 | Collaboration. Communication. Community. |
|
Customer Intelligence
Customer intelligence is a process that leverages the capabilities of business intelligence in the context of customer relationship management.
There is no greater truth than what your customer is telling you. —Guy Abramo, CIO, Ingram Micro
A fair bet is that most businesses do not know much about their customers. Even companies furthest advanced on the journey to achieving profitable customer intimacy are still struggling with what
to make of the tidal wave of information they are collecting on their customers. They are struggling to understand who the customer is, what the customer wants, when, how, and why the customer
wants it—the fundamental questions that all companies should be seeking to answer. Most businesses are now on a mission to become more customer-centric. The emphasis is shifting from transactions, processes, products, and channels to the ultimate source of immediate and long term profitability—the customers. This change is driven by intensified competition, deregulation, globalization, and saturation of market segments. Last, though not least, this change is driven by the Internet, which offers a bonanza of choices online, thereby raising customer expectations. Product differentiation is eroding. The competition is just a click away, and the switching costs for customers are becoming ordinarily negligible. Companies find it harder and harder to differentiate on factors that prevailed in the 1980s—product quality, operations, logistics, and business processes. The quality of many products has improved, many businesses have streamlined their supply and distribution chains, and companies have benefited from the widespread business process reengineering exercises of the 1980s and 1990s. In this new ruthless environment, it has become essential for companies to find new ways to attract new customers, to maximize the value of each existing customer, and to retain the most profitable ones. Numerous studies show that it is easier and as much as six times less expensive to sell to an existing customer than to acquire a new one. This can be achieved by:
To accomplish all of the above, you will need to collect and analyze all the pieces of information available to your organization that relate to the customer. This will allow you to understand the customer’s profitability, as well as his or her expectations and preferences. With this knowledge you can then determine the customer’s lifetime value to your organization. Note that to understand the value of the customer, you must look not only back in time, but also forward in an attempt to predict his or her future potential. Once you understand the lifetime value of the customer, you can then define and take the actions required to materialize that value. It is this collection and analysis of information, and the resulting actions based on our intimate understanding of the customer, that we call customer intelligence. Customer intelligence will ultimately allow you to deliver the best service and interaction with the customer.
A study by the Meta Group analyst firm in March 2000, revealed statistics showing that business has a long way to go to solve this multifaceted customer riddle. (See Figure 1.) In Meta
Group’s survey of 800 business and IT executives, 83 percent of respondents answered “no” to the fundamental question, Does your company know who its customers are? And the
survey found that 67 percent did not feel that their companies were effectively using client data to understand their customers. But efforts are intensifying to get to know these customers—56
percent of Meta Group’s respondents counted improving customer intimacy as among the top three priorities of their companies. Achieving customer intelligence ultimately involves two very important objectives:
First we will review these two concepts and the challenges of putting them to work, and then we will discuss the basics of customer relationship management (CRM) and how customer intelligence enables a more intelligent CRM. Finally, we will discuss the concept of Customer Value Management. The 360° Relationship
Understanding your customer base means knowing much more than their demographic distribution. It means having a full view not only of their purchase history, but also of all their interactions with
your company. Interactions can happen through multiple sales channels (direct sales, online purchases, resellers, etc.) and through multiple points of contact or “touchpoints” (customer
support center, marketing programs, sales force interaction, and so on). By tracking, observing, and analyzing these interactions, you can gain great insight into the customer’s needs and
desires at any
Traditionally, the various departments in contact with customers have a good understanding of their own communication with the customers, but rarely know much about the other interactions. A
salesperson knows how many calls she made to a customer for a potential new sale. A customer support representative knows how many times that customer has called to complain about product problems.
