The Proposed Ten-Step Filtration Process

The best practice approach described here is the outcome of learning accumulated over multiple linkage projects, across organizations in different industries, allowing us to draw from the common observations across these studies. Such an overview has led us to conclude that the ability to conduct successful linkage analyses is both an “art” and a “science.” We hope to bring forth this sensitivity and recognition in this section of the book. We propose that the ability to conduct successful linkage analysis is contingent on its ability to pass through a sequence of filters that represent varying shades of art and science required for success in this domain. An inability of the analyses to pass through a particular filter reduces the probability of success in subsequent steps. Each filter thus affects the probability of success in subsequent steps of the project. Overall, the proposed best practice approach lists and discusses 10 such filters (Figure 9.1). We provide a theoretical rationale as well as empirical support for these filters through real-life case studies. In addition, we discuss these filters along the chronological evolution of the project—that is, from early on in the process to its conclusion. While some of these can be more salient in certain environments, we have experienced that each of them is important in its own way. The sequencing of the filters and the relative importance of each of them can however vary across industries, requiring appropriate customization of the proposed approach. And to reiterate, while the framework here is presented within the context of linking customer attitudes to customer behavior and financial activity data, the discussion is equally applicable for linkages among other key organizational metrics.

At a higher level, the 10 filters we discuss here toward linking attitudinal perceptions of individual customers, to their behavior and financial activity can be organized along four broad themes: project support, understanding of marketplace dynamics, data considerations, and effective communication of results (Figure 9.1). By their basic nature, linkage analyses frequently involve active interaction between the sponsoring and the research organization, as well as between multiple units of the sponsoring organization. This necessitates the need for adequate project support, and these needs are discussed in filters one and two of the proposed best-practice approach. Linkage analyses also require business managers to revisit their understanding of the marketplace dynamics to gauge the viability and appropriateness of undertaking efforts to perform the analyses. The issues related to marketplace dynamics are discussed in the third and fourth filters. If adequate project support and the appropriateness of marketplace dynamics are estimated successfully, the next stage of evolution relates to data considerations for performing the analyses. These are multiple data requirements that require thoughtful consideration for linkage analysis, and these are discussed along fifth, sixth, and seventh filters of the framework. Finally, because of the infancy of such analyses in the applied world, even rigorous linkage efforts are often viewed skeptically, or with lack of interest—if the results are not communicated effectively. Effective communication of results is thus discussed as the last of the four stages of evolution, and specific issues related to communication of results are discussed in the last 3 filters of the 10-step filtration process.

Project Support

Filter 1: Senior Management Buy-In

Linking customer perceptions to actual customer behavior and bottom-line data typically involves the coming together of different functions of the sponsoring firm. For instance, a key requirement for performing linkage analyses is the availability of customer perception data, and the financial activity data for such customers. Typically, the customer data resides with the marketing (research) department of the firm, and the financial data with the accounting or the finance department. More often than not, these departments work isolated from each other and zealously guard the data that reside within their control. Working collaboratively on linkage analysis therefore requires a cultural shift for the departments, which more often than not, is met with resistance.

In such environments, if support for the proposed linkage analysis comes from the very top in the organization, it provides an impetus to these departments to share their data and their learning in a collaborative manner. When the mandate does not come from the very top, and worse when the leaders of the organization do not share the motivation for linkage analysis, the exercise gets challenging. Inordinate amounts of time and resources then need to be spent to elicit the basic information required for analyses. In most of the successful linkage assignments that we have conducted, the interest, the motivation, and the buy-in for linkage have come from very senior managers that have high levels of vested authority. In one project, it was the chief operating officer (COO) of a large financial organization, in another it was the president of a utility company, and in yet another, a senior C-level executive in a health-care organization. All these senior managers were convinced about linking cross-functional performance metrics, and their buy-in and sponsorship made it easier to elicit the cooperation of different people and divisions within each of these companies to work toward a common goal.

