identifying issues in customer relationship management at
TRANSCRIPT
Identifying Issues in Customer Relationship Management at Merck-Medco
Gholamreza Torkzadeh and Jerry Cha-Jan Chang Department of MIS College of Business
University of Nevada, Las Vegas
Gregory W. Hansen Vice President
Customer Service Operations Merck-Medco
September 2004 Revised March 2005
Please direct correspondence to:
Reza Torkzadeh Department of MIS College of Business
University of Nevada, Las Vegas 4505 Maryland Parkway - Box 456034
Phone: (702) 895-3796 E-mail: [email protected]
2
Identifying Issues in Customer Relationship Management
at Merck-Medco
Abstract
This paper reports the results of a study designed in close collaboration with Merck-Medco to identify key barriers to the success of their customer relationship management. To identify the key factors, we first used focus groups of principle users of the system to brainstorm and generate a list of scenarios and issues. A team of managers, supervisors and customer service representatives then consolidated this list. A 54-item survey was derived from the list and used to collect 1460 responses from the user groups within the company. Data were equally divided into two sets. Exploratory factor analysis was used with the first data set to identify principal factors that explained the majority of problem areas. Structural equation modeling was used with the second data set to further examine, shorten, and confirm the initial list of factors. The study results suggest a five-factor 13-items model that describe barrier to the success of customer relationship management in terms of ‘standard operating procedure compliance’, ‘accountability and ownership’, ‘callback information content’, ‘customer contact process’, and ‘dispensing and replacement process’. These factors explained the majority of customer relationship problems in the company. These measures can be used by the company to plan and monitor remedial response. Evidence of reliability and construct validity is presented for the measurement models and decision-making implications are discussed. Subject Areas: Information System Application, Customer Relationship Management, Call Centers, Structural Equation Model, Instrument Development.
3
1. Introduction
To manage prescription drug benefit for sixty million customers, Merck-Medco has made
a significant investment in its customer relationship management (CRM) over the recent years.
Computer applications and system procedures are developed and used to schedule customer
service representatives and balance call traffic to ensure speed and service quality. A network of
six call centers in five states within the continent of the US handles over 40 million customer
calls per year. Frequently, clients have service penalties associated with the speed of answer and
thus the service is painstakingly managed to avoid penalty.
Ideally, the company would like for the customer service representatives to analyze
customer data online and be able to resolve customer issues at the first contact. Yet, the call
centers and customer service representatives are not always able to resolve all issues online. The
unresolved cases are queued for follow-up by a team of customer service representatives within
each dispensing pharmacy. The majority of customer and client complaints have been traced to
this queuing process. The company has for the past few years struggled to refine this process
with limited success. Reasons for member dissatisfaction are numerous and difficult to prioritize.
In this paper, we report the results of a collaborative effort between academe and practice
to improve customer relationship management at Merck-Medco. With a few exceptions such as
BP and IBM [12, 26], very few studies of this kind are reported in research journals. This study
was designed to accomplish two objectives that together will help management develop
strategies for increasing CRM success. The first objective was to identify primary factors that
result in member dissatisfaction with customer relationship management, more specifically with
the call center. The second objective was to produce a reliable and valid set of measures that can
4
be used by the company and others to monitor employee training effectiveness and remedial
plans. The methodology, sample, and procedures were decided with these objectives in mind.
The subjects in this study were people who directly interacted with the system on a daily
basis to make decisions and serve customers. Through close collaboration with the firm, we
collected a large sample that represents over 75% of all user groups. The level of participation
was influenced by the users’ desire to improve the system and the management involvement in
the study. In the following section we will review the literature on CRM and clarify the
construct. The literature review describes call center issues in a broader perspective of customer
relationship management. Section 3 provides the background for Merck-Medco. Section 4
describes research methodology followed by section 5 that describes data analysis and results.
Discussion and conclusions are provided in sections 6 and 7, respectively.
2. Customer Relationship Management
Customer relationship management incorporates information acquisition, information
storage, and decision support functions to provide customized customer service [23]. It enables
customer representatives to analyze data and address customer needs in order to promote greater
customer satisfaction and retention. It helps organizations to interact with their customers
through a variety of means including phone, web, e-mail, and salesperson. Customer
representatives can access data on customer profile, product, logistics and the like to analyze
problems and provide online and rapid response to customer queries.
Companies use CRM to not only create a customer profile, but also to anticipate
customer needs, conduct market research, and prompt customer purchase [26]. It is suggested
that it costs up to twelve times more to gain a new customer than to retain an existing one [33]. A
5
problem needs to be resolved on the first contact or the chances are the customer with an option
to go elsewhere will never call back [39].
Because of the potential benefits, organizations commit significant hardware, software,
and human resources and often restructure their processes in order to implement CRM. It took
IBM a 4-year initiative to re-engineer its customer relationship management [26]. However,
despite the extensive commitment, it is suggested that many of these systems fail to fulfill
expectations [33, 35, 46]. The lack of proper integration of data across organizational functions
is suggested as one of the reasons why many companies struggle with their CRM systems [41].
The interplay between technological, organizational, and individual factors also affects outcomes
of these systems [22, 24].
In a recent study, Goodhue et al. [19] examined challenges and opportunities of CRM in
several organizations. They suggest that the growth of CRM is driven by the changing demands
of the business for quality service, the availability of large amount of data, and the role of
information technology. They suggest that in order to benefit fully from CRM, firms may need to
undergo a major change in organizational culture and business practices. Organizational change
requires significant commitment and has high potential in terms of opportunities and challenges.
The authors recommend different levels of integration, transformation, and application for CRM
depending on the organizational needs and maturity.
A significant portion of customer dissatisfaction is due to employees’ inability or
unwillingness to respond to service failures [38]. In the financial services sector, for example,
more than 70 percent of customers defect because of dissatisfaction with service quality [8].
Organizations with service failure and recovery problems need to communicate commitment to
customers and strengthen bonds [4], CRM can help these organizations.