A marketing manager knows how many times he called that customer to serve as a reference. Typically, none of that information is shared across these different functions, leaving all employees
involved with only partial information. To address their expectations effectively and efficiently, each of the company’s This holistic picture of the customer is what we call the “360° view.” The business intelligence software implementation of this 360° view can take several forms, including a dashboard of business indicators on a screen, a set of business reports accessible via a Web portal, or an environment that delivers direct ad hoc access to the customer data. The key is to integrate in a single environment the related data that comes from all points of interaction with the customer. Achieving this is critical to:
Why Most Companies Do Not Have the 360° View Most companies do not have a 360° view of their customers. The reason for this is that they have yet to integrate the disparate departmental systems, the data silos that we talked about earlier. In this age of multidivisional companies, business units, decentralization, mergers, and acquisitions, it is often impossible for the left hand of the organization to know what the right hand is doing. Add to this the challenge of dealing with inconsistent data—the credit card division might use its own call center software and a certain customer numbering convention, while the mortgage division may use a different one. The credit card system may contain a Ben Jones of 321 De Anza Drive as Customer #127, while the mortgage system contains a Ben Jones of 321 Deanza Ln as Customer #235. It is difficult for the company to realize that Ben Jones #127 is the same person as Ben Jones #235. Disparate systems and inconsistent identification methods are the primary deterrents to attaining a 360° view. The parent company must therefore balance the requirement for common data conventions with the need to maintain the autonomy and empowerment of each of its business units. To accomplish this, it will need to enforce at least one of the following:
Customer Profitability: A Tale of Two Customers
How should your credit card division handle a $15 disputed fee? From a business perspective, the answer should depend on the business value of the customer making the inquiry. Look at customers A
and B in Table 1. When you look at the whole table, it is readily apparent that A is a far superior customer for whom you would literally bend over backward. B, on the other hand, is not a great
customer. Now look only at the third row of the table, which represents the perspective of the individual business unit that runs credit cards. From the business unit perspective, A and B are
identical. You might deny the adjustment to both and might deeply regret having done so when customer A takes her business elsewhere. It was once thought that Pareto’s Law applied to profitability—20 percent of the customers generate 80 percent of the profit. As organizations started to focus on customer profitability, many realized that Pareto’s Law should be reversed: 80 percent of the customers generate 120 percent of the profit. The other 20 percent of the customers actually cost you money, and you would either be better off without them, or at least you would be better off restructuring your offers to make the relationship a profitable one.
The issue is that most companies cannot discern the difference in profitability between customer A and customer B. Most organizations can only see customer information within their stovepipe or business unit—the credit card division, for example, can only see customer interaction within that division. With the 360° view of the customer, the person, or computer, interacting with that customer at a given touchpoint—order entry, email response management, call centers, customer support systems, the Internet, kiosks, and in person—can know everything about the customer’s relationship with the business. Another common problem is that companies often use only lag metrics—such as past sales, or past costs generated through calls to the call center—to define the profitability of a customer. This method can only give an idea of the past profitability of a customer and does not necessarily give great insight into what the potential future profitability of that customer may be. Future profitability is the most important metric to track, but it is hard to identify, because it is based on prediction. By combining different sources of information, not only past transactions, but current interactions and behavior, the forward profitability is a metric that can be estimated and therefore acted upon. Beyond the 360° View, the 360° Relationship Achieving the 360° view of the customer is very important, but in some ways, it is not enough. A company does not simply want to be able to see all aspects of the customer relationship; it wants to be able to act on that information and to act in a consistent manner regardless of the customer interaction or touchpoint. What is needed is not simply a 360° view of the customer, but a 360° relationship with the customer. Once your organization has achieved a 360° view, you can then see that only one Ben Jones exists in the company records, although his name appears under two slightly different addresses, in the mortgage and the credit card divisions, respectively. You want to be able to not only recognize Ben Jones consistently across multiple divisions, but to take consistent actions with him, regardless of the point of his interaction with your company. For example, if you have decided that Ben is a good target for a home equity loan, then you would like to (1) mail him an offer, (2) discuss this possibility with him next time he visits a branch, (3) pop up this idea the next time he visits your Web site, and (4) discuss this idea with him the next time he calls your call center. Conversely, customers expect the companies they do business with to be knowledgeable about them, regardless of what communication channel they choose to use. If a customer has ordered a sweater through an online store and a defective piece of merchandise arrives, she expects to be able to call an 800 number and speak with a customer support representative who has details of the transaction at his fingertips. A 20-minute telephone ordeal with a support representative who has to request information that the customer supplied on the Web site does nothing to earn repeat business.
