Published in: Journal of Cost Management, Spring 1993, pp. 48-58.

Activity-Based Total Quality Management
at American Express

David A. Carlson
Assistant Professor of Information Systems
Graduate School of Business
University of Colorado at Boulder

S. Mark Young
Associate Professor of Accounting
School of Accountancy
University of Southern California
Los Angeles, CA 90089-1421

The commitment to a Total Quality Management philosophy is driven by the highest office at American Express. In a recent speech, James D. Robinson III, Chairman, CEO and Chief Quality Officer for the Company, stated that quality must be integrated with business strategy. He said, ``Business unit heads must present their strategy for quality improvements as part of their annual budget and business plans.''

Beginning in 1992, Robinson required each American Express business unit to undergo a self-assessment based on the criteria specified in the guidelines for the Malcolm Baldrige National Quality Award. He also wants to ensure that the investments in training and technology are paying off. Will the changes being made within American Express business units improve the bottom line? The activity-based costing method used at American Express Integrated Payment Systems (IPS) in Denver is able to address this concern.

Early in 1990, IPS instituted a TQM philosophy that would involve each of its 1,000 employees in the Denver metro area. Charlie Fote, IPS President, hired Brian Higgins as Director of Quality Assurance to lead this initiative.

Fote and other IPS executives were not aware of any particularly troublesome, quality-related problems. In fact, the company was, and still is, a service leader in the markets it serves. The group pursued a quality philosophy to expand its competitive advantages by further improving service, and increasing income without increasing costs. The definition of continuous improvement in any service-based organization is, however, a difficult task. Consequently, the Quality Assurance staff sought a means for defining and measuring quality in the organization. They set out to answer several questions:

• In the spirit of TQM, how do IPS' customers and suppliers feel about the reliability of activities performed by the organization, or about the contribution those activities make toward their requirements?

• What do IPS' employees do that contributes value to the company's services, or advances its mission?

• Conversely, what do the employees do that does not contribute value?

The TQM initiative began by building an infrastructure that involved all employees in continuous improvement efforts. Processes were established for assessing customer, vendor, and employee satisfaction on a regular basis, and studies were begun to develop a better understanding of organizational activities and performance measures. Activity-based studies were begun in the customer service department and telephone operations and quickly became interwoven with concerns about product costing. In particular, senior management wanted a better understanding of the fixed versus variable costs for each product line so that they could better manage growth at IPS. Product costing was related to the quality strategy in that both objectives sought answers to the difficult questions about where information workers spent their time.

In this article, we illustrate how activity-based costing contributes to the achievement of each of these concerns. As will be described later, activity costs are determined by asking each manager to estimate the time that his or her department spends on each activity, and by splitting those activity estimates across each product line. In order to achieve the objectives of TQM, these activity costs are augmented with perceptual data that various stakeholders have expressed about these activities. In the end, we suggest how similar approaches to total quality management can benefit other service organizations, or the service functions within manufacturing companies.

The Restructuring Imperative

In a recent Harvard Business Review article Stephen S. Roach stated (p. 91), ``Services need an accounting framework that can identify which activities add the most value... Activity-based managerial accounting is a step in the right direction, but much more work in this area remains to be done... It should go without saying that a metric for quality is equally important. Admittedly, quality in the service sector is hard to define.''

As suggested by Roach, white-collar productivity improvement is not being embraced by business leaders. This is probably due to the nature of knowledge work; poor productivity is very difficult, if not impossible, to remedy. Knowledge work is considered by many to be non-structured, self-directed, and intangible; it is often difficult to relate activity costs to the value provided. Traditional approaches, proven effective on the manufacturing floor, have not migrated into offices because they simply do not work. White-collar productivity improvement is not beyond hope, it's simply that many do not understand, nor have the right tools, to adequately address the difficulties associated with knowledge worker productivity. IPS appears to have, at least partially, solved this dilemma.