6
In the pharmaceutical industry, CRM has become a great investment and plays a
significant part in managing customer requests. Accurate data, effective processing, and cross-
functional integration are critical success factors in improving customer satisfaction. Product
information (e.g., specification, inventory, price, delivery method) are readily available to
customer representatives to facilitate immediate and accurate response to customers in their first
contact. Callbacks are minimized as they involve cost (in employee time) and the risk of losing
the customer. Processes are streamlined and quality control is imbedded within the system in
order to ensure continuous and consistent monitoring of customer service. Measures of service
quality have been developed and used in research studies of marketing [29], MIS [45], and call
centers [13].
Information technology plays an important role in CRM success [5, 26, 31]. Describing
the role of technology in service quality, Harvey et al. [21] point out the important gains that
come from producing and delivering more value to the customer. They suggest a model that
describes how services that provide ‘value-added partnerships’ can be created through
information technology. Improved service quality is perceived through close interaction and real-
time flow of information.
Integration of telephone communications, database, local area networks and other
information system applications have clearly enhanced the CRM function [1, 36]. Information
technology applications can be used to create customer ‘empowerment’ that will ultimately result
in customer satisfaction [36]. Organizational web sites are increasingly used to deliver services
as well as accumulate customer information. CRM professionals who are trained in information
technology and marketing [32] have been in great demand in recent years [47].
7
Despite the important role that information technology plays in CRM, academic IS
research has been lacking [40]. Benbasat and Zmud [3] encourage practice of relevance where
results of research studies can be used in practical decision-making processes. This study was
carried out at Merck-Medco, a large U.S. pharmaceutical corporation, in order to analyze CRM
issues and to formalize recommendations for remedial plans.
3. Background
Merck-Medco is a wholly owned subsidiary of Merck and Co. Inc. that specializes in
prescription benefit management. Merck focuses on discovering breakthrough medicines and
subsequently manufacturing, marketing and selling them. Merck-Medco dispenses medications
produced by over 60 manufacturers on behalf of its clients. Merck-Medco clients include major
employers, government agencies, health management organizations, and insurance providers.
The employees or members of these organizations are provided with a prescription drug benefit
that Merck-Medco manages on behalf of its clients. Currently, Merck-Medco represents over 60
million Americans through its various clients and fills over 500 million prescriptions per year.
This benefit is utilized in two ways. First, Merck-Medco negotiates contracts with retail
pharmacy chains to provide its clients prescription coverage at discounted costs. Members are
provided with a prescription drug card that they present to the retail pharmacist along with their
prescription. The retail pharmacist uses this card to access member-specific coverage
information through a proprietary Merck-Medco online system. This system provides the
pharmacist with information about the member’s drug history, co-pay and eligibility. After
filling the medication for the member, the pharmacist uses the same system to process a claim to
be reimbursed for the negotiated drug cost and administrative fee for dispensing the medication.
8
Members also have the option of using a home delivery pharmacy. Merck-Medco
operates 12 home delivery pharmacies across the country. These pharmacies are divided into
prescription processing centers and prescription dispensing centers. A member can submit a
prescription through five different means: mail, fax, phone voice response system, Internet or by
speaking with a customer service representative. Each client submits all of his or her
prescriptions to the same prescription processing center. Depending upon the type of medication,
the actual prescription could be dispensed through a different prescription dispensing center.
Once a prescription processing center receives a prescription, it is responsible for
entering all information necessary to prepare for dispensing. The process for accomplishing this
varies based on whether the order is for a new prescription or a refill of an existing prescription.
In the case of the refills, members are issued a bar-coded sticker that can be scanned to access all
of the necessary information from the original prescription. A new prescription is more complex.
All documents received through the mail are scanned into images that can be retrieved
later for any purpose. In the case of new prescriptions, once the images are available, a
pharmacist enters the information into the system on the right side of the screen while looking at
the images on the left side of the screen. The order is then subjected to a series of administrative
and professional edits. The administrative edits look for current eligibility, drug coverage
information, account balance and address or personal profile information. The professional edits
include drug utilization review that ensures they are not taking multiple drugs that interact with
each other or an allergy that would be impacted by the drug. If there were any questions about
what the doctor prescribed then a call would be made by a pharmacist.
There are also edits that target specific medications that have lower cost but
therapeutically equivalent alternatives. Pharmacists then call doctors to discuss switching the
9
medication to the alternative. This is primarily focused on saving the plan and customer money
while maintaining the same therapeutic outcomes. There are also some disease management edits
that concentrate on prescribing behaviors relative to specific disease states. These would also
prompt a call to a doctor to discuss the appropriateness of the medication prescribed by the
doctor.
Once all of these edits are identified and resolved, the prescription is then ready to be
dispensed (see figure 1). As mentioned earlier, the type of prescription dictates where it will
actually be dispensed from. There are two automated pharmacies that dispense primarily pills,
tablets and capsules. Therefore, a prescription that is capable of being dispensed through
automation would be electronically transferred to one of these two pharmacies. If the
prescription is a prepackaged item, refrigerated medication, a narcotic, a controlled substance, or
a compounded medication, it is dispensed in any one of seven pharmacies across the country.
The remaining three pharmacies are strictly prescription processing centers and do not dispense
any medications. The following table outlines the locations and functions of each pharmacy in
the United States:
Prescription processing centers only Spokane, WA, Irving, TX and Fairfield, OH Prescription dispensing centers only Willingboro, NJ Prescription processing and dispensing Las Vegas, NV, Columbus, OH, Pittsburgh, PA
Harrisburg, PA, Wilmington, MA, Parsippany, NJ and Tampa, FL
Automated dispensing pharmacies Las Vegas, NV and Willingboro, NJ
------------------------------------- Insert Figure 1 about here
--------------------------------------
Integrated throughout the prescription processing and dispensing process are extensive
customer service capabilities. All information relative to past prescriptions and prescriptions
currently in process is accessible by a customer service representative (CSR) via another
10
proprietary system. In addition, client plan design and personal profile information is also
available within this system. With 60 million customers to support, customer service handles
over 40 million calls per year through its network of six call centers. They are located in Tampa
(Florida), Parsippany (New Jersey), Columbus (Ohio), Dublin (Ohio), Irving (Texas) and Las
Vegas (Nevada).
A complex set of systems and applications are used to schedule CSRs and balance call
traffic to ensure the speed with which calls are answered meets the expectations of clients.
