Increasing the Efficiency of Corporate Travel Management through Macro Benchmarking RUSSELL A. BELL

AND

RICHARD C. MOREY

This article demonstrates a macro benchmarking methodology to estimate the potential of a corporate travel department to reduce its business travel costs, with no change in its travel demands or operating environments. Results indicate that with a large enough pool of corporations appropriate benchmarking partners can be identified that match the volume and complexity of the firm’s multidimensional travel demands, as well as the firm’s operating environment. Targets are also provided for the different types of support costs, as well as sensitivity analyses related to changes in service levels and travel demands.

PRACTICES IN CORPORATE TRAVEL MANAGEMENT

Current Trends Recent trends in business travel indicate that organizations becoming more concerned about the escalating costs of business travel conducted in concert with their normal business activities. Two studies report results from biannual surveys of corporate travel policies and practices. American Express (1991) reported that 60% of 1,564 companies, in a survey of travel management practices, agreed that rising travel and entertainment costs represent one of management’s top concerns. This figure was up from 45 % in 1986. The responding organizations also reported a doubling in the number (from 16 % to 32 % ) that have appointed or hired travel managers over the same time period. In a similar vein, Runzheimer International (1991) reported that 90% of the 980 companies surveyed centralize their travel policy development, 85 % centralize cost monitoring, and slightly over half now centralize reservations (56 % ) and ticketing (53 % ). These last three numbers are all up by 10 to 12 percentage points in just two years. Thus, it becomes clear that a growing number of organizations are translating their concerns into actions. are

Overview of Practices Before the advent of travel management, corporations had been practicing travel cost control through the normal budget and management oversight processes. Corporate travel management nowadays is, however, a distinctly different process that includes the establishment of a travel management department that oversees the entire travel program. It includes the consolidation of travel arrangements into a centralized organization that is either staffed by employees or outsourced. It also involves the use of consolidated volume to obtain Russell Bell is an Associate Professor in the Marketing and Tourism Faculty at Cornell University in Ithaca, New York. Richard Morey, from Tulane University, is a Visiting Professor at the same institution.

from travel providers such as airlines, hotels, and firms in the negotiation of corporation-wide contracts at reduced prices. In addition, the travel manager modifies the travel behavior of the organization’s employees (e.g., by means of a corporate hotel directory) to maximize cost savings without impairing their normal business effectiveness. Bell (1993) provided a detailed description of these programs and their evolution and proposed a four-stage model for this evolution. The four-stage model does imply that as these programs mature or stall at one of the stages, they reach a point of diminishing returns in terms of incremental savings achieved from the addition of new components. In other words, after the major components have been instituted and the majority of compliance from the organization’s travelers has been received, the cost savings can typically only be maintained or marginally increased. This situation leads an organization to begin the process of seeking new methods to achieve continuing incremental savings and improved service. A relatively new process termed &dquo;benchmarking&dquo; is a strong candidate to achieve this increased efficiency.

leverage rental

car

BENCHMARKING

Background Although benchmarking is now fairly widespread, it is a relatively recent phenomenon, especially in the service sector. The Japanese are given credit for inventing the concept by visiting a wide range of businesses and learning how the most efficient organizations operated. Taiichi Ohno recounts that he adopted the U.S. supermarket inventory replenishment system for Toyota’s parts inventory system after visit in 1956 (Ohno 1988). Xerox is often cited as a leader in benchmarking among U.S. corporations (Bresada 1991; Richard 1991; Jacob 1992). Xerox’s benchmarking manager has written a text on benchmarking, which provides extensive details on this process and develops a working definition of benchmarking: &dquo;the search for industry best practices that lead to superior performance&dquo; a

Inherent in this definition are two issues: &dquo;Who is best and what are the best practices?&dquo; Even though these questions can be addressed independently, the answers to each separately are of little value without the other. Thus, the process of benchmarking must be a search for the best organizations as well as a search for the processes that produce the desired results.

(Camp 1989, p. 12).

