Airline Reservation Systems – A Report & Overview (1991)

Download A Copy of This Report

Airline computerized reservation systems (CRS) are the primary form of travel agency computerization in the world. These systems manage the millions of reservation requests and cancella­tions, fare, and reservation pricing requests[1] that are initi­ated by travel agencies using these systems–not to mention the thousands of database changes that occur daily. The CRS function as extremely powerful and valuable distribution and marketing tools for their airline owners. Today’s travel agency competitive environment is largely defined and controlled by airline CRS.

This report describes:

  • Airline reservation and distribution systems
  • CRS processing and communication concepts
  • Relationships between CRS and other industry components
  • CRS and airline competitive strategies, as these pertain to reservation technology
  • Non-reservation systems that interconnect with airline CRS
  • Profiles of the major international airline CRS

Airline distribution has been greatly shaped by deregulation and the rise of travel agency computerization. Critical to understanding how travel distribution is managed is an appreciation of the influence computerization exerts over travel agency reservation and purchasing practices.

The role of automation in this area must not be underrated, as automation provides the vehicle for airlines to effectively enfranchise dealerships among the travel agency community[2].

These dealerships play an important role in broader carrier distribution strategies. Computers are routinely cited as representing significant competitive barriers to domestic U.S. and international carriers operating within the U.S. (as well as in other areas), when in reality market leverage is gained through much more subtle and effective means.

The marketing relationships described here are supported by automation within the travel agency community, based upon the conditions existing in the U.S. market. Although not directly applicable to many areas of the world, these U.S. experiences are indicative of what other markets may expect as regulation is relaxed and competition increases. As U.S. carriers expand beyond their national boundaries, both as to routes and agency distribu­tion, other countries make expect U.S. style distribution practices to be introduced and competition to become more aggressive.


The fundamentals of airline reservation handling have not changed substantively for decades. From the first basic, no-frills scheduled air service between Amsterdam and London, inaugurated May 17, 1920 by KLM – Royal Dutch Airlines[3] using a leased De­ Havilland DH16[4], passengers have required reservations and reservation record management of some type.

Objectives of a Reservation Process

The systemization of commercial air travel in the 1920s and 1930s introduced the same basic trip components that are used today:

  • Scheduled services that potentially navigate a number of intermediate stops, where a passenger may desire to travel only a portion of the aircraft’s complete route.
  • Fares[5] that apply to each portion of the passenger routing.
  • Documentation, in the form of tickets which certify payment has been made and a traveler has a right to transportation on a specific flight schedule between a specific origin and destination.
  • The necessity of managing reservations and cancellations, so that passengers may rely upon the airline’s ability to accommodate them on their desired flights.

Commercial aviation and retail travel services revolve around the public’s desire for certainty of accommodation on airline flights, and therefore, the necessity of pre-reserving and pre­issuing airline tickets[6].  Some air services, such as no reservation air shuttles, eliminate the need for pre-planning (all travelers are accommodated) and the need for complex ticketing (a single or very simple fare structure is used). In general, however, the role of airline tickets are ticket issuance has not changed in 70 years.

Early Reservation Management

Prior to the electronic age, reservations were managed manually using record books and manifests, reservation cards, and other printed media. Many simple reservation requirements are still (1990) met in this way[7].

After the Second World War, air travel and airlines expanded rapidly. The “jet age” arrived in 1958[8], and with it more passengers and larger aircraft, as air travel quickly replaced train and bus as the main form of mass public conveyance. As airline schedule planning became increasingly complex, reservation management systems reached unwieldy proportions.

Reservations were typically recorded on cards, each card corresponding to a trip between specific city-pairs. Cards were then organized by  departure date and stored in tubs that were rotated among various reservation agents as necessary. Reservation requests from outlying offices were either telephoned to a central facility or transmitted via  teletypewriter[9] and processed manually upon receipt. Reservations could only be made 30 days in advance of flight departure.

In the early 1950s, when the idea of electronic reservation systems first became practical, it took approximately 2 hours to completely process an average reservation transaction[10]. By the early 1960s, this time had been shortened to 45 minutes–still in a mostly manual mode.

Early Computerization

During these same years, electronic, digital, computers were entering what is termed the “second generation” of modern computational devices. Whereas the “first generation” relied upon vacuum tubes and a number of competing mass storage systems,[11] “second generation” devices used the much more flexible and reliable transistor and magnetic core memory almost exclusively.

In the early 1960s, work commenced on “third generation” digital computers. This era, usually described as lasting between 1964 and 1975, is characterized by the introduction of integrated circuits[12] [13], but  which  also  produced operating systems, machine instructions, and methods of data and memory management that were sophisticated enough to form the basis for many of the large-scale commercial computers used today.

Thus, it is not uncommon to find large, important programs and systems that are rooted in this era.

