification, Web interactions, and sales calls. The data often are fed into a data warehouse, where much of it lies hidden in “data tombs,” forgotten about for years. Often, even when a firm has the customer analytics resources necessary to unleash the value of its data, it soon discovers that much of its information is “dirty” (expired, irrelevant, nonsequential, or nonsensible) and needs to be “cleaned” (eliminated, updated, correlated, and refined). As customer analytics tools and technology become more affordable and easier to use, however, enterprises are starting to feel competitive pressure to improve their capabilities in this area.⁵ The various activities involved in readying customer data for analysis, and the analysis process itself, include:

• Classification, or assigning instances to a group, then using the data to learn the pattern of traits that identify the group to which each instance belongs.
• Estimation, for determining a value for some unknown continuous variable, such as credit card balance or income.
• Regression, which uses existing values to forecast what continuous values are likely to be.
• Prediction, or using historical data to build a model to forecast future behavior.
• Clustering, which maps customers within the database into groups based on their similarities. (See more about clustering in the following section.)

Customer analytics is especially useful for consumer marketing companies that collect transactional data through call centers, Web sites, or electronic points of sale. Banks, credit card companies, telecommunications firms, retailers, and even airlines adopted customer analytics as a vital part of their business operations earlier than other companies. These kinds of companies tend to generate large volumes of customer-specific information in the natural course of operating their businesses, often resulting in vast data warehouses, containing terabytes of data.

Royal Bank of Canada has been focused on customer relationships in its retail banking business for well over a decade and a half now and became a “best practice” case study in this area years ago. One of the secrets of the bank’s success is the fact that it constantly monitors the behavioral cues in its customer database in order to optimize current income results against likely changes in lifetime value for individual customers. It has a great deal of data, but it must also do the right analysis in order to spot the cues. For instance, until recently the bank’s “Behavioural Based Modeling” system calculated the effects its various products and services had on customer lifetime values by using customer-specific revenues, but using the average (i.e., non-customer-specific) cost-to-serve figure. The problem is that banking customers don’t all cost the same to serve. Different customers incur different costs. One customer might prefer dealing with the bank online, for instance, while another might prefer the more expensive teller window. Customers will generate different levels of credit risk, processing charges, and other expenses. After upgrading its software, Royal Bank of Canada began tracking customer-specific costs as well as revenues. The result was that the accuracy of its lifetime value figures improved immensely, with more than 75 percent of its consumer customers moving two or more deciles in rank as a result.

In evaluating its actions for different customers, Royal Bank of Canada optimizes “overall efficiencies,” a term the bank uses to include both current income and lifetime value (LTV) changes in the calculation. One example of a policy change based on maximizing overall efficiencies has to do with “courtesy overdraft limits.” This product is now provided for the vast majority of consumer customers rather than just its heavy-hitters. Each customer’s overdraft limit is set based on that particular customer’s overall relationship with the bank. Anyone who has been a customer for at least 90 days, has a low-risk credit score, and has made at least one deposit in the last month will have some level of overdraft protection. Not only does this enhance each customer’s experience with the bank, but it actually increases the bank’s efficiency during the check-clearing process, reducing the number of write-offs and allowing account managers to focus on sales activities. (It also reduces the number and cost of contact center calls handling irate customers who want their overdraft fees reversed.)

Overall, since initiating the program, the bank has increased the profitability of its average client by 13 percent and increased the number of high-value clients by 20 percent.⁶

Customer analytics contributes to better sales productivity and lower marketing costs in many different ways:

• By making it possible to send more relevant information and offers, analytics helps to improve shopper-to-buyer conversion rates.
• Instead of offering one product to many customers, analytics makes it possible to offer specific and more targeted cross-sell and up-sell opportunities, which can result in measurably increased sales.
• By taking steps to keep customers longer, analytics can help increase customer lifetime value, profitability, and Return on Customer.
• Companies use analytics to improve operational effectiveness through smarter, more relevant (and therefore, usually, faster and less costly) customer service.
• Analytics can be used to reduce the interaction time and effort, making information exchange or transactions easier, faster and, therefore, more likely.
• Analytics improves the customer’s perception of the level of service as a result of relevant messaging during an interaction.
• Analytics makes improved service levels for best customers possible.⁷

For example, Coach, the global designer and retailer of accessories and gifts, has successfully used insight gleaned from customer analytics to deliver exceptional customer experiences with maximum cost effectiveness. Because the Coach mission statement is “treating customers as guests in their own home,” the customer’s in-store experience is key. Predictive analytics have allowed Coach not only to predict how many sales it will generate and identify its top 1 to 2 percent of customers but to predict when those customers will most likely come into the store. Coach incorporates weather forecasts and traffic patterns into its data analysis, ensuring that its staff will be ready when the best customers are most likely to visit rather than needlessly allocating staff resources when they’re not.⁸