The Micro

Benchmarking Process The practice of micro benchmarking, as detailed by Camp and widely followed by practitioners, is dominated by the search for specific practices that will enhance the efficiency of a specific operation (e.g., warehousing). This process can be characterized as a micro approach (in contrast to a macro approach) since the focus is on the discovery of specific practices that can be added to produce an improvement in the desired facet of the operation. Most frequently, the gain in efficiency is measured in improvements in simple ratios (e.g., inventory turn ratios, sales per square foot of retailer space, etc.). Often there is little recognition of the detailed environment in which the improvements must take place, as well as of the impacts on service, quality, and other measof output. This search for leadership companies, in spite of the limited focus, appears to be complex, lengthy, and as much art as science. Camp recommends the use of consultants, vendors, functional experts within an organization, industry associations, and public databases as important sources of information in the selection of comparison companies. The level of detail provided for the selection of benchmarking partners seems insufficient when one recognizes the range of operating environments that firms face and the multitude of outputs produced. This is especially true when one appreciates that many resources are public in nature in that they are used to produce many different types of outputs, with no clear allocations by types of output. An example in corporate travel management is the use of corporate staff and technology to jointly reduce expenditures to airlines, hotels, and rental ures

car

agencies.

Site Visits at

Benchmarking Partners

Once suitable

have been identisite visit focused on acquiring knowledge from the comparative companies on the specific processes and practices that make them relatively more efficient. For example, in the context of travel management (as discussed by Bell and Morey 1994), &dquo;What steps are used in the negotiation process with airlines? Are formal request-for-proposal documents that encompass the entire set of travel itineraries used or are negotiations on specific routes where they have high volume use? Are negotiations with hotel chains held at the corporate level for a chain-wide contract, or are they held with specific hotel properties in specific markets ? What practices are employed to monitor and enforce compliance of vendors with negotiated rates? What methods are used to monitor satisfaction and enforce compliance among their travelers to measure the impact of their policies on overall business success? What information systems are used and which ones are most useful? How are these information systems used to increase their savings? Have organizations typically hired at the upper quartiles, or have they hired at the average levels and then used extensive amounts of training?&dquo;

benchmarking partners

fied, the next step is typically

12

a

MACRO BENCHMARKING

Searching for Benchmarking Partners If appropriate choices in the selection of benchmarking are to be allowed, those programs that operate in environments that are relevant to the constraints that exist in similar environments should be employed. Similar operating environments would include factors such as volume and complexity of the travel activity, similar trip patterns, usage of airports that have similar competitive environments, and, of course, similar levels of service provided to corporate travelers. The goal of the benchmarking process should be to produce the maximum net dollar savings possible (i.e., reduction in total travel costs offset by the total of all expenses to achieve these reductions). This is to be accomplished in the firm’s operating environment, all the while delivering the firm’s chosen level of service and continuing to meet all of the firm’s travel demands. This approach is in marked contrast to the micro approach of seeking improvements in only one facet.

partners

An

Approach to Travel Management Benchmarking In order to select travel management benchmarking part-

Macro ners

that have similar environments and service,

a

new, multi-

input/multi-output approach is required. Such an approach would require comprehensive knowledge about the travel characteristics of a large number of organizations for inputs. The suggested approach can be characterized as the identification of a peer group of other corporate travel departments, operating in the same time period, and matched on amounts and complexities of travel, the service levels provided for travelers, and travel environment operating difficulties. In addition, the actual data from the previously discussed sister firms can be used to yield concrete targets for resource consumption by type of expenditure. Peer grouping approaches have successfully been used in evaluating productivities and efficiencies for school districts (Charnes, Cooper, and Rhodes 1978), utilities (Fare, Grosskopf, and Logan 1985), criminal superior courts (Lewin, Morey, and Cook 1982), military recruiting districts (Lewin and Morey 1982), and hospitals (Sherman 1984; Morey, Fine, and Loree 1990; Morey et al. 1992; Morey et al. 1994). A powerful, analytical tool known as data envelopment analysis (DEA) (see Charnes, Cooper, and Rhodes 1978) has been proposed by Collier and Storbeck (1993) in the telecommunications industry to aid in selecting benchmarking partners. DEA is particularly applicable when costs are not known, but rather when only the physical levels of consumption