Development of “PARS”

This is the period when commercial data processing technology matured to a point where large-scale airline reservation systems were practical. The early airline reservation system projects were managed by IBM[14] and UNIVAC[15], with the most successful projects developed in connection with IBM. There were several initial development projects pursued by competing airlines, including United Airlines, Eastern Airlines, Trans World Airlines and SABRE,[16] developed in conjunction with American Airlines.

American invested what was, for the time, the tremendous sum of $40 million which “would have bought the airline four Boeing 707s at a time when the entire fleet consisted of fewer than 25 aircraft”.[17]

SABRE development began in 1958 and required four years of planning and two years of programming. The system was introduced in 1964, using a modified IBM 7094, a second-generation system. The essential SABRE operating system was proprietary to American at this time.

In 1964 IBM also introduced the System 360, which represented an almost complete departure from many machine design concepts employed by IBM prior to that time.

System 360 architecture was the most successful computer design up to that time, and among the most successful of any to date. Although the 360 did not use integrated circuits in many of the key areas usually associated with third-generation systems, its design has much more in common with third than second-generation devices.

Understanding the basic premises of the 360, which was supplanted by System 370 in the early 1970s, is important to appreciating many of the strengths and limitations of PARS-type CRS. System 370 architecture is still (1990) central to IBM’s mainframe commercial processor line.

Key Aspects of System 360 Architecture

  • Large, complex instruction set intended to build many capabilities into the basic design of the machine, thereby limiting, to a degree, the need for many complicated programming steps.
  • Employed instructions embedded in microprogramming, stored in ROM components, allowing application portability across a number of machines using the 360 instruction set but employing a range of circuit designs and speeds.
  • First large-scale commercial use of emulation, whereby applications not directly employing the System 360 instruction set could be run by simulating conditions on the hardware for which they were designed.
  • Introduction of memory cache to aid overall machine speed. Successful introduction of large-scale virtual memory operating systems and applications, permitting time-sharing among a very large group of simultaneous sessions.

American faced serious implementation problems with the early SABRE system, among them capacity limitations. Shortly after the introduction of System 360, American implemented a machine that provided supportive processing functions to SABRE, but did not migrate to the newer and much more powerful and efficient System 360 environment.

In 1968 American reached an agreement with Eastern Airlines that allowed American to modify and implement Eastern’s Programmed Airline Reservation System (PARS) on its System 360. PARS[18] was another IBM project, working along different lines from American’s SABRE effort, and built on behalf of Eastern. Other airlines had launched similar, but unsuccessful, reservation system development projects[19]. These eventually adopted the basic PARS system, and modified it to meet their own needs.

PARS concepts, designs, and features still form the primitive foundation of most airline reservation processing and management systems today. An international variant of the PARS system, IPARS, followed the initial PARS project.

PARS was designed to run on IBM’s latest 1960s vintage commercial processors, System 360, and used a variety of existing and proprietary tools and software systems. Since its introduction, IBM has assumed a leadership role in developing, enhancing, and upgrading airline systems, supportive programs, protocols, and the hardware on which they run.

System 360/370 architecture assured PARS’ usable life in the 1990s. As IBM upgraded its commercial mainframe processor line, newer, faster, more powerful machines replaced older systems in existing PARS installations. This was made practical because upward compatibility was an important feature of the System 360 concept. Even though many essentials of the PARS environment remained fundamentally unchanged over the years, hardware advances have allowed PARS to continue meeting the needs of ever expanding CRS networks.

What Is PARS?

Rather than a monolithic program set or product, PARS is the essential foundation and conceptual basis for airline systems. Because these concepts were replicated in later modifications and revisions of the original PARS, there is great similarity among all PARS systems, particularly the major U.S. CRS. This applies to operational practices, operator formats, system capabilities and limitations, and ongoing enhancements. There has been significant cross-sharing of PARS functionality because of IBM’s role as developer and maintainer of the software tools that comprise much of the PARS environment.

Although this text speaks of “PARS and IPARS systems”, it does so in a generic sense, as frequently there is little or no real difference in the software of one PARS-based system as opposed to another. Where there are functional or architectural distinctions, these are clearly identified. In reality, however, all major U.S. CRS have operated as independent systems since the late 1960s.

Inventory vs. Passenger Systems

The initial airline automation efforts, including PARS, were inventory systems, and not passenger systems. Based upon the management priorities defined in the late 1950s, gaining centralized inventory control was the most important priority. These early transaction systems processed reservation requests against declining inventory allotments, but lacked even the relatively simple passenger record, service request and file access capabilities of later PARS-type systems. The extensive schedule information for offline (non-host) airlines was also lacking.

More extensive databases were partially visualized, but were only incorporated into later PARS software releases after the initial inventory systems had been operative for some time. These were also implemented in varying degree of sophistication, depending upon the airline and system in question.  There is more commonality between the schedule, reservation, and booking modules of various PARS-type systems than there is in other databases and functions.