Filter 2: Willingness of Sponsoring Individuals to Act as Active Participants

Unlike many other research programs, linkage analysis, by its basic nature, is a highly interactive process that involves frequent communication and exchange of information between the sponsoring and the research firm. For instance, a good strategy to kick off the project is to have a flowprint meeting between the sponsoring firm and the linkage research and analysis team. Such a meeting allows business managers to detail the links that they expect to see in their marketplace, and allows the research analysts to translate such links into potential relationships to be explored in the data. Similar engagements from managers of the sponsoring organization are important throughout the process, requiring significant time commitments from these individuals over the span of the analysis. The success of linkage projects is therefore contingent on the commitment of business managers to act as active participants during the life of the project. And from a process perspective, it becomes important to set realistic expectations early on in the process about the involvement and role of relevant individuals from the sponsoring firm for the successful completion of the linkage exercise.

In our experience, it helps to detail the expected involvement of individuals from the sponsoring firm by flowcharting the overall process of the linkage project. One such example that was used in an engagement is presented in Figure 9.2. The flowchart of the process clearly identifies the interim points where active participation of individuals from the sponsoring firm is sought, allowing these individuals to recognize the importance of their participation toward successful completion of the overall project. We have consistently experienced that involvement from required individuals, if obtained effortlessly, contributes immensely to the success of linkage projects.

Understanding of Marketplace Dynamics

Filter 3: Understanding the Impact of Marketplace Characteristics on the Strength of the Link

Customer loyalty research posits that customers with more favorable attitudes exhibit more favorable behavior, which in turn provides financial benefits to the firm. It is however important to recognize that this relationship is not ubiquitous. Instead, the presence of a positive link between customer perceptions of the firm and their behavior is contingent on certain marketplace characteristics. Therefore, these characteristics must be understood before the customer-financial linkage analysis is attempted in order to gauge the viability of such analyses. In general, a positive link between customer perceptions and their behavior is observed in markets where customers can freely act on their perceptions and attitudes. Therefore, a positive link can be expected in markets where dissatisfied customers can easily defect to alternate suppliers without incurring significant monetary or psychological costs related to such defection. By way of contrast, in other environments, where customers are unable to act as freely on their perceptions of the firm, the presence as well as the strength of the positive link may not be strong. For instance, in order to avoid monetary penalties for breach of contract, unhappy customers might continue to do business with the firm even though these customers may be motivated to seek alternatives. Similarly, in environments characterized by psychological costs associated with defection, such as the effort required to educate a new financial institution about all the complex borrowing and investment needs of banking customers, dissatisfied customers may not defect. Further, for certain basic goods and services, increased customer spending with the firm may be less a function of their loyalty toward the firm and more a function of their personal needs. For example, utility consumption will be related to the size of a customer’s residence and family structure (e.g., presence vs. absence of young children at home), rather than how happy the customer is with a service provider. In such environments, a positive link between customer perceptions and behavior can be expected to be weak or missing.

Interestingly, in work we have done for some industries, where the exit barriers are high for customers, we observe a negative link between customer attitudes and customer behavior. In working with a telecommunication provider, for example, we separated the data into two sets (Figure 9.3). The first of these data was for competitive geographies, wherein the customer was free to select their telecommunications provider. For this set of data, as should be expected, we did find a positive link between customer attitudes and behavior. Customers that reported more favorable attitudes toward the firm also exhibited more favorable behavior in terms of growth in their business with the firm. The other set of data was for geographies where the customers had very limited, if any, choice in selecting their telecommunication provider. For all practical purposes, the firm held a monopoly position in these second set of markets. When we studied the linkages between customer attitudes and customer behavior for this set of data, we were surprised to find a weak negative relationship. We observed that customers who reported less favorable attitudes actually exhibited higher absolute spends as well as higher growths in spends over the last few periods. However, subsequent investigation revealed that in the causality in such cases does not necessarily run from customer behavior to customer attitudes. Instead, customers who spend more with the firm are more likely to have higher expectations and therefore less likely to be satisfied. Thus, in such monopolistic situations, less favorable attitudes do not always lead to less favorable behaviors. Instead, more favorable behaviors may lead to less favorable attitudes.