(i.e., man-years of labor, numbers of fax machines, etc.) available. Unfortunately, DEA will rate an operation as efficient when it is relatively efficient from a technical standpoint, but will indicate it is inefficient when cost data are included. As a result, conventional DEA modeling could result in benchmarking partners that would not yield the maximum cost savings possible. Our suggested approach differs from Collier and Storbeck in that it uses a powerful extension of DEA known as allocative data envelopment analysis (ADEA) (see Banker and Mainderatta 1988; Sickles 1988; Morey and Retzlaff-Roberts 1993; or Morey, Fine, and Loree 1990). It is similar to DEA in its use of linear programming to handle the multifaceted nature of a firm’s operating characteristics and in producing are

peer grouping of operating units built up from the actual performances of other units operating at the same point in a

time. However, the thrust of ADEA is to estimate the cost expenditures of a unit that would produce: at least the same level of all outputs, do so in no easier an operating environment, deliver at least the same level of service, and achieve the lowest possible cost (see Figure 1). The emphasis in this approach would be on the identification of appropriate benchmarking partners that used a different mix of resources that was more cost efficient than that used by the firm under evaluation. This is in contrast to DEA, where the thrust is on maintaining the firm’s same &dquo;technological recipe,&dquo; but cutting back all resources by the same proportion. The site visits would uncover the detailed processes, cultures, and so on that helped give rise to the final dollar savings. Travel management clearly has the multifaceted environment that could benefit from the ADEA approach. Organizations involved in travel management vary significantly in the volume and complexity of their travel activity. Large corporations such as IBM, AT&T, and General Electric have travel budgets in excess of $500 million, while smaller organizations such as law firms, consulting organizations, and regional manufacturing firms have travel budgets of much smaller magnitudes. These same diverse organizations operate in vastly different competitive situations, depending on their destination patterns and their air hub patterns. Also, each of them could be using different service level philosophies for their corporate travelers.

The Data Pool

Concept This approach would require the establishment of a centralized pool of information that would encompass a uniform data set about each organization. It would include each organization’s volume and complexity of travel, environmental constraints, service levels delivered, and detailed cost expenditures. For example, the travel data might consist of standard airfares (such as the coach rate available at the time of travel) applied to the actual travel patterns of each organization. This would help establish a kind of baseline of nominal costs to measure any savings in the organization’s actual expenditures for air travel. Similar comparisons would be made for hotels

(perhaps using the hotel corporate rate for the base) and rental expenditures. The support cost data would consist of the costs for office space, exempt and nonexempt personnel, the car

amortized cost of hardware and software for the technology that is employed, and, of course, fees for consulting and commissions paid to travel arrangers. A number of environmental constraints (e.g., a local cost-of-production index) would also be included. Finally, the level of service provided could be categorized as average, superior, etc. This data pool could then be used to measure the overall efficiency of any given participant at any point in time as well as to set targets for resource consumption. The unit of time selected could be every year for a management report, or perhaps after a significant change in management. An index of efficiency could be calculated that would measure how far the organization was from the most efficient firm in its peer group. An index of .85 would mean that the organization under review was 85 % as efficient as the leading company in its peer group. The data pool could also become a valuable tool in the selection of actual benchmarking partners since the procedure would identify the peer companies. The confidentiality of each individual organization’s information could be maintained at a credible and unbiased location such as a major university or an industry association.

MACRO BENCHMARKING ILLUSTRATED: A COMPARISON OF 31 TRAVEL DEPARTMENTS In order to demonstrate the power of ADEA in providing insights into the efficiency of corporate travel management programs, an illustrative set of variables was developed. To deal with the credibility issue, the illustrative data were reviewed for reasonableness by a panel of individuals involved in corporate travel departments (see Appendix). The data represent

outputs for 31 different travel departments (airline expenditures, etc.), a set of inputs (costs for labor, technology, fees, etc.), and a set of environmental variables (levels of service, of airline competition on the company’s major travel routes, etc.). Both large- and medium-sized corporations were included. Since the major goal of a travel department is to reduce the cost of travel, but not to dictate the trip legs and