Activity-Based TQM

Activity-based costing (ABC) is central to recent efforts to redesign cost accounting systems in order to account for a wide variety of changes that are occurring in high technology manufacturing and service firms. These changes require accountants, and all employees in the firm, to alter their mindsets from cost accounting to ``cost management.'' Cost management emphasizes a proactive approach to planning and managing an organization's costs, whereas cost accounting usually focuses on the historical reporting of costs. The expanded view proffered by cost management integrates consideration of corporate strategy, leading to the notion of ``strategic cost management.'' The objectives of strategic cost management (SCM) appear to be aligned with the merger of qualitative and quantitative information in the Functional Administrative Control Technique (FACT) approach used at IPS. They have explicitly tied the results of its FACT studies to achievement of its Total Quality Management (TQM) strategy, as well as to its product strategies. (See the exhibit ``Activity-Based Total Quality Management.'')

The FACT approach has its roots in value engineering rather than in accounting. Value engineering is a technique for increasing the value of a product or organization rather than for just decreasing its cost. A product-oriented value engineering study would first create a functional description of a physical product, and then map the product's parts onto the functions that those parts performed. Thus, a costed functional description is produced. Customers' requirements are balanced with the function costs in assessing the value. During the 1960's, General Electric refined some of the value engineering disciplines developed in the 1940's which had proved so successful in reducing material costs. This refinement, referred to as Cross-Functional Analysis, was designed to examine cross-functional effort applied to organizational activities. Using cross-functional analysis, a business model is constructed which contains the activities performed on an entity-wide basis. Typically 150 to 200 such activities are defined and grouped into major business processes. Effort and cost are then recast into this business model. The contribution of each functional organization is examined to identify specific activities that are unique or that contain duplication, redundancy, or overlap. An organization's resources are then redeployed -- added to some activities and reduced in others -- in order to increase the value of the business to employees and to customers.

Brian Higgins was working as a value engineer when he began designing the FACT analysis process in 1972; his efforts began without knowledge of GE's cross-functional analysis. The FACT process uses a five-phase project plan for analyzing and improving organizational performance. (See ``FACT Project Plan'' for an outline of these phases.)

The duration of a complete FACT project varies with the scope and complexity of the organization. In most instances a project may be completed in twelve to sixteen weeks. The Project Planning Phase is guided by the FACT project leader who carefully selects a small team of individuals to conduct the study. Depending on the project scope, anywhere from two to ten people may be required, either part or full time. The project leader typically introduces the objectives and plan for the FACT study in a meeting of all managers involved.

In the Data Collection Phase, FACT project team members gather information through standardized data collection forms, and they conduct interviews with three questions in mind:

1. What does the organization do?

2. What does it cost to do what it does?

3. What is the acceptance or worth of what it does?

The interviews are further structured by posing questions that provoke managers, customers and vendors to identify problem areas and potential opportunities for change. The interviewees are told to imagine what their organization would look like if everything were working perfectly and to answer the questions with that scenario in mind. For each primary activity impacting the interviewee, three questions are asked:

1. If the organization was working in the most efficient and effective manner, how would they be organized and how would the activity be performed?

2. What inhibits the organization from working in the most efficient and effective manner?

3. How else can the activity be provided?

The last three phases of the project plan are crucial for building consensus among the team members and management when identifying problems and possible solutions. Specialized corrective action teams (CAT teams) are established at the start of the data analysis phase. These teams champion organizational changes through collaborative participation of all affected individuals, departments, customers, and suppliers. A team retains control and responsibility of the project until final recommendations are presented to management, and often until final implementation is complete.

Higgins believes that the main lesson everyone has to learn is that crucial assumptions are made when, as in many traditional accounting systems, you have a large fraction of your product cost allocated from pools of overhead expense. At IPS, half of their business unit's expenses were allocated overhead, but the accountants admitted that they were comfortable with only 15% of the allocations of that overhead to products. For example, data processing expenses were allocated to products based on transaction volume. Since IPS processes hundreds of thousands of Money Order transactions per day, that product got the lion's share of data processing allocations even though everyone admitted that MoneyGram's on-line systems were more CPU-intensive per transaction. Higgins concludes, ``You have to question the assumptions people made when their decisions are based on allocating huge overhead pools to products. Such assumptions are often wrong!'' By directly allocating expenses to activities and to products, a more accurate picture can be obtained.