For any proposed linkage analysis, it is therefore important to scope out the viability of the analysis by paying special attention to identifying the characteristics that define the marketplace that may affect any potential linkage between customer and financial measurements. In the telecommunication example discussed previously, for example, we used the information to demonstrate the financial opportunity at risk in markets that are still monopolistic. We looked at the financial vulnerability of the company in monopoly markets when these markets were to become competitive in the future. The approach helped reinforce the importance of delivering favorable customer experiences not only in competitive markets but also in the currently noncompetitive markets that could open to competition over the next few years.

Filter 4: Availability of an Appropriate Unit of Analysis

Customer perceptions can be meaningfully linked to customer behavior and other bottom-line data, but only if the market and data structures provide a meaningful unit of analysis to perform such linkages. This meaningful unit of analysis could be an individual customer, a household, a retail unit, a group of customers, a geographical region, or even points in time. For instance, for a company that sells its merchandise directly to the customer and has information on the perceptions and behavior of individual customers, an appropriate unit of analysis could be an individual customer. In these cases, individual customer loyalty toward the organization can be meaningfully linked to the financial contribution that each of these customers provides to the firm. In retail environments of hotels and banks, an individual franchise or property could also be treated as a meaningful unit of analysis, wherein the customer loyalty levels at each of these franchise or property levels can be meaningfully linked to the financial performance of these units.

Oftentimes, however, the choice of the unit of analysis is governed by the availability of data. In one recent business-to-business assignment for instance, we collected individual customer perceptions toward the firm from respondents located in various physical sites across the nation. Customer perception data were thus available at an individual site level, which in this study was our preferred unit of analysis. The financial data for these customers was however not available at the level of individual sites. Instead, as a historical practice, the customer activity and financial data were stored only at a regional level, with each region consisting of multiple sites. Thus while we collected customer perception data at the individual site level, our ability to use individual sites as units of analyses was restricted by the nonavailability of customer activity data for individual sites. We therefore modified our research agenda to select these regions as the unit of analysis, even though the choice led to two handicaps: (a) region-level data provided fewer data points to perform the customer-financial linkage analyses; and (b) the quality of data was compromised because individual site-level idiosyncrasies were ignored in aggregating the data to a regional level. The learning however provided stimulus to collect and record financial data for customers at the level of individual sites, to improve the quality of future analyses.

Data Considerations

Filter 5: Availability of Appropriate Data at the Chosen Unit of Analysis

Establishing a link between customer perceptions and downstream measures typically involves eliciting information from databases that may have little in common. It has been our experience that in some organizations the customer and financial databases evolve over time to cater solely to their primary users. The isolated evolution of these databases reduces their commonality and the ability to be linked. For example, the absence of a common customer identifier across the two databases can preclude linking individual customer perceptions to customer activity data, if an individual customer were the preferred unit of analysis. Likewise, the absence of certain financial data for all or some of the customers that report their perceptions to a survey instrument can limit the scope of the linkage analysis. On the other hand, certain survey research programs allow respondents anonymity because of the sociocultural environment or regulatory mandates in which these surveys are administered. When many respondents exercise their right to stay anonymous, the scope of the linkage analysis is curtailed, because these anonymous customers cannot be identified and their business activity levels cannot be determined.

Firms that practice a direct-to-customer business model typically possess better quality of data about their customers. The databases of these firms, by virtue of their business model, are better designed to record behavioral activities of individual customers. Also, by the nature of their business model, it is easier for these firms to furnish more accurate contact information about their customers to collect customer perceptions. For firms that do not follow a direct-to-customer business model, the quality of data about their customers is inversely related to the layers of intermediaries between the firm and the customer. The more separated the firm is from the end user, the lower is the likelihood that the firm will possess rich and reliable data about these users. Researchers thus need to be cognizant of any nonavailability or the inaccuracy of data to help scope their analysis. Such realizations often result in modifications to future data collection strategies, in order to perform the desired linkage analysis at a later point in time.