degree

FIGURE 1 MACRO BENCHMARKING INPUTS AND OUTPUTS

(This chart also appears

in

Bell and Morey

1994.) 13

their times, it was also necessary to create a set of nominal costs for each type of expenditure against which to measure the degree of savings achieved. The nominal costs were different for each of the companies to reflect a range of possibilities, such as trip length and hotel type usage patterns. Each of the nominal amounts did not include the cost of commissions or fees payable to the travel arranger, since many of the firms have succeeded in negotiating rebates of some portion of these commissions. Rather, these types of costs were included as a type of support cost. Furthermore, costs for meals and entertainment were not included, since no organization used in the creation of this data set actively negotiated for discounts in this category. The estimated spending for travel expenses and the associated support costs for these 31 large- to medium-sized companies vary in size of travel budgets (7,000 to 833,000 trips), degree of success in negotiation for each of the travel components, variation in hotel type usage, percentage of trips using rental cars, and internal versus external orientation for travel arranger personnel, etc. The internal versus external orientation represents a choice of carrying the travel arranger personnel on the payroll of the corporation or paying commission payments to the travel management company or travel agency. Cost variables were created to represent a realistic range for exempt labor, nonexempt labor, part-time labor, technology and equipment costs, building space, and fees paid. Commission payments were included in fees as mentioned earlier, since the nominal and actual costs of travel excluded commission charges. Two environmental variables were created to illustrate constraints in a company’s ability to achieve travel cost discounts. For example, the level of service desired for a company’s travelers (such as next-day ticket delivery from a centralized office) could well be a cost driver in the travel arrangement process. Also, the degree of competitiveness present in the company’s major travel itineraries would affect their ability to achieve a lower level of discounts (e.g., if a commuter leg was required at the beginning and the end of most itineraries, the ability to reduce airfares would be lessened). Two levels of service (high and average) and two levels of degree of difficulty (high and average) were spread across the 31 companies in this data set.

Summary

of Data

The 31 companies in this demonstration had an average of $93.5 million in nominal airfare that exhibited a range of $5 million to $583 million. In comparison, their actual airfare costs averaged $58.5 million and ranged from $4 million to $333 million. Similar comparisons are available for hotel and rental car expenditures. The labor costs averaged $.17 million for exempt, $.13 million for hourly, and $.02 for part-time payroll, with fees averaging $3 million, space costs $.02 million, and technology costs $.04 million. Finally, 15 of the 31 corporate travel management departments delivered &dquo;excellent&dquo; service and 15 of the 31 faced a &dquo;difficult&dquo; operating environment. Details for each of the 31 companies are contained in the Appendix. A Numerical Illustration This section concretely illustrates the types of inputs and outputs used in identifying a given corporate travel department’s efficiency in its operations as well as targets for achieving full efficiency. The method uses past data and constructs 14

for each corporate travel department a custom-tailored peer grouping of other actual, efficient corporate travel departments, operating in the same time period, and matched in terms of size, volume, difficulty of environment, and level of service rendered. The prime focus is on minimizing the total costs to achieve the given department’s actual set of travel demands, in its actual given operating environment, and delivering its actual service level.

Notation and Definition of Variables The mathematical details of the relevant linear programformulation can be found in Bell and Morey (1994). The various types of insights available will be illustrated using corporate travel department #9. (See Appendix for the details of #9.) To summarize, travel department #9 had a nominal airline bill of $29.82 million, its nominal hotel bill was $8.2644 million, and its nominal rental car bill was about $1.276 million. Once again, these figures represent the cost that would have been incurred using coach fares, standard corporate hotel rates, etc., applied to the travel patterns (trips taken, actual hotels stayed in, and actual rental cars rented) for each company. Its level of difficulty was normal for the calendar year under review, and its level of service delivered was excellent. Its actual airline expenditures for the calendar year were $23.43 million, its actual hotel expenditures were $6.6456 million, and its actual rental car expenditures for the same period were $1.2268 million. Hence, the overall discounts obtained by the ninth corporate travel department for its airline, hotel, and rental car operations, respectively, were 21.4%, 19.6%, and 3.92%. The average discount over 31 firms (see Appendix) was 30.39 % for airlines, 11.67 % for hotels, and 11.72 % for rental cars. Furthermore, corporation #9 also spent a total of $1.1497 million to achieve the above discounts, of which $.176 million was for exempt labor, $.1656 million for hourly labor, $.03 million for part-time labor, $.054 million for technology costs, $.0132 million for space, utilities, etc., and $.7029 million for fees (to consultants, travel agencies, etc.). Also, note it is one of the smaller corporate travel departments, facing a total nominal travel bill of about $39.33 million (versus the average of $129.89 million). Finally, it is one of the 15 corporations delivering a superior level of service. These numbers are summarized in the first column of Table 1.