Frequently, clients have service penalties associated with the average speed of answer so this is
painstakingly managed to avoid any penalty. These six call centers have the ability to route
telephone calls transparently to a customer. Any question can be answered by any CSR in any
call center via the system. This allows customer service to balance call traffic based on staffing
and demand in real time.
Although the call centers and CSRs have integrated systems, they are not always capable
of resolving a member’s issue via the system. In these cases, there is a follow-up system to
address the member’s question or concern. A common example is when a member calls about
the status of their order that is currently in process. If the member needs the medication earlier
than when the system projects it will be dispensed, then a message is sent electronically to the
dispensing pharmacy. Within each dispensing pharmacy is a team of CSRs who are staffed
exclusively to follow up on member issues that the call center CSRs are unable to resolve. This
team would work these electronic messages called queues. Each queue is defined based on its
pharmacy location and the nature of the request. For instance, a member who had their
medication dispensed from Tampa, FL and needs to have it replaced because it was lost in the
mail would be queued to FLRPLC or Florida Replacement. There are 26 different queues that are
11
used by each dispensing pharmacy. If the specific nature of the request does not fit into one of
the first 25 queues then the CSR would use the 26th queue, which is a general queue. Figure 2
shows a simplified customer service process. The majority of customer and client complaints are
traced to the queuing process of Call Center CSR Enter Unresolved Issues into Queues (third
box) and Dispensing Pharmacy CSR Resolve Queued Issues (fourth box) in Figure 2.
------------------------------------- Insert Figure 2 about here
--------------------------------------
4. Research Methods Measures of service quality have been developed by researchers in marketing [29, 30],
information systems [45], and call centers [9]. These measures, however, are more general and
do not relate to specific issues such as CSR and queuing process that are important components
of the CRM system at Merck-Medco. Thus, we decided to start with a clean slate for identifying
and prioritizing problems that caused customer dissatisfaction and were primary concerns of the
management at Merck-Medco.
The process started by conducting focus groups in two of the call centers and four of the
pharmacies. Teams of six customer service representatives at each site were asked to brainstorm
a list of scenarios and issues that cause member dissatisfaction relative to the queuing process. A
team of managers, supervisors and CSRs in the Las Vegas site combined the lists and
consolidated them. They primarily sought to eliminate redundancy. Once the consolidated list
was complete, they sent the list out to all of the pharmacies and call centers asking them to add
any items that were not already addressed in the list. Therefore, every site had input. The
descriptions were then edited into the form of neutral problem statements.
12
The next step was to use a team of three CSRs from the call center and three CSRs from
the pharmacy in NV, since they were co-located, to categorize the list. They reviewed the entire
list and then brainstormed various categories. These categories were then narrowed to a list that
the team felt would capture all of the individual statements. The statements were then assigned to
each category through consensus discussion. Once every statement had been assigned a category,
one more consolidation step took place to eliminate categories with only a few statements. The
final list included 54 statements in five categories.
The list was then formatted as a survey where a five-point scale from strongly agrees to
strongly disagree was applied. A brief opening paragraph provided context for the survey
participants. A pilot survey of 25 CSRs was completed to assess the time to take the survey and
ensure there were no confusing statements. After completing the survey, the team of 25 was
asked for input on the clarity of the survey itself. The consensus was that the survey form was
fine and required no further modifications.
The survey was then converted to both an online version and hard copy. The online
version was administered to a sample of CSRs in the call centers via the system they had access
to. Because the CSRs in the pharmacies used a different system, they were unable to complete
the survey online. They therefore filled out the hard copy version. The current staff of 1500
CSRs in the call centers and the staff of 400 CSRs in the pharmacies completed a total of 1460
surveys – a 77% response rate.
5. Data Analysis
There are two objectives in this data analysis. The first is to identify salient factors that
affected CRM process at Merck-Medco. This would help the company develop an action plan for
improving the process. An exploratory factor analysis is an appropriate tool to identify these
13
salient factors. The second objective is to develop a reliable and valid instrument that can be
used to assess the effectiveness of the action plan. A confirmatory factor analysis is
recommended to reinforce confidence in the instrument [7].
In order to accomplish both objectives, data were randomly split into two equal parts. The
first half is used with exploratory factor analysis to determine the salient factors and produce a
set of items that measure CRM process failure at Merck-Medco. The second half is used with
confirmatory procedures to modify and finalize the factors and their measures produced based on
the first part of data.
Exploratory analysis First, as suggested by Churchill [11], the researchers purified the items (to eliminate
‘garbage items’). Two criteria were used to eliminate the items: corrected-item total correlation
(each item’s correlation with the sum of the other items in its category) and reliability. The
domain sampling model provides a rationale for corrected-item total correlation procedure. The
key assumption in the domain sampling model is that all items, if they belong to the domain of
the concept, have an equal amount of common core. If all the items in a measure are drawn from
the domain of a single construct, responses to those items should be highly inter-correlated. After
this, an exploratory factor analysis of the remaining items was conducted to identify items that
were not factorially pure. Items that loaded on more than one factor at 0.50 or above were
eliminated. This cut-off point is higher than what has been used by other researchers [15].
Exploratory factor analysis using principle components as extraction method with
Varimax rotation was used to determine the number of factors. If the factor analysis would result
in ambiguous structure or produce many items with multiple loadings, we reevaluated the
corrected-item total correlation results and carefully examined close-call items. To the extent
14
possible, we also considered item content to make sure all items within a factor measured similar
content.
This was possible because the researchers were familiar with the nature of the problem
through discussions with the staff at Merck-Medco. This was also necessary since we had
developed the initial list of items based on practice rather than theory. Thus, an iterative process
of using corrected item-total correlation and exploratory factor analysis was used to determine
the number of factors. Reliability was calculated at each stage to make sure it remains higher
than 0.80.
The process started with 13 initial factors and resulted in 7 factors with 21 items shown in
Table 1 and described in Table 2. Eigen values for the seven factors are greater than 1.0 and
range from 5.374 to 1.002. All factor loadings are above 0.64, much higher than the commonly
used threshold of 0.5. The seven factors accounted for 65% of variances. The seven factor
solution was easily interpreted and labeled as standard operating procedure compliance (5
items), accountability and ownership (4 items), callback information content (3 items), customer
contact process (3 items), billing issues (2 items), dispensing and replacement process (2 items)
and queuing procedure (2 items).