While this 360° relationship makes a lot of sense, few organizations are able to both recognize a customer across multiple divisions and perform consistent actions with that customer
regardless of the interaction point. Putting the infrastructure in place necessary to do this is a requisite first step to reaching customer intelligence. The Utopian Segment of One: Segmentation 101 The second key concept in customer intelligence is customer segmentation. The way companies have traditionally tried to get closer to their customers is through realizing that customers are not all the same, and that each of them has potentially different needs and desires when it comes to the use of products and services. In the early days of the industrial era, when production and distribution ruled, this was not the case. Henry Ford famously proclaimed on the availability of Model T’s: “Any color you want as long as it is black.” Such arrogance would be punished today, of course. Customers are not the same, and they want different things from a product or service. If they cannot find satisfaction from your company, they will go to your competitor. Some people will buy a Mercedes Benz as a status symbol, others for durability, and others for resale value. Knowing which customers are buying for what reasons can help in focusing your marketing and business strategy. Because customers are different, companies have long known that they should attempt to group similar customers into buckets, or market segments, in which the customers have common characteristics or interests. The goal of customer segmentation is, of course, not only to group the customers into common segments, but also to treat them in a way that meets the unique needs and interest of the segment. Pitching a senior citizen a skateboard, or pitching a preteen a new Mercedes is a waste of everyone’s time and effort. Sometimes, you might think that the segment and its needs are obvious. But it does not always work that way, and segmentation can sometimes surprise you. A case in point is Mobil Oil. Mobil assumed that buyers of its highest grade of gasoline would predominantly be affluent people who drove expensive vehicles—a Lexus, a Mercedes, a BMW. These people would be discriminating motorists certain to feed their cars with the highest octane fuel to maximize performance and extend the vehicle’s longevity, or so the thinking went. Based on this assumption and a limited degree of data analysis, Mobil’s target audience for advertising and marketing its premium fuels would be people who watched televised golf and PBS, read Conde Nast and Esquire, and shopped at Nordstrom. Guy Abramo had a surprise for Mobil. The oil company hooked up with the Center for Data Insight—a research center created in 1997 through a partnership between Northern Arizona University and the consulting firm KPMG—for which Abramo was working as managing director of marketing intelligence. He later joined Ingram Micro, the $29 billion computer distributor, where he now serves as CIO. As a proof of concept, the center ran data from Mobil’s retail stores division through a series of data mining exercises. The researchers at the center took about a year’s worth of customer point-of-sale data from credit cards, and enhanced it with demographic information from a third-party provider. The results upended Mobil’s belief that buyers of its highest grade fuels would be affluent people.
Abramo and others from the Center and KPMG presented their findings from this proof-of-concept exercise at a meeting in the summer of 1997, at Mobil’s headquarters in Fairfax, Virginia:
“Our data showed it was just the opposite,” said Abramo. “They expected to see Lexus and Cadillac drivers, but what we found was that people buying super unleaded owned older
model vehicles and tended to be part of the blue collar demographic. They were the ’69 Buick drivers. That was not what they expected. I think the biggest impact was recognition of what they
did not know and the fear of creating multimillion dollar advertising campaigns on the basis of perception rather than reality.” Further analysis upended another assumption—that buyers of premium fuel would tend to be loyal to the Mobil brand. In fact, they were not as loyal as Mobil thought. Customers of Mobil’s regular unleaded gasoline had much higher loyalty scores. Armed with intelligence from data that had remained unexploited in its data stores, Mobil was now prepared to tailor its advertising and marketing accordingly to target markets, build customer loyalty, and drive revenues. The Limits of Demographic Segmentation Many companies use demographic segments—segments based on standard demographic measures, such as age, income, geography, gender, and marital status—to divide up their customer base and create marketing programs. The problem with demographic segments is that they tend to be very large or coarsely grained. If you have 8 million customers, as do many phone companies, retail banks, and consumer e-commerce companies, and you place them in eight segments, you would end up with an average of 1 million customers per segment. While you have segmented your customer base in recognition of the fact that not all customers are the same, you have done so at an extremely coarse granularity. The problem is that you may be missing major marketing opportunities in much more finely grained or smaller segments in the customer base. Because of the coarse grained segmentation, major differences among individual members of a segment—say, the 1 million people who own minivans and have kids who play soccer—are overlooked. Clearly, many customers are interested in many different things. Telling them apart and finding much finer grained segments has become a new goal. In fact, the inability to get beyond very coarse grained segments is a key reason that standard response rates for direct marketing, such as direct mail, are only about 2 percent. For every 100 promotional mail pieces sent out, only two responses, on average, are received. As one direct marketer stated: “You can say no to me. I have very thick skin. Ninety-eight percent of the offers I make are rejected and thrown away.” Clearly, there is a huge opportunity to eliminate waste and improve productivity. This can be accomplished by delivering better focused marketing campaigns through finer grained segmentation. Doing so will result in significantly higher response rates. One example is British Airways. The airline’s customer intelligence system helps it determine how to best serve its customers, reduce its costs, and boost revenues by analyzing multiple variables. The airline analyzes routes, market share data, and bookings through travel agents and computerized reservation systems to calibrate its routes and schedules. Analyzing market share data against British Airway’s internal information gives the company interesting insights into who its competitors are and where the passenger feed comes from for particular routes, says Peter Blundell, British Airways’ knowledge strategy manager. Understanding changing market dynamics enables marketing campaigns to be launched into areas likely to be fruitful.