Filter 6: The Need to Have Information on Customer Buying Requirements

Assuming the linkage analysis effort has successfully passed through the first five filters, the research strategy should now begin to focus on some specific, microissues related to availability of appropriate data. The first of these issues relates to an understanding that in competitive markets, differences in the level of business activity of various customers of the firm is contingent on at least two key factors: differences in the level of loyalty of these customers and differences in the volume of their buying needs for the product category. One customer might thus buy more from a firm vis-à-vis another because this customer shares a greater sense of loyalty toward the firm, as well as because the customer has greater need for the product category in general. Differences in loyalty perceptions are typically gathered in the Customer Loyalty and Relationship Management (CLRM) program survey instruments through measures such as “overall satisfaction.” However, by themselves, these measures may be insufficient for creating effective linkages between customer perceptions and customer behavior. Differences in the buying needs of individual customers should also be acknowledged to meaningfully link customer perceptions and behavior data.

In our experience, most firms undertaking linkage analyses projects do not acknowledge differences in buying needs of their customers while attempting to link customer perception and behavior data. Among firms that do recognize this need, either of two options is typically employed for collecting appropriate data. One, they ask respondents to report the volume of their total product category buying needs as a part of the survey. These self-reported buying needs are then used as proxies for the actual buying needs of these customers. Our experience tells us that generally customers are very forthcoming in sharing data on their total buying needs in a survey-based feedback environment. In sharing data on their buying needs and therefore allowing the firm to estimate a customer’s share of business allocated to the firm, the customers want to send a clear message to the firm. In cases where the customer allocates a large share of their business, say 100%, the customer wants the firm to recognize the level of partnership they expect in reciprocation from the firm. On the other hand, where the current share-of-business allocated by the customer is small, it wants the firm to recognize the total potential available to the firm if it were to work harder in fostering a stronger relationship with the firm.

The other option available to a firm for estimating the buying needs of customers is through proxy variables such as the number of employees that work for the organization, or the total sales of the organization. For instance, the number of white-collar workers reported in a published secondary data source can be used as a proxy for the office stationary needs of the firm, while the square footage can be used as a proxy for the office furniture needs of the same firm. While each of these proxy measures are likely to have some error associated with them, in our experience, they provide a lot of useful information vis-à-vis the alternate option of not having the information at all. For example, in one linkage analysis that reported buying requirements estimated through the proxy method discussed previously, the link between the customer perceptions and behavior improved considerably when such information was included in the analysis (Figure 9.4).

Filter 7: Proper Temporal Alignment of Customer and Financial Data

Customers, in their role as information processors, process their recent experiences with a firm, form overall affective and cognitive attitudes based on these and past experiences, and then act on such attitudes at a future point in time. A satisfied customer will thus plan to continue doing business with a firm, while a dissatisfied customer might plan to defect at a future point in time. Meaningful linkage analysis should therefore have the perceptual and behavioral data that are properly aligned in time, with perceptual data preceding the behavioral data. Absence of the proper temporal alignment can influence the quality of linkage analyses. It has been our experience that most first-generation linkage analysis models commence with data that are not most appropriately aligned in time. These models are still able to demonstrate positive links between customer perceptions and actual behavior, largely because in a mature marketplace, few customers change their behavior drastically over consecutive buying cycles. The linkage results of these first generation models, however, usually improve when customer perceptions are matched with future behavior data, because the alignment is more consistent with the decision-making processes of customers.

In a recent assignment, we performed the analysis with data that were cross-sectional, that is, where customer perceptions did not precede customer behavior that was reported in the data. Instead both the perceptions and behavior data came from the same period of time. Subsequently, we got an opportunity to monitor the behavior of these same customers over the next few months, and to reattempt the linkage analysis with the data that were more properly aligned—with this new behavioral data following the perceptual data in time. In this revised version of the model, the link from customer perceptions to their actual behavior improved significantly, validating the importance of proper temporal alignment of such data. We have similarly observed in other studies that when a link between customer perceptions and lagged customer behavior is attempted, it usually provides an improvement over comparable models that do not allow proper temporal alignment of data. While the amount of lag differs across industries, it should always be greater than or equal to the average buying cycle of the product category, to allow customers a subsequent buying opportunity to exhibit their (lack of) loyalty to the organization.