Managers and supervisors who closely collaborated with the researchers throughout the
study could easily relate to these factors. These factors would be useful for Merck-Medco to
make closer examination of their system and develop a useful remedial action plan.
------------------------------------- Insert Table 1 about here
-------------------------------------- -------------------------------------
Insert Table 2 about here --------------------------------------
Confirmatory analysis
15
The results of exploratory factor analysis were encouraging and provided a theoretical
basis to conduct confirmatory factor analysis. Figure 3 presents the confirmatory measurement
model to be tested. The confirmatory factor analysis follows the measurement property
assessment paradigm used in other studies [16, 37, 44] and was conducted using SIMPLIS in
LISERL 8.3. The results are presented in Table 3. Since some items have low factor loading and
the fit indices were less than satisfactory, further refinements were needed.
------------------------------------- Insert Figure 3. about here
-------------------------------------- -------------------------------------
Insert Table 3. about here --------------------------------------
Following Segars’ [37] procedure to refine and purify the measures, each factor with
more than 3 items is tested separately to improve model fit first. This is done by examining
modification indexes of each single factor measurement model and adding error correlations that
were suggested. One error correlation is added to factor 1 since reading of the item suggests they
could be correlated and no modification was necessary for factor two. These are the only two
factors with more than three items. The next step is to test measurement models with pair of
factors. In this process, items with cross loading were identified and eliminated. A total of 21
paired tests were conducted with only one item eliminated.
The last step is to combine all seven factors into a single measurement model and test the
model fit. This model included the reduced number of items and the error correlation identified
earlier. At this stage, items with factor loading less than .45 [34] or cross loadings were
eliminated one by one. Factors 5 and 7 ended up with only one item and therefore were removed
from the model. The process resulted in a five-factor 13-item model shown in Figure 4. Table 4
16
presents the standardized parameter estimates, fit indices, and final item descriptions for this
model. Compared to the initial model, the final model has much better model fit, as shown by the
fit indices. This procedure establishes convergent validity and unidimensionality.
------------------------------------- Insert Figure 4 about here
-------------------------------------- -------------------------------------
Insert Table 4 about here --------------------------------------
The next step in confirmatory analysis is to examine discriminant validity and construct
reliability. Discriminant validity can be established by comparing the model fit of an
unconstrained model that estimates the correlation between a pair of factors and a constrained
model that fixes the correlation between the factors to unity. Discriminant validity is
demonstrated when the unconstrained model has a significantly better fit than the constrained
model. The difference in model fit is evaluated by the chi-square difference between the models.
A significance of the chi-square difference is a chi-square variate with one degree of freedom.
Tests of all possible pairs for the five factors were conducted and the results are presented in
Table 5. All chi-square differences are significant at .001 level, this supports discriminant
validity. Construct reliability can be assessed using either a formula for composite reliability or
average variance extracted [20, 37]. Those numbers are also presented in Table 5. All factors
except dispensing and replacement process have acceptable composite reliability and average
variance extracted.
------------------------------------- Insert Table 5 about here
--------------------------------------
6. Discussion
17
Organizations continuously rethink their processes in order to improve them by extending
the boundaries of information technology application. Practitioners and academics alike have
long accepted the delivery of quality service to customers as a success factor. Information
technology has played an increasingly important role in the successful delivery of this service to
the customer. This study illustrates the complexity of customer relationship management
function and employee concerns regarding the processes involved. These issues are different
from what existing service quality measures and studies address. Service quality studies address
issues of customer (i.e., front end issues) while the current study addresses issues internal to the
system and processes (i.e., back end issues).
There are great expectations for what CRM can accomplish in terms of customer profile,
product information, rapid response, predicting customer needs, retaining customers, conducting
market research, promoting sale, and reducing cost. However, despite considerable
organizational and executive commitments these expectations have not always been
materialized. The perception of a widening gap between the potential of customer relationship
management (that is, what it can ideally achieve) and its actual accomplishments has increased
the need for better understanding of the nature of the problem and for better measures of factors
that influence outcomes. This research was designed to address the perceived gap that exists
between the potential of customer representative management and what it actually accomplished
in a large U.S. pharmaceutical company. Although this research is company specific, we believe
that our findings have relevance to other CRM environments (similar to [17, 27]).
Because of the nature of the problem situation, the approach taken in this study includes a
combination of qualitative and quantitative methods. Case study methodology was used to
determine the scope and boundaries of the problem. Quantitative methods, especially
18
multivariate analysis, were used to more specifically identify influential factors within the
determined scope. We feel that a combination of these two methodologies is quite appropriate
for situations where there is no widely accepted theory base for study questions or where existing
theories might not be appropriate for the practical problem at hand. In studying information
technology failure, for example, Lyytinen and Hirschheim [25] argue that any analysis is an
interpretive activity to understand the problem and to find solution. It is prudent to formulate
study questions based on experience and careful case study approach rather than borrowing a
well-established theory from other disciplines that poorly fits the problem.
The company, Merck-Medco, greatly depends on CRM to interact with their customers.
Doctors and patients in the company’s member plans send their prescriptions by mail, fax,
phone, or through website. At automated pharmacies like the ones in Las Vegas, Nevada or
Willingboro, New Jersey, 99% of the prescriptions are filled within 24 hours of receipt. Some of
these pharmacies may fill more than 800,000 prescriptions per week [18]. Before a prescription
is filled, potential drug interactions are automatically checked and issues are flagged and
forwarded to a registered pharmacist for investigation and resolution. There are numerous other
issues that customer service representatives use the system to respond to. For example, a member
patient may request drugs to be sent to a different address while on holiday or may request a
speedy delivery before travel. These messages are received at call centers and queued for action
at pharmacies or packaging centers responsible for that particular package.
For some time, the company has been experiencing problems with their CRM that result
in complaints from customer representatives. This led to a close collaboration between the
company and the researchers and influenced the design, implementation, and data collection for
this study. The study involved broad participation that included senior executives, managers,
19
supervisors, and customer representatives. Customer representatives throughout the company
showed interest in addressing CRM issues and streamlining customer query and response
system. Participants from call centers throughout the organization were involved in different
stages of they study including design and data gathering. More than 75% of system users
responded to survey questions.