Gathering Information

The foundation of the FACT model lies in the concept of an activity -- a response to the question, what does the organization do? An activity is stated as a brief verb-noun description of an organizational process. For example, ``Resolve customer problems'' or ``Coordinate corrective action.'' Unlike GE's cross-functional analysis, and unlike many accounting applications of activity-based costing, FACT derives significant benefits from using a hierarchical structure of organizational activities. A response to the question ``How is an activity performed?'' leads toward the more specific activities in the hierarchy (from left to right in the figure), and a response to the query ``Why is an activity performed?'' leads from the specific to the general activities (from right to left). This structure permits both macro and micro level analyses of the organization. In all, about 200-250 such activities are usually defined. The level of detail expressed by these activities depends on the purpose for which the FACT model will be used.

The FACT process has been applied to individual work groups as small as five people or to entire organizations encompassing 1000 people with salary and overhead expenses exceeding $60,000,000. In small departments, the activities would address such processes as ``Document customer problem'' or ``Reconcile invoices,'' whereas in large organizations the activities are conceptualized at a higher level: ``Research market,'' ``Establish pricing,'' or ``Identify inventory variation.'' The associated figure (see ``Activity Hierarchy'') shows the top levels of an activity hierarchy for the IPS business unit.

A FACT study begins by defining the costed hierarchy of activities, but it doesn't end there. This activity structure provides the basis for assembling a collection of information that will later be used in creating and evaluating solutions to identified problems and in linking the activities to business and quality strategies. Managers follow an incremental process for constructing an organizational model that begins with the definition of a costed activity set. Each activity's cost is determined by assigning a fraction of each person's time (and therefore salary) or a fraction of a departmental expense item to a particular activity. Thus, all of the organization's costs are allocated to the activity hierarchy. The fractions are determined by a variety of methods (e.g., time sampling or interviews with managers), depending on the scope of the organization being studied. Once the costs are allocated, additional information is collected and related to the model.

The principal non-cost information added at this stage represents attitudes of customers and suppliers---those who receive the outputs of the organization and those who supply its inputs. This attitudinal information includes both quantified variables and statements of perceptions. Quantified data consists of rating each activity for its contribution and reliability on ten-point scales. Perceptual data includes verbal comments which provide either favorable or unfavorable opinions related to one or more activities. These data are crucial for satisfying the objectives of TQM. The Malcolm Baldrige criteria place 30 percent of their emphasis on customer satisfaction; the activity hierarchy provides a framework for comprehensive assessment of satisfaction across the total business.

Each activity is classified as being either mission-related or non-mission-related work. Information about cost drivers -- those events or environmental situations that cause costs to be incurred -- may also be associated with each activity. These two information sources work in tandem to inform managers about which activities contribute to accomplishing the business strategies and about how the costs are generated.

Finally, activities may be associated with one of the four ``Cost of Quality'' categories (prevention, appraisal, internal failure, and external failure) or with the ``Cost of Business'' (e.g., advertising or production). Whereas many organizations have spent a year or more developing an accounting system to estimate cost of quality (many service organizations don't even attempt these analyses), an accurate estimate can be produced in a matter of hours with FACT. Many people express disbelief that cost of quality can be computed so quickly. However, with an activity and cost database as extensive as that developed in a FACT study, the assignment becomes straightforward. The activities are sufficiently detailed that one simply assigns each to a cost of quality category, or splits the activity cost between two categories (`Train agents' is a cost of prevention and `Resolve customer problems' is a cost of external failure). Teams responsible for the FACT study quickly come to consensus on these assignments. Computing cost of quality is now trivial; you simply sum the activity costs in each category. (See the figure ``Activity-Based Organizational Model.'')

The project plan of a FACT study is a relatively short, but intensive, process. The data is first collected and then entered into a proprietary software program which produces summary reports for analysis. A study encompassing all 1000 employees at IPS took about 2 months to gather all data and another 2 months to analyze the data and to develop initial recommendations. The FACT study team consisted of twelve part-time members, allowing the study to be conducted without disrupting the ongoing organizational processes.