ming

Questions Given the above situation for to be addressed are:

corporation #9,

the four

questions

(1) Which (if any) of the 30 other corporations would be

(2) (3)

(4)

suitable benchmarking partners for corporation #9? Is corporate travel department #9 operating relatively efficiently, and if not what is its efficiency rating? What are concrete targets for each of its nine types of cost expenditures, if it were as efficient as the best in its peer group, among those matching on service, operating environments, nominal airline charges, nominal hotel charges, and nominal rental car charges? What is the estimated decrease in firm #9’s total cost if it were to downgrade its service from excellent to average ? What would the total increase in its costs be if its air routings, city pairs, and time of week changed so that its difficulty factor went from easy to difficult?

Answers Consider the answers to these questions. Corporate travel department #9 was found to have four benchmarking partners, namely firms #4, #15, #22, and #26, of which the first two could be classified as &dquo;major.&dquo; (The first two firms are referred to as &dquo;major&dquo; because these two firms together constitute more than 96 % of the total weights used to build the &dquo;ideal&dquo; firm.) Each of these four travel departments had weights (in the building of the composite firm) from the linear programming model that were greater than zero. To help establish the credibility of the benchmarking partners for #9 and to illustrate the matching process, consider Table 1, which focuses on the 14 facets of a travel department: column two identifies those for corporation #9; column 3, those for corporation #4 (one of its suggested benchmarking partners); column 4, the characteristics of the composite ideal corporation on the best efficiency frontier, which was used to set targets for department #9; and column 5, the suggested changes for department #9 to have been rated as efficient. First, consider the degree of matching achieved by the

suggested benchmarking partner #4 (for the evaluation and improvement of travel department #9). Note that travel department #4’s nominal airline charges and its nominal hotel charges are both larger than those for #9. Its difficulty factor and delivered service level are at least equal to those of #9’s. The only facet where #9 has a slightly easier situation than #4 is in its nominal rental car charge (i.e., $1.2759 million for #4 versus $.9686 million for #9). Note that #4, however, met all three of its travel demands (a total of $57.4476 million compared to #9’s $39.33 million) at a total percentage cost (of its sum of total nominal costs) of 63.44 %, compared to 87.47 % for #9. Hence, intuitively, it appears corporate travel department #9 might benefit from learning some of the processes and procedures that travel department #4 actually used. The only reservations for exclusively using the performance of travel department #4 to set targets for #9 are (a) #4’s slightly lower nominal car rental charges and (b) #4’s substantially higher nominal airline charges and slightly higher nominal hotel bill. In order to deal with these imperfections, we shall &dquo;adjoin&dquo; (see Bell and Morey 1994) to actual corporate travel department #4 three other real departments,

TABLE 1

ANALYSIS FOR TRAVEL DEPARTMENT #9 (ALL DOLLARS ARE IN MILLIONS)

15

namely #15, #22, and #26, with respective weights of .47 for #4, .493 for #15, .022 for #22, and .015 for #26. Note the

operating environment, related to the ease for negotiating airline discounts, was changed from easy (its present status)