“You want to segment your market so that you do not flood the market with deals that nobody’s going to take up,” Blundell says. “Getting the right offer to the right group
is very important. Customer intelligence enables us to understand what our booking profiles and customer profiles look like, so that we can make the right offer to the right person.”
“A student, for instance, is time rich and cash poor. A student might have a huge lifetime value, and if you can capture the student’s business at that point with the right offer you
have been successful,” Blundell says. “The offers that a student would be interested in are much different than the offers you would make to a businessperson who is cash rich but time
poor. That person is much more interested in having the best possible schedule, maximizing his or her time, The 1:1 Future In a bold attempt to address this rather desperate situation, Don Peppers and Martha Rogers introduced the notion of one-to-one marketing in their hit book, The One to One Future. In this book, the authors argue that businesses should move away from coarse grained segments of one million customers toward fine grained segments, with the ultimate goal of reaching the segment of one—to treat each customer as the individual that he or she is. They argue that the business world should move from one-to-one million marketing to one-to-one marketing. The key recommendations of the 1:1 future were to treat customers as individuals, based on a holistic view, with consistent actions across touchpoints, and to think in terms of share of wallet, not share of market. Fundamentally, the book “turned sidewise” many classic ideas based on product-centric marketing concepts, such as positioning and the drive for market leadership in product categories. While those concepts are still quite valid, Peppers and Rogers reminded marketers to think simultaneously across two dimensions —the classic product dimension, as well as the newly conceived customer dimension. In practice, the primary result of the one-to-one movement is seen in the personalization of e-commerce Web sites and the widespread use of personalization engines to deliver content that has been tailored based upon an analysis of available customer data. However, the basis of personalization is all too often only e-commerce data and not data gathered from customer interactions with other touchpoints, such as the call center, customer support, and store sales. With an understanding of the 360° customer view, segmentation, and one-to-one marketing, companies are ready to explore ways of acquiring, building, and caring for customer relationships. The ABCs of CRM At its heart, CRM is about doing three things to maximize the lifetime value of customer relationships. 1. Acquire. Attract new customers intelligently. An intelligent CRM implementation can enable your company to:
2. Build. Build your customer’s value over time. Once it has acquired a new customer, your company must work hard to increase the value of that relationship over time. This means understanding your customers enough so that they are enticed to repeat business, through cross-selling or up-selling programs. And it also means having the ability to measure and track that customer value over time. 3. Care. Provide the best level of service to customers. Good customer care translates into customer retention. With a solid informational infrastructure built on customer intelligence, you are able to better understand their needs, make your company easy to do business with, and provide customers with a multi-channel, self service environment. Optimizing the value of a customer relationship is like moving through a three-dimensional chart (see Figure 2). The y-axis of this chart is the number of customers that your organization has, the z-axis is the value of the relationship, and the x-axis is the duration of the relationship. Customer acquisition (acquire) moves your company up the y-axis. Maximizing the value of the customer relationship through techniques such as cross selling and up selling (build) moves your company along the z-axis. Caring for customers over the long term (care) moves your organization along the x-axis, increasing the duration of the relationship. Customer intelligence can help your organization take steps to move from wherever you are to the upper right hand corner of this chart.
A for Acquire: New Customer Acquisition When we talk about acquiring customers, we are fundamentally talking about optimizing the “target-to-conversion” loop. Key questions include:
Targeting
Targeting, as the name implies, is figuring out the desired target audience for a marketing promotion. In the context of marketing to existing customers, targeting is typically done using the
segments that you have created for your customer base—you select a few of the predefined customer segments as the target for a promotion, or you perform some basic operations on those
segments to create a target. For instance, you may seek to target the number of high-value customers who have not bought recently. To do so, you will need to identify the intersection between the
segment of high-value customers with the segment of customers who have not bought recently. This is obviously a very important target, as these are high-profit customers whose inactivity could be a
sign that they are defecting to another supplier. A timely promotion—a “love letter”—along with a nice coupon with an incentive for them to make a quick repeat purchase may
be just what the doctor ordered to keep them in the corporate fold. Once you have segmented your customer base, you are in a position to improve your targeting practices. You may evolve from the standard approach of having customer marketing personnel running programs to existing customers to having market segment managers who manage the company’s relationship with customers in one or more segments, and to maximize the revenues and profits derived from those segments. Therefore, the targeting process will be driven by a market segment manager who has specific revenue and profit objectives for a segment. The most traditional way of targeting is to define the target audience for either a new marketing program, or more broadly, a new marketing campaign that consists of a series of integrated programs. For example, say you are launching a new telecommunications switch and wish to build an integrated campaign that uses direct mail (both email and snail mail), advertising (both print and banner), and trade shows to inform the target audience (network designers