The results of this study point to numerous processes that are either not followed or
clearly understood by customer service representatives. Through exploratory factor analysis, this
study specifically identifies issues that relate to standard operating procedure, accountability and
ownership, call back information content, customer contact process, billing issues, and
dispensing and replacement process. Collectively these issues have resulted in CRM failure in
the company. Two factors (billing issues and queuing procedure) were dropped during the
confirmatory factor analysis phase in order to improve measurement rigor. Below we describe
each factor that has been identified as a barrier to CRM success in the company and will briefly
address possible remedies. Although the shorter five-factor model is easier to use and supported
by confirmatory analysis, the discussion will include the larger list of actors identified in the
exploratory phase of the study to help the management of CRM at Merck-Medco.
Standard operating procedure – Data analysis suggests that compliance with standard operating
procedure explains current complaints more than any other factor. It includes wrongly sequenced
queues, closing of incomplete files, forwarding incomplete electronic forms, and the like. This
issue relates to employee behavior and may be addressed through training as well as improved
description of procedures. However, training alone or improved description of procedure may
not be adequate if non-compliance is widely spread and has become routine or if the system’s
information product does not readily support compliance with standard operating procedure. The
20
current CRM system can be revamped, improved, or redesigned so that it will enforce much of
standard operating procedure and make compliance a part of interaction with the system. Joshi
and Rai [22] studied the influence of information product on work and suggest greater attention
to the need for designing quality systems that not only meet primary information delivery
objectives, but also take into account the task and organizational design issues for the user.
Accountability and ownership – The issue of accountability is another readily identified factor
that has created discontent among CRM users and customer service representatives. Under the
current system, it is difficult to determine who is accountable for events such as filing customer
contact forms without action, violating queue sequence for customer contact form, keeping
commitments made to customers, and not verifying address prior to refill and shipping. This has
created role ambiguity and role conflict with adverse effect on information product intended
outcome [22]. The interaction between task needs and technology application has created a
problem that cannot be addressed through defining responsibilities alone. Levels of responsibility
and accountability need to be established and communicated and violations are readily identified.
Information technology is expected to provide appropriate control to management for work
process and quality performance [42]. Information content must be extended to include control
for accountability.
Call back information content – The third factor influencing CRM outcome at the company
relates to the quality of information generated through the use of customer contact form. This
information is unclear or even wrong and is used to generate customer queries and determine
queues. This in turn influences queue procedure that is identified as another factor affecting
customer representatives. Information technology is expected to empower the individual
employee to provide accurate and timely response to customers. Information content plays a
21
critical role in employee satisfaction with information system [2, 10, 14]. Lack of employee
confidence in information content has an adverse influence on system use [43]. The system
needs to be revamped to more effectively and more easily control for information quality at data
entry, data integration, and data manipulation level.
Customer contact process – Information technology and CRM are expected to provide the
mechanism through which long-term, individualized relationships with customers can be created
and maintained. Massey, et al. [26] suggest that CRM centers on gaining a steady or increasing
business from current customers, not necessarily a constant stream of new customers. To
accomplish this, the company must ask what makes a specific customer unique and then tailor
services in response to that uniqueness [28]. The current system fails to create an environment
that helps sustain a steady business from customers. The customer contact form was designed to
collect information on a single issue rather than multiple ones. As currently practiced, this form
generates multiple issues and that in turn complicates routing of issues to appropriate pharmacies
or packaging centers. The system should facilitate generation of separate forms for multiple
issues and avoid providing the option of multiple issues on a single form.
Billing questions – This factor illustrates ineffective integration of accounting function (e.g.,
accounts receivable) with CRM in the company. Customer service representatives are not
familiar with how to respond to customer queries for billing; the system does not help them
explain expenses to customers. Customer service representatives have difficulty interpreting
accounting codes on the billing screen. The system needs to more fully integrate accounting
function and help customer representatives respond accurately to billing questions. Lack of
proper integration is suggested as one of the causes of failure in CRM [41].
22
Dispensing and replacement process – Since a number of pharmacies and packaging centers are
involved in dispensing medicines to customers, it is important to link back recurring customer
requests or changes to specific center or pharmacy designated to fill that prescription. Customer
service representatives often have difficulty queuing a replacement request. The system does not
automatically identify where a replacement request should be queued. The dispensing and
replacement process is unclear to many customer representatives and the existence of ‘front end’
and ‘back end’ pharmacies makes it more difficult to know where to queue issues. The process
needs to be streamlined and the system should be redesigned to assist customer representatives in
managing dispensing as well as replacement requests. These issues may not have existed at the
time the system was developed. Rapid growth and packaging and dispensing automation have
further complicated the process.
Queuing procedure – There seems to be a significant confusion over the queue system and how
and where customer contact form should be queued. In the current system, queuing occurs
because a customer service representative in a call center cannot systemically resolve the
customers concerns online. There is a need for an environment that integrates all systems at the
company and that will enable customer service representatives in call centers to resolve the vast
majority of member needs without the need to queue anything. The system needs to be
redesigned to eliminate the need for customer service representatives to memorize procedures for
each pharmacy and generate standard form that is easily understood by all representatives and
pharmacy people alike.
These issues can be viewed under two broad categories of ‘people’ and ‘system’. This
grouping facilitates generalization of the issues and determination of remedial actions. Table 6
summarizes our analysis of these issues relative to ‘people’ or ‘technology’. This breakdown is
23
similar to Bostrom’s [6] argument that each work system is made up of two interacting
subsystems: the technical and the social. The technical subsystem relates to processes, tasks and
technologies while social subsystem involves people attributes such as skills and attitudes as well
as organizational attributes such as reward systems and authority structure. This model provides
a useful framework that helps identify the interaction between two sets of influential variables.
The interaction between system and people issues is evident in majority of the factors identified
as causing dissatisfaction with CRM in this company. Further, these factors interact among
themselves and sometime exasperate the situation.