Restructuring Through FACT Analysis

An oft-cited phrase in discussions of advanced manufacturing technologies is that implementation of just-in-time processes uncovers a ``hidden factory'' of overhead expenses for logistical, balancing, quality, and change transactions that account for a large portion of the value added by production. A similar discovery may be made in service organizations by identifying ``hidden service providers.'' This point is clearly illustrated by a situation uncovered through the FACT study conducted at IPS (see ``Hidden Service Providers'' for a fictionalized example). Higgins reveals these data to senior managers at IPS by saying, ``You don't realize how much help you're getting from around the company.''

Similar discoveries of hidden service providers were made by the Controller, where costs related to the activity `Administer Financials' range far beyond the Accounting department budget used within typical cost accounting systems. At first glance, the Controller thought that the FACT data was wrong since reported costs were twice what he expected. However, after reviewing the details, he agreed that the numbers were right. Managers throughout the company ``help'' the controller `Budget/Forecast Operations' or `Document Employee T&E'. Similarly, in the Human Resources department, all managers contribute to the activities `Hire Exempt Employees' or `Review Employee Performance'. In each of the three examples, the actual cost incurred to provide these activities far exceeds the apparent cost shown in the respective department's budget.

These sorts of realizations -- that actual activity costs often significantly exceed or fall short of presumed costs -- are of central importance in both promoting and guiding the restructuring imperative. Restructuring is promoted by creating disenchantment with the status quo, and it is guided by providing a sound basis for restructuring decisions.

Promoting efforts to restructure

Managers resist change. At least, they resist unfounded or unjustified change. Once a foundation has been laid and changes are justified, resistance quickly breaks down and may even turn into enthusiasm for the change.

The above example of hidden service providers illustrates one way in which resistance to change is overcome at IPS. Other means are used which are also based on the data collected during the initial phases of the FACT process. Pareto analysis, commonly taught as an analytical technique in quality circles, involves sorting a set of items in descending order and focusing quality improvements on the top of the list, i.e. on the largest items. A Pareto analysis of activity costs is illuminating in that managers are often surprised to discover how few activities (at the top of the list) account for a large fraction of the total organization cost. It is also surprising to note which activities show up at the top. Managers often have an intuition that an activity cost is too high, but intuition cannot justify significant investments. The relative positions of activity costs can provide information heretofore unavailable for justifying corrective investment. In a similar manner, a descending list of activity costs is produced for each cost of quality category, allowing managers to focus on the top offenders in each area.

Focusing one's attention on the top of the sorted list, however, only tells half of the story. When Higgins gave an eight-page listing of descending activity costs for the MoneyGram product to an IPS executive, he was prepared to throw away the last seven pages and focus his attention on reducing the top activity costs. However, Higgins pointed out that the activities at the bottom of the list are also candidates for attention. Important activities may not be receiving enough resources. A costly activity at the top may represent an internal or external failure cost, while an ignored activity at the bottom represents a preventive item that, with a small investment, could dramatically reduce the failure-related activity. That's the difference between value analysis and cost reduction.

Perhaps the most powerful analyses enabled by the FACT process are cross-functional and cross-product summaries of activity costs. A matrix is constructed with an indented list of the activity hierarchy on the left side and either the major functional areas or the product families across the top. As shown in the accompanying exhibit (see ``Cross-Product Analysis'' for a fictionalized example), one might note that the percentage of cost allocated to the `Support & Service Agents' activity has a wide variance across products. The hierarchical activity list also indicates how the total cost is divided among subactivities. By itself, this observation is interesting, but of limited value. The obvious next question is ``Why?'' Fortunately, the FACT model provides a response.

Activity Product A Product B Product C
1 Develop Products      
4 Support & Service Agents 3% 7% 5%
41 Receive & Document Inq. 1.5% 2% 2%
42 Handle Agent Problems 0.5% 3.5% 0.7%
43 Supply Agents 1.0% 2.5% 2.3%
Total Cost




The FACT study includes a record of how each person's time (and therefore salary) is allocated to the activities. A department's expense items, other than personnel, are also allocated to the activity base. It is important to note that a department does not allocate its personnel costs as one opaque cost pool. Thus, when an activity's total cost is out of line or when the cost is oddly split between product lines, the supporting data is easily available for explaining these anomalies. In general, this audit trail of detailed data is crucial for investigating the basis for activity costs.