Those who use customer analytics, therefore, are trying to create an unobstructed view of the customer, allowing the enterprise, essentially, to see things from the customer’s own perspective. By delving into a customer’s history, analytical programs can help the enterprise customize the way it serves or manufactures a product for a customer to suit that customer’s individual needs. In essence, customer analytics helps the enterprise to transform its customer data into critical business decisions about individual customers. Customer analytics software can reveal hidden trends about a customer and compare her behavior to other customers’ behavior.⁹ In addition, customer analytics can play an important role in customer acquisition, by helping the enterprise decide how to handle different prospects differently and by predicting which ones are more likely to become the most valuable customers.¹⁰

In 2001, Tesco, the U.K. retailer with the highly successful Clubcard frequent-shopper program, bought a 53 percent stake in Dunn Humby, its data-mining partner. In 2006, it raised its stake to 84 percent. Tesco knows that its customer data are its most valuable asset. Dunn Humby and Tesco became partners in 1995, when Tesco was launching its Clubcard initiative, and since then the firm has helped Tesco evaluate and act on what it learns from its Clubcard customers, managing massive data sets and enabling Tesco to increase the value of its customer base.¹¹

Tesco and other astute customer-strategy enterprises have learned that customer data have a dollar value associated with them, and the more accurate the information, the better the enterprise can compete. Customer analytics can provide metadata—information about information—spotting characteristics and trends that enhance customer retention and profitability. Furthermore, customer analytics can be a technique for examining the profitability of specific products that individual customers purchase. As the late Fred Newell pointed out in Loyalty.com, analytics helps profile customers so that characteristics of loyal customers can be identified to predict which prospects will become new customers. Data mining can manage customer relationships by determining characteristics of customers who have left for a competitor so that the enterprise can act to retain customers who are at risk of leaving. Moreover, analytics helps an enterprise learn the mix of products to which a group of customers is attracted so it can learn what the customers value. “With this knowledge,” Newell writes, “we can mine the customer file for similar customers to offer suggestions they are likely to value. Without data and its being analyzed to develop information and knowledge about the way things are happening in the real world, all we have are opinions. Every expert we have talked to gives the same answer: ‘Data mining is knowledge discovery.’”¹² Customer analytics is not a technology—it is a business process.

The next level of analytics might be applying financial characteristics to the data analysis, in order to yield a more accurate view of the actual economic consequences of particular customer actions. For instance, an enterprise might know that a promotion should go to customers fitting a certain profile, but it probably will have more difficulty correlating the cost of the promotion with its likely outcome, at least on a customer-specific basis. In the next epoch of customer analytics, the mathematical algorithms will look across a range of promotions and associated costs to determine which tactics will generate the most profit, ideally taking into consideration the current return as well as the long-term effect on equity simultaneously. Ultimately, customer analytics will generate a revolution in how marketing decisions are made, driving companies increasingly toward solutions based on highly detailed marketing simulations.

In the end, however, the reason for analyzing all of these data is simply to develop a deeper relationship with each customer, in an effort to increase the overall value of the customer base, or as some would say, “optimize” the enterprise’s customer relationships.

In the next section, Dr. James Goodnight, the founder and CEO of SAS Institute, explains how customer intelligence works in this era of data-driven relationships to customers.

Here’s an example of a company that uses customer insight to increase profitability.

Here’s another example that illustrates how data mining can be elevated to insight and prediction of customer behavior.

Summary

Big data and data in the cloud have opened up deeper and wider analytical capabilities to companies whose goal is to build stronger relationships with and better experiences for customers. But one thing is clear: No matter how powerful the algorithms, they are only as good as the data we have about customers, and our intelligence about how to use them.

But note: In a speech given by Claudia Perlich, Chief Data Scientist at Distillery and Adjunct Professor at NYU, in January 2016, she made the point that in studies of Internet activity, there may be a map of the United States that shows “mountains” where the most people are active online generally (or with a particular social Web site or looking for certain topics). So the pile of users in the big cities can be quite tall. But all these maps have a default spot; if the location of the user is unknown through GPS or other means, that person’s location defaults to the center of the specified area (state, city, country, etc.). In the United States, therefore, the most populous region for online activity in the entire country is a Kansas cornfield. The point is, we have to be careful about what we think we know.

We have now covered two critical parts of customer-based enterprise “measurement”: We have examined some of the ways an enterprise can measure the success of its customer value–building initiatives, and we have explored how advanced customer analytics can help predict how a customer will behave in a relationship, how a firm can positively influence that relationship behavior, and how much it’s worth to the enterprise to do so. Analytics affect the “customer” issues for the company, but understanding how value is created for today and tomorrow affects decisions for the chief financial officer; the chief information officer; human resources recruiting, training, evaluation, and compensation; product development; the chief executive review and board decisions; appraisal of merger-and-acquisition opportunities; and even public reporting of competitive advantage.