------------------------------------- Insert Table 6 about here
--------------------------------------
Finally, it is important to realize that while some of these issues are linked to employee
and user behavior, many of them are inherent attributes of the system and existing processes.
Over the course of the last few years, extensive training efforts have been aimed at addressing
these issues at Merck-Medco with minimal effectiveness, suggesting that systems related issues
are as problematic. Given the number and extent of issues and concerns and the potential impact
that they have on the business, a complete redesign of the systems that customer service
representatives use to access customer and prescription data may be necessary. In any redesign
of the system, one of the key objectives must be to eliminate or minimize the number of issues
that need to be queued. To accomplish this, the multiple systems that are not currently integrated
need to be able to directly communicate with each other. An integrated system will enable a
customer service representative in a call center to resolve member needs online without the need
to queue. Results of this study further suggest a need to streamline processes used to service
customers.
24
7. Conclusions
This study identifies factors that affect the success of customer relationship management
at a US pharmaceutical company, Merck-Medco. Through a close collaboration with the
company, we generated a comprehensive list of potential issues, developed a survey, collected
company wide data, and used multivariate methods to determine salient factors that caused
problem in their CRM. Results of the exploratory factor analysis suggest seven factors (standard
operating procedure compliance, accountability and ownership, callback information content,
customer contact process, billing issues, dispensing and replacement process and queuing
procedure) that describe CRM issues at Merck-Medco. A confirmatory analysis produced a five-
factor model measurement instrument. These measures are useful in assessing the results of
remedial plans. These factors are summarized and discussed in organizational and technical
contexts. More research is needed to better understand the interaction between the technical and
organizational dimensions of technology application. This study used customer representatives
as respondents. Future research should consider the use of actual customers for collecting data.
This study also benefited from management involvement for data collection. Data collection
using actual customers will eliminate that intervention and its possible influence.
25
References
[1] O.Z. Aksin and P.T. Harker, Modeling a phone center: Analysis of a multichannel, multiresource processor shared loss system, Management Science 47, 2001, pp. 324-336.
[2] J.E. Bailey and S.W. Pearson, Development of a tool for measuring and analyzing computer user satisfaction, Management Science 29, 1983, pp. 530-545.
[3] I. Benbasat and R.W. Zmud, Empirical research in information systems: The practice of relevance, MIS Quarterly 23, 1999, pp. 3-16.
[4] L.L. Berry and A. Parasuraman, Marketing services: Competing through quality Free Press, New York, 1991.
[5] M. Bitner, S.W. Brown and M.L. Meuter, Technology infusion in service encounters, Journal of the Academy of Marketing Science 28, 2000, pp. 138-150.
[6] R.P. Bostrom, Conflict handling and power in the redesign process: A field study investigation of the relationship between management information system users and systems maintenance personnel, University of Minnesota, 1978.
[7] M.-C. Boudreau, D. Gefen and D.W. Straub, Validation in Information Systems Research: A State-of-the-Art Assessment, MIS Quarterly 25, 2001, pp. 1-16.
[8] J.W. Bowen and R.B. Hedges, Increasing service quality in retail banking, Journal of Retail Banking 15, 1993, pp. 21-28.
[9] A. Burgers, K. de Ruyter, C. Keen and S. Streukens, Customer expectation dimensions of voice-to-voice service encounters: a scale-- development study, International Journal of Service Industry Management 11, 2000, pp. 142-161.
[10] M.M. Caldeira and J.M. Ward, Understanding the successful adoption and use of IS/IT in SMEs: An explanation from Portuguese manufacturing industries, Information Systems Journal 12, 2002, pp. 121-152.
[11] G.A. Churchill, A paradigm for developing better measures of marketing constructs, Journal of Marketing Research 16, 1979, pp. 64-73.
[12] J. Cross, M.J. Earl and J.L. Sampler, Transformation of the IT function at British Petroleum, MIS Quarterly 21, 1997, pp. 401-423.
[13] A.M. Dean, Service quality in call centres: Implications for customer loyalty, Managing Service Quality 12, 2002, pp. 414-423.
[14] W.J. Doll and G. Torkzadeh, The measurement of end-user computing satisfaction, MIS Quarterly 12, 1988, pp. 259-274.
[15] W.J. Doll and G. Torkzadeh, Developing a multidimensional measure of system-use in an organizational context, Information & Management 33, 1998, pp. 171-185.
[16] W.J. Doll, W. Xia and G. Torkzadeh, A confirmatory factor analysis of the end-user computing satisfaction instrument, MIS Quarterly 18, 1994, pp. 453-461.
[17] P.A. Essex, S.R. Magal and d.E. Masteller, Determinants of information center success, Journal of Management Information Systems 15, 1998, pp. 95-117.
26
[18] H. Forcinio, What does pharmacy automation mean for packaging?, Pharmaceutical Technology , 2002, pp. 22-26.
[19] D.L. Goodhue, B.H. Wixom and H.J. Watson, Realizing business benefits through CRM: Hitting the right target in the right way, MIS Quarterly Executive 1, 2002, pp. 79-94.
[20] J.F.J. Hair, R.E. Anderson, R.L. Tatham and W.C. Black, Multivariate Data Analysis Prentice Hall, Upper Saddle River, NJ, 1998.
[21] J. Harvey, E. Lefebvre and L. Lefebvre, Technology and the creation of value in services: A conceptual model, Technovation 13, 1993, pp. 481-495.
[22] K. Joshi and A. Rai, Impact of the quality of information products on information system users' job satisfaction: An empirical investigation, Information Systems Journal 10, 2000, pp. 323-345.
[23] P.K. Kannan, Introduction to the special issue: Decision support issues in customer relationship management and interactive marketing for e-commerce, Decision Support Systems 32, 2001, pp. 83-84.
[24] H.C. Lucas and J. Baroudi, The role of information technology in organization design, Journal of Management Information Systems 10, 1994, pp. 9-23.
[25] K. Lyytinen and R. Hirschheim, Information systems failures - A survey and classification of empirical literature, Oxford, UK, (4), 1987, pp. 257-309.
[26] A.P. Massey, M.M. Montoya-Weiss and K. Holcom, Re-engineering the customer relationship: Leveraging knowledge assets at IBM, Decision Support Systems 32, 2001, pp. 155-170.