If the old cliche ``numbers speak louder than words'' applies anywhere, it applied here. The numbers are arrived at through a logical estimation process which is validated through a structured interview with each manager. Although the numbers may be only estimates, the ratios of activity costs provide a powerful message to managers. For example, when the ratio between the activities `Resolve agent problems' and `Train agents' is 20:1, then, even if the respective costs are off by 10 percent, even 25 percent, the ratio still points to a potential issue. Perhaps agent problems may be reduced through increased training. A similar conclusion may be drawn from a cost of quality analysis. Agent problems, a cost of external failure, may be reduced by investing in agent training, a cost of prevention. The data provided by FACT supports either of these analyses.

Cost of quality has been faulted by Deming, who asserts that if you focus on process improvement then cost of quality will take care of itself. However, given the ease of computing cost of quality within the FACT framework, it provides a useful means for creating disenchantment with the status quo and for promoting restructuring efforts.

Guiding firms in restructuring

In order to avoid the slash and burn mentality of most cost-cutting efforts, a better basis must be provided for rational decision making. Stephen Roach warns that ``cost cutting must be judicious'' if service organizations are to avoid the hollowing that occurred when the manufacturing sector chose to sell short its future by trading its long-term capacity requirements for the sake of near-term financial gain.

One of the most useful mechanisms for guiding restructuring is cost driver analysis. Cost drivers cause activities to be performed, and they may be positive (customer orders) or negative (customer complaints). By eliminating negative cost drivers, their associated activity costs will also be reduced.

During the data collection phase of the FACT process, each activity is identified as being either mission or non-mission related. Employees may be spending a disproportionate amount of time on activities which are not directly related to the mission of their work group. Attempts to restructure the organization should be directed at shifting effort to mission-related activities. Are the members of the sales department performing the activity `Close sales' or are they `Resolving agent problems' or `Preparing department reports'? Only the first activity contributes to their department's mission.

Major work activities are often diffused across the organization. Many different individuals may be involved in an activity which represents only a few full-time-equivalent employees. When activities are highly fragmented, they should automatically be subject to further analysis. Consolidation of such activities may present an opportunity to improve performance; however, some activities should remain distributed among managers. Budgeting, personnel reviews, and other similar tasks are some examples of normally fragmented activities which contribute to effective management.

Empowerment or Enforcement?

The way in which data generated by a FACT study is used in an organization can have an impact on how the results, and the process itself, are assimilated into the ongoing strategy. This warning is especially true in promoting a successful TQM strategy where one wishes to gain the support and involvement of every employee. If the data about misallocation of resources is wielded as an enforcer's hammer, then you can expect neither the honest participation of employees in producing accurate estimates of activity costs, nor their commitment in creating and implementing improvements.

There are exceptional situations where an activity-based analysis is the best means for quickly down-sizing an organization having severe profitability problems. However, unless carefully managed, a FACT study used in this context is likely to be viewed as an axe for paring organizational fat. While this approach is superior to the slash and burn mentality that removes necessary activities along with the fat, there is little likelihood that the benefits will be incorporated into long-term changes in the strategic management processes.

The approach employed at Integrated Payment Systems seeks to empower employee teams (referred to as Corrective Action Teams (CAT)) to develop and implement changes, armed with copious data for guiding their efforts in the most fruitful directions. The FACT process facilitators serve as consultants to the CAT teams, helping them to interpret their data and teaching them additional data analysis techniques as required to restructure their respective organizations, to refocus their efforts, and to refine their processes. These approaches are consistent with TQM objectives for instilling attitudes for continuous improvement and for establishing quantitative performance measures to assess the current quality level and to monitor improvement.

Implications for Service Organizations

IPS' FACT study already has provided valuable insight into its business operations and yielded substantial savings in several operational areas. Not surprisingly, it uncovered some information that could come from a similar study conducted in a manufacturing organization, i.e. high cost areas, costs for difficult systems changes, and validation of overhead costs. In IPS' instance, however, the study went further to identify less tangible, but no less valuable, opportunities for quality improvement.