weights sum to one. Hence, we have created a composite corporate department with a nominal airline bill of: .47 ($45.0569 million, the nominal airline charges for #4) + .493 ($6.244 million, the nominal airline bill for #15) + .022 ($95.784 million, the nominal airline bill for #22) + .015 ($235.48 million, the nominal airline charges for #26) = $29.82 million (exactly the nominal airline bill for #9). Similarly, the composite travel department spent, for example, for its actual airline expenditures, a total of: .47 ($25.746 million, the actual amount for #4) + .493 ($4.46 million, the actual for #15) + .022 ($54.732 million, the actual for #22) + .015 ($108.128 million, the actual for #26) = $17.095 million. Note exactly the same weighting factors are used to compute the nominal and actual expenditures for the composite corporate travel department; those same weights are used to estimate the individual consumption levels for all six support costs for the composite department for every performance dimension. This amount of $17.095 million (for the efficient composite &dquo;actual&dquo; airline expenditures on line 7, column 4) is to be compared with $23.43 million, the amount actually spent by #9’s travel department to meet exactly the same nominal air-

to difficult?

line bill, in an easier negotiating environment than that faced by the composite. Using exactly the same weightings of the actual expenditures of these four travel departments yields the amounts the composite would have spent on hotels and rental cars, as well as on exempt labor, fees, etc. These quantities are shown in the fourth column for the composite efficient travel department and provide targets for #9. Note the airline, hotel, and rental car nominal amounts for the composite corporation (rows 1, 2, 3, column 4) turned out to be exactly the same as for department #9. Upon inspection of this column, we see the composite efficient corporate travel department would have spent only $17.095 million for its actual airline purchases, $6.80 million for hotels (versus the $6.646 million actually spent by #9), and $1.026 million for rental cars (versus the actual of $1.227 million actually spent by #9). Hence, the composite efficient corporation would have put more of its efforts into negotiating deeper discounts on the airline side and rental car side than did #9, and less on the hotel side. The ratio of the optimal total of the amount paid for air, hotels, and rental cars to that actually paid is:

The composite efficient corporation would also have spent somewhat smaller total amount (in labor, fees, etc.) to have done this (i.e., $.853 million, Table 1, column 4, row 11) versus $1.14 million spent by #9. This yields a ratio of .748 (see fourth column of table for #9). In all (for the total of expenditures to airlines, hotels, and rental car agencies), it would have spent $25.86 million, compared to the actual of $32.434 million, for exactly the same nominal bills, giving rise to its overall rating of 79.7 %. (See column 5 of Table 2 for department #9.) Finally, consider the answer to question #4 (i.e., what would be the estimated savings available to an efficient corporate travel department #9) if (a) it downgraded its service from excellent (its present staus) to average or (b) if its a

16

For the first question, if #9, operating efficiently, lowered its service level, it is estimated that it could reduce its total cost from $25.8646 million to $25.6297 million, a savings of $235 thousand (or .9%). This is obtained by choosing a different peer member group that can lower the costs once the original service level no longer has to be met. If its operating environment were made more difficult the cost would increase from $25.8646 million to $27.5778 million, an increase of 6.6 %. Both of these estimates are obtained by resolving the linear problem formulations in Bell and Morey (1994). For example, in the case where the difficulty factor for #9 is changed to &dquo;difficult,&dquo; the only peer group members allowed are those also facing a difficult environment. The composite group is now forced to change as the original peer group included two departments, namely #22 and #26, that had &dquo;easy&dquo; operating environments. This is no longer permitted under the new scenario.

Summary of Results The analyses just described in detail for #9 was repeated 30 more times, one for each corporate travel department. Due to the paucity of the data set (i.e., only 31 departments), it was not possible to identify appropriate benchmarking partners for 17 of the 31 corporations; this is because of the required matching in the analytical peer generation process (on each of the nominal levels, its environment, and its service level). Larger data sets would minimize the chance of this occurring. For 14 of the 31 corporations (i.e., more than 45 % of them) appropriate benchmarking partners were indeed identified ; they are shown in Table 2. For the 14, we see the average inefficiency score was 83.9 %, with a low of 65.4 % and a high of 99.2 %. The 83.9 % means that about 16.1 % of the total cost being expended by the average inefficient operation could be saved if it operated as efficiently as the composite firm formed by a combination of the performances of the peer group members. We also noticed an interesting result for departments #7, 11, 13, 18, and 28. They are not spending enough on support costs, given the volume and complexity of their corporate travel needs. (This follows from the ratios of more than unity in the third column of Table 2.) That is, in order to minimize the total of the firms’ nine expenditures (i.e., the six support costs, plus three expenditures for air, hotel, and rental cars), the above departments need to increase their support costs. These increases in labor, technology, fees, etc. would result in even larger decreases in air, hotel, and rental car expenditures so that the total is actually lowered. Finally, note that for the 14 inefficient operations, the average drop in total travel expenditures would have been nearly 20 % (i.e., 1-.801) (see last row of Table 2, second column) if they were to become as efficient as the best in their peer group. CONCLUSIONS