in telecom companies) about the new product. Because you know your market well, you know there are a large number of trade shows, magazines, Web sites, and mailing lists available to use in the product launch. The question is which trade shows should we attend, which magazines and Web sites should we advertise in, and which mailing lists should we rent? Customer intelligence provides value in this context by giving you the power to analyze the results of historically similar programs. On the mail side, and even on the advertising side, you can drill into this information to see which subsets of the total mailing list you mailed to, which cuts worked best, and which demographic versions of a magazine your advertisements ran in. Armed with this information, your marketing managers are in a strong position to make good decisions when targeting their marketing programs. Another way to perform targeting is to exploit relationship networks. Savvy realtors will ask a purchaser if they have friends interested in selling a house, and car salespeople will ask a similar question about cars. Network penetration refers to the idea of “building out” from a single customer to reach his or her family and friends, or in the business-to-business realm to start with a single department and then branch out to other departments, and then other divisions. Businesses such as Amway have had great success exploiting these relationship networks. Now, new techniques known as affiliate networks are emerging on the Web to perform the virtual equivalent of Amway—using relationships to penetrate people’s business or personal networks. A simple example of this is Amazon.com’s affiliate network program, under which tens of thousands of affiliates have signed up to sell books on Amazon. The affiliate simply joins the program and places a few links on its site that either list, or more powerfully, recommend various books. When a customer of the affiliate’s site is persuaded to buy a book, the customer presses the link, gets directed to Amazon, and has his or her order taken and filled. Then the affiliate is given a finder’s fee (typically in the single digit percentage point range). This model is very powerful because it provides true added value. If someone is visiting his or her favorite management consultant’s site and reading information on leadership, what better time and place for that management consultant to recommend the person’s favorite readings. Enterprise Marketing Automation Once a marketing department runs a campaign, the question becomes what to do with those people who respond. In some cases, such as a direct mail campaign pitching a child’s CD, life is simple—you take the orders and mail the disks. In most cases, however, particularly in business-to-business environments, life is much more complex. No one orders 10 aircraft engines or mainframe computers as the result of a direct mail piece. That is not to say that direct mail is not valuable in the sales of those products. It is just more difficult to measure, because the industrial sales cycle is complex, in that the purchasing decision is made by a group of people within the targeted organization, and a company will interact numerous times over an extended period with each of those people. An operational CRM system known as an enterprise marketing automation system, or a campaign management system, provides a systematic means of handling the responses from marketing programs. No matter what technology or implementation approach is chosen, the business goal remains the same—to handle the influx of leads, score them, and then place them into a process as a result of the score. For example, customers who “have budget and are going to buy in the next 30 days” might immediately be sent to sales. But what do you do with customers who are "just looking" and who indicate no intent to purchase? You may have spent $50 to $150 to get the response. Do you throw them into the virtual bit bucket because they are not ready to buy right now? Of course not, you place them into a process often known conceptually as "the incubator." From here, they may be periodically contacted and reminded of your company’s products.
Customer intelligence plays a key role in enterprise marketing automation, because it allows you to measure not only the effectiveness of your targeting, but the relative effectiveness of the
different incubation paths into which you place customers. Say that all A leads are passed to sales and all B and C leads are incubated, and that you try four
Customer intelligence will let you analyze the relative effectiveness of these incubation methods. As many things in marketing still remain a “black art,” the enlightened marketer is driven by the desire to constantly experiment, and then to measure the variances driven by that experimentation. Customer intelligence provides marketing with the information systems backbone to do just that. Sales Force Automation Say the lead has been successfully incubated and is now ready to be passed to the sales organization, and that, depending on either thesize of the opportunity or its location, you might pass it to your sales force, or to one of your distribution partners. A number of questions need to be answered:
Answering these problems is the province of sales force automation (SFA) systems. SFA systems take leads, assign them to sales representatives, help them execute a sales process and methodology, help to track attributes of the deals (e.g., competition, application), and help to keep track of individuals and their relationships within companies and when they change companies. They also assist with forecasting and commissions generation. Business intelligence is often used for standard and ad hoc reporting and analysis, against the underlying SFA database. While the real power of business intelligence is unleashed in a 360° customer warehouse, using it to access the SFA database will lead to answers to the preceding questions. That is the practice at Heineken’s beer sales operation in Madrid, Spain. Some 150 salespeople for Heineken’s El Aguila brand of beer are outfitted with handheld devices to manage taking orders, deliveries, discounting, and routes while in the field. The data they collect is entered into a sales force automation application. There, through a Web-enabled customer intelligence solution, the information is accessed by the Heineken sales managers and administrative personnel to measure efficiencies, assess profitability, size up sales performance, and track problems and issues that affect customers.