``The study was helpful in creating a fact base for potential organizational improvements,'' said Eula Adams, IPS Executive Vice President, who has 20 years of audit experience at a Big Eight (now Six) accounting firm. ``In the past, these kinds of changes at service firms were driven largely by opinion -- by what people thought would work. Our study provides a solid fact base and, consequently, justification for changing the organization.''

According to Adams, the study garnered support for organizational improvements by identifying areas of overlap and duplication, and focusing attention on non-value-added activities. ``Overlap and non-contributing areas are usually fairly evident in manufacturing environments,'' he said. ``But that wasn't necessarily the case in the service sector, until now. This study was helpful in identifying those opportunities and indicating how we might address them.''

Bob Kuhnemund, IPS' Chief Financial Officer, agrees with Adams' assessment and also suggests that the study is helping to link customer satisfaction to specific organizational activities. ``In addition to hard numbers, it helped us develop an understanding for mission-related activities -- what's hot and what's not,'' he said. ``That's extremely helpful from an organizational perspective and can have a dramatic impact on overhead costs -- reducing in some areas and increasing in others.''

Just two months after the FACT study was completed and results tabulated, IPS documented savings in several areas totaling more than $1 million. While creating considerable savings, the changes also created some equally dramatic customer service improvements.

``Information from the study led us to ways that we could streamline the operating processes of some very important customer functions for MoneyGram,'' said Dan Carrington, Vice President of MoneyGram Operations. ``Normally, improvements like these are associated with large one-time expenses. But, we realized initial net savings of $350,000 and improved the speed of providing these services.

``These are the kinds of changes we have to make because we're competing in what many people perceive to be a commodity service area. Also, we have a competitor with more experience and we have to set ourselves apart from their way of doing business.''

In spite of the rather unlikely birthplace of activity-based costing within IPS' Quality Assurance department, these testimonials point to the fact that ownership and backing of the FACT approach to cost management transcends the quality function. As Higgins concludes, ``The FACT results have now become Accounting's numbers. They own it. But we're also finding out that other areas can own it too -- areas like customer service and operations.'' IPS is currently considering ways of supplementing its existing cost accounting systems with an activity-based approach based on the initial FACT study.

The FACT project plan provides a structure and methodology for assessing the current situation: where effort is being expended and what value is contributed by those efforts, from the customer's viewpoint. An activity-based model is used for this assessment, and a number of approaches were described for guiding a firm's restructuring decisions. But an answer has not been supplied for the CEO's concern about whether American Express's investment in TQM can be tied to improvements in the bottom line.

The results of TQM efforts in a service organization -- changes to processes that affect an activity's performance -- can be measured by using activity costs to quantify the expected benefits of the change and by computing a return on investment. For finer resolution, activity costs may be separated by department, product, employee classification (manager, individual contributor, or clerical), or all of the above in order to better estimate the benefits of a proposed change. In relatively small scale FACT studies, activity costs can be traced to the percentage of time devoted by individual employees. One can then accumulate these cost savings figures by activity for an accurate estimate of TQM-derived benefits. These analyses are often performed within IPS's CAT teams as they progress through the phases of the FACT job plan -- information, creativity, evaluation, planning, and implementation.

In his HBR article, Roach concludes by saying, ``Only with the proper measurement tools in hand can services assess restructuring options.'' FACT provides such a tool. Roach suggests that an activity-based value analysis could assist banks in reallocating information technology away from low-value-added activities like transaction processing and administration to more analytical applications like interest-rate swaps. To more directly address Roach's concern with deploying information technology in a firm, the FACT framework might be extended to allocate capital investments to the activity hierarchy. A cross-functional analysis could then yield summaries of information technology investment by functional area for each activity. In a similar manner, both expenses and capital investments could be summarized for those activities related to two or more competitive strategies, or for non-mission-related activities, in order to analyze the relative cost consumed by each one. When costs of strategies are dramatically out of line with strategic priorities, some form of restructuring should be considered.


Questions or problems regarding this web site should be directed to
Copyright � 1996-2001 Ontogenics Corp. All rights reserved.