Summary The thrust of this article is to introduce a new approach for selecting benchmarking partners and corporate targets. In this respect, it adds to the list of benchmarking tools that contains the suggestions of Camp (1989) and Collier and

Storbeck (1993), among others. The macro focus is imporbecause many of the resources being consumed (e.g., technology, rent, etc.) cannot be allocated to any one type of travel activity. Hence, any micro approach focusing on a single type of output is dependent on an arbitrary allocation of &dquo;public&dquo; costs. Additionally, micro approaches ignore differences in operating environments and service levels provided. For these reasons, simple engineering ratios (e.g., discount off corporate hotel rate) do not tell the whole story. To accommodate a more macro approach toward selecting benchmarking partners, we need an approach that can simultaneously handle multiple outputs, service level, and environmental factors. It would also be desirable for the approach to use as few assumptions as possible, be transparent, and provide concrete targets for the firm seeking help. Allocative tant

data envelopment analysis has all of the above desirable features. We have exercised the logic on a representative data set and illustrated the types of quantitative insights available.

Caveats and Future Needs The strengths given above notwithstanding, approach is not a panacea and some caveats

our

suggested

are

in order:

our approach focuses not on averages but on the best that can be accomplished, it is sensitive to data outliers. Hence, considerable care needs to be taken (range and alpha-numeric checks) to eliminate spurious data. This is a legitimate criticism of other approaches

(1) Because

to

benchmarking.

TABLE 2 BENCHMARKING RESULTS FOR 31 ILLUSTRATIVE CORPORATE TRAVEL DEPARTMENTS

a

weighting factor used to construct the composite corporation in the efficiency frontier was at least .15. Illustration for department #9: The identified major benchmarking partners for department #9 are department #4 and department #15; for department #9 its ratio of the optimal expenditures for air, hotel, and rental cars to its actual was .796; for department #9, the ratio of its total optimal support costs to its actual was .748; finally, the ratio of the total of the nine optimal costs (that is, three travel expenditures plus six support costs) was .797. (It is the sum of these nine costs that the optimization program minimizes, while meeting the same level of nominal costs, m an environment that is similar, while delivering at least the same level of service.) Major

means

that the

JOURNAL OF TRAVEL RESEARCH

17

ex post facto, or retrospective, so that the identification of appropriate benchmarking partners is based on comparisons of past performances, hopefully not too distant in the past. There is no assurance that by the time the physical, on-site audits are done, the firms selected as benchmarking partners will have continued to operate efficiently. Also, their operating environments or service levels might have drastically changed so that their &dquo;best practices&dquo; may no longer be relevant to the firm looking for help. Again, this is a weakness of the other benchmarking tools. The travel demands, difficulty of operating environment, and service levels desired were modeled using the firms’ &dquo;nominal&dquo; costs and categorical indices. These proxies, of course, are far from perfect and need to be refined as more data and other measures become available.

(2) The approach is

(3)

These caveats and others notwithstanding, we have exercised the suggested logic on a nontrivial database (31 companies, each with 18 separate factors) with representative data, and have shown the credibility of the benchmarking partners selected and insights yielded. Given the present interest surrounding benchmarking, as well as the many institutions being formed with benchmarking as their focus (such as the International Benchmarking Clearinghouse of Houston, Texas, part of the American Productivity and Quality Center, and Best Practices Benchmarking and Consulting, Inc. of Lexington, Massachusetts), researchers should develop better analytical tools to handle the very large corporate databases becoming available. In this manner, those firms whose practices and cultures can be of most help to other firms will be identified and hope-

fully replicated.