[27] L.R. Matheson and R.E. Tarjan, Culturally induced information impactedness, Journal of Management Information Systems 15, 1998,
[28] R.M. Morgan and S.D. Hunt, The commitment-trust theory of relationship marketing, Journal of Marketing 58, 1994, pp. 20-38.
[29] A. Parasuraman, V.A. Zeithaml and L.L. Berry, SERVQUAL: A multiple-item scale for measuring consumer perceptions of service quality, Journal of Retailing 64, 1988, pp. 12-40.
[30] A. Parasuraman, V.A. Zeithaml and L.L. Berry, Alternative scales for measuring service quality: A comparative assessment based on psychometric and diagnostic criteria, Journal of Retailing 70, 1994, pp. 210-230.
[31] J. Peppard, Customer relationship management (CRM) in financial services, European Management Journal 18, 2000, pp. 312-327.
[32] M. Pontes and C.O. Kelly, The identification of inbound call center agents' competencies that are related to callers' repurchase intentions, Journal of Interactive Marketing 14, 2000, pp. 41-49.
[33] B. Price, Making CRM come to life, E-Business Review , 2002, pp. 25-31.
[34] B. Raghunathan, T.S. Raghunathan and Q. Tu, Dimensionality of the strategic grid framework: The construct and its measurement, Information Systems Research 10, 1999, pp. 343-355.
27
[35] D.K. Rigby, F.F. Reichheld and P. Schefter, Avoid the four perils of CRM, Harvard Business Review 80, 2002, pp. 101-109.
[36] L.A. Schlesinger and L. Haskett, The service driven service company, Harvard Business Review , 1991, pp. 71-81.
[37] A.H. Segars, Assessing the unidimensionality of measurement: A paradigm and illustration within the context of information systems research, Omega 25, 1997, pp. 107-121.
[38] A.K. Smith and R.N. Bolton, An experimental investigation of customer reactions to service failure and recovery encounters, Journal of Service Research 1, 1999, pp. 65-81.
[39] J. Stuller, Making call-center voices smile: A business case for better training, Training , 1999, pp. 26-32.
[40] R. Subramanyam and M.S. Krishnan, Business value of IT-enabled call centers: An empirical analysis, Proceedings of the Twenty-Second International Conference on Information Systems, New Orleans, LA, 2001, pp. 55-64.
[41] R.S. Swift, Excutive response: CRM is changing our eras, the information we require, and our processes, MIS Quarterly Executive 1, 2002, pp. 95-96.
[42] G. Torkzadeh and W.J. Doll, The Development of a tool for measuring the perceived impact of information technology on work, Omega 27, 1999, pp. 327-339.
[43] G. Torkzadeh and D.J. Dwyer, A path analytic study of determinants of information system usage, Omega 22, 1994, pp. 339-348.
[44] G. Torkzadeh, X. Koufteros and K. Pflughoeft, Confirmatory factor analysis of a computer self-efficacy instrument, Structural Equation Modeling Journal 8, 2001,
[45] T.P. Van Dyke, L.A. Kappelman and V.R. Prybutok, Measuring information systems service quality: Concerns on the use of the SERVQUAL questionnaire, MIS Quarterly 21, 1997, pp. 195-208.
[46] N. Woodcock and M. Starkey, I wouldn't start from here: Finding a way to CRM projects, Journal of Database Marketing 9, 2001, pp. 61-74.
[47] L. Yu, Successful customer-relationship management, Sloan Management Review , 2001, pp. 18-19.
28
Table 1. Factor structure
Items Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6 Factor 7
F1-1 .702 F1-2 .772 F1-3 .770 F1-4 .812 F1-5 .649 F2-1 .754 F2-2 .814 F2-3 .818 F2-4 .707 F3-1 .647 F3-2 .796 F3-3 .722 F4-1 .695 F4-2 .779 F4-3 .681 F5-1 .796 F5-2 .815 F6-1 .791 F6-2 .807 F7-1 .757 F7-2 .681 Variance explained 15.22% 12.53% 8.51% 8.18% 7.14% 6.89% 6.47% (Total) 64.95% Eigenvalues 5.374 1.839 1.753 1.442 1.194 1.035 1.002 Factor correlations Factor 2 .423** Factor 3 .429** .292** Factor 4 .343** .258** .244* Factor 5 .110** .176** .159** .254** Factor 6 .088* .132** .211** .096** .206** Factor 7 .386** .158** .249** .218** .172** .142**
29
Table 2. Factors and items descriptions
Factor 1: Standard operating procedure compliance F1-1 CCFs that are queued are not worked in a timely manner. F1-2 CCFs are closed without documented comments. F1-3 CCFs are returned via the FEED process with no resolution. F1-4 CCFs are closed with no action taken. F1-5 CCFs are closed with no resolution documented.
Factor 2: Accountability and ownership F2-1 No one is accountable when a CCF is closed without any action. F2-2 There is no accountability in the queuing process. F2-3 No one is accountable when a CCF is mis-queued. F2-4 No one has ownership of a member's issue.
Factor 3: Callback information content F3-1 The system auto-populates wrong information on the CCF. F3-2 Cancel callback queues contain wrong reason codes for the cancellations. F3-3 Cancel callback queues contain unclear reason codes for the cancellations.
Factor 4: Customer contact process F4-1 CCFs are created with incomplete or unclear information. F4-2 CCFs are routed to the wrong pharmacy. F4-3 CCFs are created which contain multiple issues.
Factor 5: Billing issues F5-1 It is difficult to know how to handle members calls related to billing. F5-2 Member calls for billing explanations should be routed to A/R.
Factor 6: Dispensing and replacement process F6-1 The pharmacy dispensing process is unclear to me. F6-2 The system is unclear as to where I should queue a replacement request.
Factor 7: Queuing procedure F7-1 It is difficult to remember the control prescription procedures for each pharmacy. F7-2 CCFs need a standardized format for ease of queuing.