“Our vision has been to provide tools to the sales force that allow them more time to dedicate to customer management,” Darrell Proctor, information systems manager, says. “As
they must manage their sales routes profitably it is important they have the most up-to-date information about their customers, including deliveries, pricing and discounts, and any incidents that
have occurred and impacted the customer.” Measuring Marketing Effectiveness One of the most powerful ways to use business intelligence for customer analysis is to apply it to the particularly nasty problem of evaluating marketing effectiveness in complex business-to-business sales cycles. This can be done stand-alone or as part of a broader effort to measure customer profitability. The key problem is that business-to-business sales processes are not as simple as filling in a card and giving a credit card number. Multiple programs touch multiple people over an extended period of time that result in an initial order and then follow-up purchases. What is needed is a way to look at a set of individuals in an organization over time, the marketing programs that they experienced, and then the net resulting revenues for the organization. Only in this way, by aggregating the costs and benefits of business-to-business marketing, are you able to get a realistic picture of the return on investment (ROI) of business-to-business marketing programs. Evaluating such programs in isolation is often done, but the results are questionable as they ignore the greater reality of business-to-business selling. Often, in performing such ROI analyses, it is better to start backward. That is, rather than ask the question “what happened to all the leads I generated in Q1,” the ROI analysis is more effective by looking at the question backward: “Of all the sales in Q3, tell me the marketing I did to these organizations.” The power of set-based analysis, explored in the next section, is very useful in doing this type of analysis because you can identify a set of individuals, perform a series of actions on them, and watch the evolution of that set of people over time. One interesting analytical test to perform on a single program, however, is to look at which was the final marketing program that “pushed the prospect over” into becoming a customer. So, rather than calculate ROI based on this figure, you are actually trying to determine the correct sequencing for marketing programs as a function of where someone is in the sales cycle. Finally, we should add that this whole analytic process may be further improved through integration of third-party data with your internal data. A number of firms provide third-party evaluation of the evolution of customers down the demand pipe for different industries, by measuring items such as awareness, positive opinion, consideration, trial, purchase, and repeat purchase. Moreover, they can calculate standard conversion ratios for each movement along the demand pipe. For instance, these analyses can evaluate what percentage of those who were aware of a product offering actually considered a purchase. This sort of third-party data is useful when analyzed stand-alone as a sanity check to look for "leaks in your demand pipe" and common causes for them. It becomes more powerful when combined with your own analysis of your lead funnel and demand pipe, or with additional third-party data, such as advertising expenditures. B for Build: Maximizing Customer Value Once acquired, you need to ensure that you maximize the value of the customer over time. This is where most of the heavy lifting of marketing segmentation is actually performed, because you will first wish to segment your customers, and then to perform additional marketing to them as a function of their segment. Segmentation for the Elite: Data Mining As we saw in the market analysis example of Chapter 5, there are two basic techniques for customer segmentation: data mining and set-based analysis. Data mining is the high-end approach to customer segmentation. Ph.D.s in lab coats use rocket science techniques such as neural networks, chi-squared automatic interaction detection, case-based reasoning, and genetic algorithms to determine segments inside the customer base. These advanced techniques are used in industries such as credit card management, for complex analysis such as credit card fraud. The purpose of data mining is to build a mathematical model that represents accurately the attributes of the customer base. The idea is that if one can find a model or hypothesis that describes well the characteristics of existing customers, the same model could be used to identify future prospects or opportunities. A frequently cited example of such a data mining model, almost folklore in the data mining industry, is the model which predicted that grocery store purchasers of diapers were also very likely to buy beer.