If we can measure it, we can manage it. That’s why the next chapters are about how to manage an organization to build the value of the customer base. Making the transition to a customer-strategy enterprise requires a careful examination of the way the company is structured and a rethinking of many business processes. Chapter 13 focuses on two key themes: What does a relationship-building enterprise, based strategically on growing customer equity, look like? What are the organizational and transitional requirements to become a customer-based enterprise? Next we take a closer look.

Food for Thought

1. From the customer’s perspective, which is better: to buy through one channel or through several channels? The obvious answer is to have multiple channels available—order from the Web, make returns at the store, check on delivery by phone—and have all of those contact points able to pick up where the last one left off. But is there any advantage—to the customer—of using only one channel? Why does research show that customers who use more than one channel are more likely to be more valuable to a firm than those who use only one?
2. Often, the challenge in using predictive models boils down to a misunderstanding of the nature of cause and effect. Although statistical analysis might reveal that two observable events tend to happen at similar times, it does not necessarily mean that one event “causes” the other. What is more important: to understand what will happen next or why it will happen?
3. Customer analytics can be used for improving retention rates. How?

Glossary

Analysis process

Includes classification, estimation, regression, prediction, and clustering.

Customer analytics

Enables the enterprise to classify, estimate, predict, cluster, and more accurately describe data about customers, using mathematical models and algorithms that ultimately simplify how it views its customer base and how it behaves toward individual customers.

Customer centricity

A “specific approach to doing business that focuses on the customer. Customer/client centric businesses ensure that the customer is at the center of a business philosophy, operations or ideas. These businesses believe that their customers/clients are the only reason they exist and use every means at their disposal to keep the customer/client happy and satisfied.”¹³ At the core of customer centricity is the understanding that customer profitability is at least as important as product profitability.

Customer equity

The sum of all the lifetime values (LTVs) of an enterprise’s current and future customers, or the total value of the enterprise’s customer relationships. A customer-centric company would view customer equity as its principal corporate asset. Also see definition of customer equity in Chapter 1.

Customer experience

The totality of a customer’s individual interactions with a brand, over time.

Customer relationship management (CRM)

As a term, CRM can refer either to the activities and processes a company must engage in to manage individual relationships with its customers (as explored extensively in this textbook), or to the suite of software, data, and analytics tools required to carry out those activities and processes more cost efficiently.

Customer segment

A group of customers who are suited for the same or similar marketing initiatives because of particular qualities or characteristics they have in common.

Customer service

Customer service involves helping a customer gain the full use and advantage of whatever product or service was bought. When something goes wrong with a product, or when a customer has some kind of problem with it, the process of helping the customer overcome this problem is often referred to as customer care.

Data mining

The process of exploration and analysis of large quantities of data in order to discover meaningful patterns and rules.

Event-driven marketing

Marketing campaigns and initiatives characterized by offers or communications that are “triggered” by preidentified events.

Mass marketing

An attempt to appeal to an entire market with one basic marketing strategy utilizing mass distribution and mass media.

Moment of truth (MoT)

Interactions with a customer that have a disproportionate impact on the customer’s emotional connection, and are therefore more likely to drive significant behaviors.

Omnichannel real-time customer dialogue

The “conversation” that happens in real time with a customer, across any and all channels, enabling marketing professionals to deliver the right message in the right channel at the right time to each customer, taking into account both contact policies and company objectives.

Optimized customer campaigns

Marketing campaigns that take into account each different customer’s propensity to buy, and how it relates to profit, based on a product’s margin. Optimized campaigns maintain the smartest balance between customer centricity and the company’s objectives by taking into consideration not just customer likelihoods, but budget and contact policy constraints as well.

Real-time analytics

Instant updates to the customer database that allow services in multiple geographies, communication channels, or product lines to respond to customer needs without waiting for customary weekly or overnight updates.

Share of customer

For a customer-focused enterprise, share of customer is a conceptual metric designed to show what proportion of a customer’s overall need is being met by the enterprise. It is not the same as “share of wallet,” which refers to the specific share of a customer’s spending in a particular product or service category. If, for instance, a family owns two cars, and one of them is your car company’s brand, then you have a 50 percent share of wallet with this customer, in the car category. But by offering maintenance and repairs, insurance, financing, and perhaps even driver training or trip planning, you can substantially increase your “share of customer.”

Social media

Interactive services and Web sites that allow users to create their own content and share their own views for others to consume. Blogs and microblogs (e.g., Twitter) are a form of social media, because users “publish” their opinions or views for everyone. Facebook, LinkedIn, and MySpace are examples of social media that facilitate making contact, interacting with, and following others. YouTube and Flickr are examples of social media that allow users to share creative work with others. Even Wikipedia represents a form of social media, as users collaborate interactively to publish more and more accurate encyclopedia entries.