Acronyms A/R: Accounts receivable CCF: Customer contact form FEED: electronic means to pass information between departments/sites
30
Table 3. Completely standardized parameter estimates and t-values
Latent Variable Item Factor Loading (λ) Standard Error t-Value R-Square F1-1 .56 ---& ---& .32 F1-2 .71 .089 13.85 .50
F1 F1-3 .70 .088 13.75 .49 F1-4 .79 .091 14.70 .62 F1-5 .74 .084 14.25 .55 F2-1 .76 ---& ---& .58
F2 F2-2 .83 .048 20.82 .69 F2-3 .76 .049 19.52 .58 F2-4 .61 .047 15.71 .38 F3-1 .44 ---& ---& .20
F3 F3-2 .75 .14 10.06 .56 F3-3 .75 .16 10.07 .57 F4-1 .67 ---& ---& .45
F4 F4-2 .66 .089 10.93 .44 F4-3 .40 .068 8.24 .16
F5 F5-1 .63 ---& ---& .40 F5-2 .58 .16 6.46 .34
F6 F6-1 .53 ---& ---& .29 F6-2 .53 .18 5.53 .29
F7 F7-2 .61 ---& ---& .37 F7-2 .41 .073 7.22 .17
& Indicates a parameter fixed at 1.0 in the original solution. Fit Indices: χ2=406.33, df=168, p=0.00000, χ2/df=2.42, RMSEA=0.044, p(RMSEA<.05)=0.96,
NFI=0.91, NNFI= 0.93, CFI=0.94, RMR=0.066, GFI=0.95, AGFI=0.93
31
Table 4. Refined parameter estimates, t-values, fit indices, and item descriptions
Latent Variable Item Factor Loading (λ) Standard Error t-Value R-Square F1-1 .53 ---& ---& .28
F1 F1-2 .67 .092 13.53 .44 F1-3 .71 .11 12.39 .51 F1-4 .83 .12 12.84 .69 F2-1 .73 ---& ---& .54
F2 F2-2 .85 .063 16.67 .72 F2-4 .61 .052 14.64 .37
F3 F3-2 .74 .14 10.06 .55 F3-3 .77 .11 11.12 .59
F4 F4-1 .69 ---& ---& .47 F4-2 .64 .11 8.72 .41
F6 F6-1 .45 ---& ---& .21 F6-2 .63 .37 3.66 .39
Fit Indices: χ2=81.00, df=54, p=0.01014, χ2/df=1.5, RMSEA=0.026, p(RMSEA<.05)=1.00, NFI=0.97, NNFI= 0.98, CFI=0.99, RMR=0.042, GFI=0.98, AGFI=0.97
Table 5. Descriptive statistics, correlations, discriminant validity tests, composite reliability, and average variance extracted
Factors Means S.D. F1 F2 F3 F4 F6
F1 3.70 (4 items) 0.97 0.78
[0.48]
F2 3.14 (3 items) 1.16 0.35**
(75.13$) 0.78
[0.54]
F3 3.46 (2 items) 0.93 0.36**
(119.21$) 0.26**
(88.03$) 0.88
[0.55]
F4 3.55 (2 items) 1.00 0.31**
(99.47$) 0.28** (56.8$)
0.29** (98.41$)
0.61 [0.44]
F6 2.49 (2 items) 1.05 0.16**
(142.52$) 0.16** (77.8$)
0.16** (111.92$)
0.14** (92.7$)
0.45 [0.30]
** Correlation is significant at 0.01 $ χ2 differences are indicated in parentheses. Differences in χ2 for 1 degree of freedom are significant at 0.001. a Composite reliabilities are on the diagonal. b Average variance extracted are on the diagonal in brackets.
32
Table 6: Summary analysis of barriers to CRM success at Merck-Medco
Organizational/People Technological/System
Factor Cause Solution Cause Solution
Compliance with standard operating procedure [22]
Lack of understanding, Lack of enforcement
System usage, training, policy & procedure
Lack of system checks
Incorporate automatic SOP enforcement
Accountability and ownership [22, 42]
Lack of policy & procedure for tracking responsibilities
Enacting policy & procedure
Lack of system tracking or monitoring
Incorporate system tracking or monitoring
Callback information content [2, 10, 14]
Lack of understanding with proper system usage
System usage training
System error, lack of assistance from system
Correct system error, improve system help features
Customer contact process [26, 28]
Lack of understanding with proper system usage
System usage training
Lack of system checks
Incorporate automated system checking
Billing issues [41] Lack of understanding with internal processes in other functional areas
Cross-functional training
Lack of system integration from other areas
Improve system integration
Dispensing and replacement process
Lack of understanding with internal processes in other areas & system usage
Cross-functional training, system use training
Lack of assistance from system
Incorporate system help features
Queuing procedure Lack of understanding with proper system usage & standards
System usage training
Lack of assistance from system, lack of standards
Incorporate system help features & standards
33
Prescription
New?
Collect Information
through Barcode
Scan Prescription
Enter Prescription
Question?
Contact Doctor
Question Resolved
Dispense Prescription
Administrative Edit
Professional Edit
No
YesNo
Yes
Figure 1. Prescription Processing
CustomerCall in Issues
Call Center CSR Resolve
Customer Issues
Call Center CSR Enter Unresolved Issues into Queues
Dispensing Pharmacy
CSR Resolve Queued Issues
Figure 2. Customer Service
34
F1 F2 F3 F4 F5
F1-1 F1-2 F1-3 F1-4 F3-2 F3-3 F4-1 F4-2 F5-1 F5-2 F1-5 F2-1 F2-2 F2-4 F2-3 F3-1 F4-3 F6-1 F6-2 F7-1 F7-2
F6 F7
Figure 3. Path diagram of the initial confirmatory measurement model
Chi-Square=81.00, df=54, P-value=0.01014, RMSEA=0.026
F1 F2 F3 F4 F6
F1-1
0.72
F1-2
0.56
F1-3
0.49
F1-4
0.31 0.13
0.53 0.67 0.71 0.83
F2-1
0.46
F2-2
0.28
F2-4
0.63
0.73 0.85 0.61
F3-2
0.45
F3-3
0.41
0.74 0.77
F4-1
0.53
F4-2
0.59
0.69 0.64
F6-1
0.79
F6-2
0.61
0.45 0.63
0.44 0.49
0.46 0.24 0.27
0.26 0.40
0.33 0.43
0.25
Figure 4. Path diagram of the final measurement model