Building the model involves a great deal of time and effort—neural networks, for instance, need to be trained on an existing set of data, and only after a significant amount of training will
they attempt to predict future behavior or do segmentation. The task also requires a great deal of expertise. It is not uncommon to have a staff of statisticians spend six to nine months to build a
data mining model. These models are then deployed to help make marketing and Data mining models are very powerful, and they have delivered great value in addressing high-end analysis problems. The problem, however, with data mining is that it is inaccessible to most people, and it has not yet been adapted to a marketing campaign that needs to be designed and implemented quickly. 2The models are built by a small group of specialists, meaning that relatively few models may be made, they take a long time to build, and the models are usually “black box”—there is no easy way to understand what the model is or how it was built. Segmentation for the Rest of Us: Set-Based Analysis What is needed is a more accessible segmentation technique that can be practically applied and used every day by mainstream marketers, without the assistance of Ph.D.s. What is needed is a democratization of the ability to do segmentation, much as a general democratization of data access was needed 10 years ago. Managers need to be able to define their own segments and quickly determine:
Set-based analysis provides just such a technique—a way of performing segmentation that is easy to understand, and regular marketing staff can easily be trained to do it. Through new
techniques like stepwise queries and visual selection, marketers can create sets, or customer segments that can be reused throughout the organization by the A very useful approach when analyzing customer segments is a technique known as stepwise query. This method lets marketers take a gradual approach when defining their customer segments. The idea is for them to take simple steps and sequence their construction of segments. For example, a marketing analyst might want to create a segment that consists of all airline “frequent flyers” who fly out of Chicago’s O’Hare Airport. This segment would also include all other flyers who use one of three named U.S.-based airlines, but excluding any customers who fly overseas or travel with children. The marketing manager would:
As organizations learn more about their customers, marketing analysis inevitably becomes more complex. With stepwise querying, the selection process remains quite simple, however complex the final query to the database. It is also much easier to deal with exclusions, which are important for marketing. Stepwise querying makes it easy to plan a marketing activity along the following lines, for example:
The power of stepwise querying is greatly strengthened when used in combination with a technique called visual selection. Visual selection is an important step in the understanding of
customer data. It allows marketers to see a high-level overview of their customer information before they begin a query process. For example, a customer service manager may create a query such as
“how many inbound calls to my service center in Seattle went unanswered after two minutes?” After the query is run, the manager learns that only 30 of its thousands of calls exceeded
the two-minute limit. Using visual selection, this same manager could have seen at a glance that while 30 calls languished at the Seattle call center, a much higher percentage exceeded the
two-minute limit at the company’s Miami call center. Instead of wasting time querying data about the successful call center in Seattle, the customer service manager could act upon the
graphical representation of hold-time information and quickly focus her efforts on fixing the problem in Miami. Visual selection lets the marketer take the structure of the data into account when defining and refining the request for data, and to change direction in midcourse if appropriate. However complex the criteria used to build the segment, the user always knows how a decision will affect the final counts of customers in the segment.
Visual selection gives users an overall grasp of the data at the outset so that they do not go down blind alleys. With a visual selection tool, the user can see how many customers fit a given
criterion before running any queries. If some groups contain very few or no customers, and are therefore not useful for further analysis, this is immediately obvious. A major advantage of set-based analysis is that it makes it easy to arrive at a result of a known size and structure. Marketers often know the structure of the answer at the outset. They commonly ask questions such as I can afford to print 1,000 brochures. Whom should they be sent to? Visual selection enables the user to see how each step in building a query affects the final count. At any point, it is possible to define a sample that fits the requirements of the results in terms of size, content, and structure. Sampling therefore makes it possible to tailor the selection process to arrive at the desired number of customers. Similarly, it is easy to fulfill requirements—such as choosing an equal number of customers in each sales territory, by skewing the results into the desired form. One 2 One. The mobile phone market has grown incredibly fast over the past decade, and the introduction of new networks and services will further fuel that growth. With high rates of churn and the constant introduction of new services, mobile phone companies have to work hard to create a personal relationship with each of their customers. One 2 One, a mobile phone provider in England, uses customer intelligence to support a range of customer services and operational requirements—from fraud management and supply chain evaluation to customer relationship management and the creation of products and services to meet customer needs. One 2 One uses a customer intelligence system to exploit a data warehouse that includes information from customer care systems, telephone switches, prepaid vouchers, and external geo-demographic data. One role that customer intelligence plays is in helping One 2 One identify customers to include in a specific promotion, and then to monitor the success of promotions. “It enables us to refine our customer segmentations,” says Adrian Daniel, One 2 One client team leader. “It ensures we can get the right population groups and then run queries on that group, enabling us to refine the product or promotional offering.” For example, One 2 One was able to produce a customer segment that included all customers who had opted for prepaid phones, but who had bought only one voucher and made few calls. “We mailed them to let them know that vouchers no longer expire, and then track the success of that promotion based on an increase in calls made over time,” Daniel says. One 2 One’s finance department also uses customer intelligence to track customer churn patterns. The company takes a group of customers that joined in a specific month and discounts those who leave at the end of a one-year contract, concentrating on the rest to see why they are leaving or staying. Daniel goes on to say that “Churn is a fact of life in the mobile market. By applying filters we can cut the total five million customer population down to a very small group, enabling us to answer specific questions about that segment.” Types of Analysis
Once you have created your customer segments, you then are in a position to perform a number of common techniques and analyses to help increase customer value. Customers may be segmented by a
variety of characteristics—profitability, risk, revenue generated, geographic areas, lifestyles and life stages, length of time as a Segmenting customers by profitability can help you understand which groups tend to purchase high-profitability items frequently and which groups produce only a marginal profit—or no profit at all. Factors such as transaction history, income, and “recency and frequency” of purchases may be fed into a customer segmentation model that includes spending bands of high, medium, or low, and propensity to purchase by various categories. From this information, forecasts on future customer purchases and profitability are developed. Recency, frequency, and monetary value are key characteristics in customer segmentation, taking into account how people buy over time and the value of their transactions. Customers may be segmented by the sequence in which they buy products, by the channel by which they came to do business with you or by which they conduct the bulk of their business, by sales force territorial region, or by customer needs.