APPENDIX TABLE A-1

ILLUSTRATIVE DATA SET ON 31 CORPORATE TRAVEL DEPARTMENTS

18

TABLE A-2 ILLUSTRATIVE DATA SET ON 31 CORPORATE TRAVEL DEPARTMENTS

TABLE A-3 ILLUSTRATIVE DATA SET ON 31 CORPORATE TRAVEL DEPARTMENTS

aOne

bOne

indicates that excellent service was delivered, while zero indicates average service was delivered. indicates that firm operated in a difficult negotiating environment, whereas zero indicates an easier environment. 19

REFERENCES American Express (1991). The American Express 1991 Survey of Business Travel Management. New York: American Express Corp. Banker, R. D., and A. Mainderatta (1988). "Non-Parametric Analysis of Techmcal and Allocative Efficiencies in Production." Econometrics, (November): 1315-32. Bell, Russell A. (1993). "Corporate Travel Management Trends and Hotel Marketing Strategies." Cornell Hotel and Restaurant Administration Quarterly, (April): 31-39. Bell, Russell A., and Richard C. Morey (1994). "The Search for Appropriate Benchmarking Partners: A Macro Approach and Application."

Omega (forthcoming). A. (1991). "Benchmarking." Financial World, 160 (September): 28. Camp, Robert C. (1989). Benchmarking — The Search for Industry’s Best Practices that Lead to Superior Performance. American Society of Quality Control. Milwaukee: Quarterly Press. Charnes, A., W. W. Cooper, and E. Rhodes (1978). "Evaluating Program and Managerial Efficiency: An Application of Data Envelopment Analysis to Program Follow Through." Management Science, 27: 668. Collier, David A., and James E. Storbeck (1993). "A Data Envelopment Approach to Benchmarking in the Telecommunications Industry." Ohio State Faculty of Management Science Working Paper. Columbus: Ohio State University. Fare, R., S. Grosskopf, and J. Logan (1985). "The Relative Efficiency of Illinois Electric Utilities." Resources and Energy . Amsterdam: North Bresada,

Holland.

Jacob, R. (1992). "How 19: 102-6.

20

to

Steal the Best Ideas Around." Fortune, October

Lewin, A., and R. C. Morey (1982). "Measuring the Relative Efficiency and Output Potential of Public Sector Organizations: An Application of Data Envelopment Analysis." International Journal of Policy Analysis and Information Systems, 5: 922. Lewin, A. Y., R. C. Morey, and T. Cook (1982). "Evaluating the Administrative Efficiency of Courts." Omega, 10: 410. Morey, R. C., D. F. Fine, and S. W. Loree (1990). "Comparing the Allocative Efficiencies of Hospitals." Omega, 18: 71. Morey, R. C., D. F. Fine, S. W. Loree, D. L. Retzlaff-Roberts, and S. Tsubakitani (1992). "The Trade-off Between Hospital Cost and Quality of Care." Medical Care, 30 (8): 677-98. Morey, R. C., and D. L. Retzlaff-Roberts (1993). "A Goal Programming Method of Stochastic Allocative Data Envelopment Analysis." European Journal of Operational Research, 71: 379-97. Morey, R. C., Y. A. Ozcan, D. L. Retzlaff-Roberts, and D. J. Fine (1994). "Estimating the Hospital Wide Cost Differential Warranted for Teaching Hospitals: An Alternative to Regression Approaches." Medical Care

(forthcoming). Ohno, Taiichi (1988). Toyota Production System: Beyond Large Scale Produc-

Cambridge: Productivity Press. Jr. (1991). "The Quest for Quality: A Race Without A Finish Line." Industrial Engineering, 23 (January): 27. Runzheimer International (1991). Survey and Analysis of Business Travel Polition.

Richard, Norman E.,

cies and Costs. Northbrook, IL: Runzheimer International. Sherman, H. D. (1984). "Hospital Efficiency Measurement and Evaluation: Empirical Test of a New Technique." Medical Care, 22: 927. Sickles, R. C. (1988). "Allocative Inefficiency in the Health Care Indus-

, edited by Ali Dogtry." In Application of Modern Production Theory nomaci and Rolf Fare. Boston: Kluwer Academic Publications.