Effective Communication of Results

Filter 8: The Importance of Validating the Linkage Results

Because of the infancy of linkage analysis in the applied world, even a rigorous and carefully conducted analysis is often viewed skeptically. Business managers often continue to doubt the veracity of analyses that demonstrate links between “soft” customer attitudes and “hard” behavioral and financial data. One effective method to address such skepticism is to provide validation to the results of linkage analysis. For example, validation using a holdout sample can be a potent approach. In applying such an analytical strategy, a random portion of the total data is “safe-vaulted” for subsequent use. All initial linkage analyses are then performed on the remaining “main” portion of the data. Once these analyses are completed and a linkage model has been developed, the holdout sample is used to validate the findings. Specifically, the parameters of the main model are used to estimate the downstream financial numbers for the holdout sample, given the favorability of perceptions observed among customers of this holdout sample. The estimated financial numbers are then juxtaposed with the actual observed financial numbers for the holdout sample, and proximity between these two sets of numbers is used to provide support for the model’s validity.

In one linkage assignment, the sponsoring firm was able to provide us with voluminous data on their large pool of customers, which allowed us the luxury of isolating a random holdout sample for validation. We ran the model on 70% of the total data and then used such modeling results to predict the financial numbers for the 30% holdout sample. The proximity between the predicted and the observed financial numbers for customers of the holdout sample significantly bolstered the confidence of the sponsoring firm in the results of the linkage analysis. In another validation exercise, we used longitudinal data to demonstrate that the firm was able to retain more of its satisfied customers over time, and that these customers actually increased their spending with the firm over a year’s time. The dissatisfied customers, on the other hand, were shown to be more likely to defect to other suppliers or reduce their level of spending with the firm.

Filter 9: Tailoring the Communication Strategy

Linking customer perceptions to their behavior often involves complex analyses, which creates a significant challenge in communicating the results to the users of such information. Business managers are less interested in the analytics than they are in seeing a compelling and easy-to-understand link between customer attitudes and customer behavior. It is thus easy to lose their attention with analytically complex information, and it becomes important to present the analytical findings in a simple and meaningful way. The ability to present the required information in the format desired by these audiences is critical for soliciting their buy-in.

In one option that we frequently exercise, we typically classify each individual customer into categories of loyalty—low, medium, and high—based on the customer’s score to a set of loyalty measures, such as “overall satisfaction.” Then, for each of these categories, we show differences in mean revenue (or other financial measures of interest, such as share of wallet) that the customers provide to the firm to demonstrate the positive link between customer affinity and their business activity with the firm (Figure 9.5). This method of communication validates the positive link between customer perceptions and behavior observed in the analytical model, and makes it easier for business managers to estimate the financial gains associated with moving their customers from the less favorable to the more favorable levels of loyalty. Such a simplified communication strategy also makes it easier to get buy-in from these users of information for the overall linkage analysis program.

Filter 10: The Need to Cater to Two Sets of Audiences Throughout Results Dissemination

Typically, the customer-financial linkage analyses cater to two very different sets of audiences. One of these comprises senior level managers who want empirical validation for a positive link between customer perceptions and behavior, to justify resource investments in CLRM programs. The other set comprises the frontline employees and managers who want more granular and specific information—for example, what should be done differently when the first customer walks in the next morning. Further, it is important for both these sets of audiences to work on the proposed improvement priorities in harmony with all other organizational initiatives so that the organization is working in tandem toward the common goal of improved customer loyalty in the marketplace.

In one research project, we catered to both these sets of audiences by linking transactional data (e.g., customer evaluations of support transactions) to the relational customer database (overall attitudes toward the firm), and then further linking the relational database to customer purchase data. The positive relationship between relational customer perceptions and customer behavior demonstrated that customers with more favorable perceptions also provide greater revenue to the firm. This catered to the needs of senior management and provided them with the required confidence to continue to invest in customer focused measurement programs. The link from the transactional to the relational database, and further to actual customer behavior, in turn helped identify the specific activities during each individual customer transaction that have the most significant impact on overall customer affinity levels, and therefore on the revenue and profits that the customers provide to the firm. For example, the analyses confirmed that specific actions, such as the ability of the firm to decrease “hold time,” where a customer is put on hold while the agent confirms something, can have a much greater positive impact on customer experience vis-à-vis a similar reduction in “wait time”—the time a customer has to wait before an agent answers the call. This helped frontline employees to focus their priorities on areas that were most beneficial to the overall marketplace and to the financial health of the firm.