Another segmentation model may examine customers at a given time in a given year to assess customer turnover. Many companies use dozens of segmentation models that are evolving to include more
finely grained data points. In a parallel analogy in which likens a business intelligence system to one of local water distribution, the customer base may be compared to a lake. A lake may remain
fairly level, but its contents are always changing, with water constantly Here is an example that shows several segments and the sorts of objectives that a company might build around them.
C for Care: Customer Service That Qualifies as Care Once a company acquires customers, segments them, and performs actions that build customer value through increasing frequency, up selling, and cross selling, it is well on the way to maximizing customer value. Now it needs to do a good job of servicing the customer and building his or her loyalty. The third dimension of the CRM cube is the duration of the relationship. Several factors influence the duration of a customer relationship:
The surest and most obvious way to lose a customer is to make the customer angry. The most effective way to permanently lose a customer is to see only a part of the total relationship and, based on that, make a bad customer service decision that alienates the customer. The way to ensure a long relationship is to deliver service that delights the customer. That means not only providing a 360° relationship, but also allowing customers to serve themselves. As numerous studies have shown, self service customers not only are happier customers, but they cost less as well. In a CRM context, that customer care might mean anything from providing a searchable database of problems and solutions to technical problems, to allowing someone to see the status of an order, or even to the more sophisticated case of allowing someone to analyze her historical orders and bills. The last example is what we call a customer care extranet, one of the fastest growing areas of customer intelligence. Organizations are flocking to develop extranets and self service data access and analysis as a way of reducing their customer service costs and differentiating their products and services. Finally, the last way, and the most defensive way, of prolonging a customer relationship is to anticipate its end and take preventive action to avoid it. This is the realm of customer churn and has been a typical application of high-end data mining within the highly competitive telecom business. But customer churn does not need to be only used in high-end data mining applications. There are a number of simple steps that can be taken to ensure that you are avoiding churn:
New Forms of Customer Value Management If you follow the advice presented so far, you will be on your way to building strong customer intelligence. Few organizations are at this level today, but among those which are, where are they today, and what problems have they encountered when implementing these ideas? For all its value, the one-to-one marketing theory put into practice presents some practical obstacles. It does not work well for very large customer bases, for example, and it does not deal with fluctuation over time in an efficient way. To solve these problems a new class of analytic applications is emerging, called customer value management (CVM) applications. In the context of a customer base of 8 million, the impossibility of creating 8 million discrete segments is a classic example of the difficulties of turning theory into practice. Sure, lots of segments sound great, but how do you manage an environment of such hyper-segmentation? The answer is simple:
Customer value management (CVM) applications are starting to provide these abilities by allowing companies to build the segments, attach measures to them, and then assign action rules on each
measure, such that the segments can largely be ignored unless a material change is automatically detected. Another issue concerns how dynamic the customer segments are over time. Indeed, experience
shows that people migrate through and among segments. A customer segment that may look stable because its size is not changing, may undergo considerable changes as the people who belong to it come
and leave regularly. Effective marketing often depends on the ability to look at how the same customers change over time. For example, an analysis of churn and the factors that affect it is of
central importance for companies wanting to address the issue of customer loyalty. Change is also the mainstay of event-based marketing: People are receptive to offers when their circumstances
change, although this receptivity seldom lasts very long.
To capture the required information using these tools, it is necessary to plan a specific data structure in advance. For instance, you would want to identify the people who were good customers in February, and then to identify those who join and leave that group in March. Similarly, you may want to identify individual customers who have recently married or had children. SUMMARY
Customer intelligence is a process that leverages the capabilities of business intelligence in the context of customer relationship management. It entails developing a 360° view of the
customer REFERENCES This article is excerpted from the book E-Business Intelligence by Bernard Liautaud with Mark Hammond, McGraw-Hill, 2001.
|
||||||||||||||||||||||||||||||||||||||||