Communication Limitations

Early airline systems lacked the sophisticated communications capabilities that are taken for granted today. Large public data networks, sophisticated data transfer protocols, and standardized interfaces were unknown. System planners operated in an environment where even long distance telephone calls were relatively rare and expensive.[20]

This necessitated low-cost, reliable, easily implementable and maintainable communications methods to support information exchange between airline host systems. Because of their role as product distribution tools, not just inventory systems, airline computers in particular rely upon data communications to exchange information rapidly and effectively. This is because interline[21] reservations and ticketing have been fundamental to world travel for many years. An airline’s biggest “customers” are other airlines’ internal reservation centers and computers, who make reservations for passengers as both as a pure convenience to their customers and as a source of passengers that may transfer or “connect” from these airlines.

Communication between airline systems is necessary because the operator of one system needs the ability to update the inventory and create reservations in another system that operates independently of the one the operator is using. The system functions as a computerized work environment and productivity tool for the operator. If inventory sales and updates in “offline” systems can be accomplished electronically the reservation process is much faster. The alternative is for the CRS operator to use the te1ephone, which always increases the time required to make a passenger’s reservation many times.[22]

The passenger also wants reliable schedule, availability, and reservation information as quickly as possible. Particularly in the U.S., travelers are accustomed to “instant” results and are dissatisfied if lengthy delays are introduced. Electronic messaging, together with airline database practices, makes virtually instant reservation confirmations practical.

Early airline communications used teletypewriter-based messages that could operate reliably at low transmission speeds, using existing, inexpensive, communication circuits. These communications operated in a network-based mode, thus eliminating the need for dedicated circuits between each reservation system. Airline messaging systems are described later in this report under the heading Connectivity.

These basic communication formats and practices are another legacy from the early days of airline computerization and still serve much of the industry today, albeit in an enhanced and upgraded form[23]. While the original interline communication methods are practical, they are neither flexible nor particularly efficient and have been extensively replaced by modern communication technology.

It is interesting to note that considerable work had been done on developing more efficient data communication methods during the same years airline systems were developing, particularly to support military projects[24]. For various reasons this technology was not incorporated into the early airline system projects[25].

Software Development in the 1960s

It is difficult for many people, regardless of their familiarity with computers and programming, to appreciate many of the constraints large-scale developers operated under in the 1960s. Today, most commercial programmers are avid PC enthusiasts and are accustomed to working “online” both at home and professionally. Using a CRT that directly accesses a computer workspace, they write program “code” or instructions electronically in a disk file, “compile” their programs (produce a set of machine commands ready for execution by a computer based upon the “source code” or instructions created by the programmer) online, and test them as each portion of the program is completed.

Sometimes emulation programs are available that permit PCs or other small systems to behave like large commercial mainframes, so that programmers can experiment in an environment much like that where their final work will reside[26]. Modern programming has come to depend upon interaction with a processor and “debugging” or error-correction almost at will.

It was not uncommon for programmers in the 1960s to never actually see the computer they worked on–indeed, programmers working exclusively on software of similar vintage are sometimes similarly disadvantaged. Instructions were written on pre-printed forms and transferred to punch cards by clerks specially trained for that task. Those cards were then “loaded” into the computer in batches and the program was run. Debugging was always necessary, but the programmer could not depend upon constant interaction with the processor, because the input, load, and run functions were much more complicated and time-consuming.

Hardware in the 1960s was limited and expensive. Computer time was valuable and also expensive, so much so that skilled programmers often spent considerable time trying to reduce the number of steps their programs required. Modern programmers have much greater and less expensive machine resources available to them, which eliminates the need for such precision[27].

Computer Languages

Computer languages are sets of symbols, instructions, and statements used by programmers to control the operation of a computer and its peripheral devices. In its basest sense, computers “understand” only binary code – series of Os and 1s, or “on” and “off” conditions. While it would be possible to construct programs using binary code, it is so difficult that it is almost never done. Other “languages” are used that can then be translated or “compiled” into instructions that the computer can execute.

Assembly Language is a term loosely used to describe symbolic instructions that a programmer uses to direct specific, predetermined responses from the computer. The programmer does not work directly with binary code, using instead alphabetical or other symbols that equate directly to binary instructions[28]. Since assembly language directly controls individual actions or series of actions that the computer performs, it is specific to each machine where it is used. In other words, an IBM System 370 uses a different variation of assembly language than does an IBM PS/2. Many early programs were written using forms of assembly language, and this form of programming is still used where programmers must optimize the responses of the machine for which the program is developed.

Most airline system and application programmers use a form of assembly language, which accounts for much of their development and maintenance. IBM is introducing enhancements to the programs used by airlines that, once implemented, will allow more flexible development tools to be used.

So-called “higher-level languages” approximate human communication and are compiled into object programs that can be executed by computers. Whereas an assembly language programmer exercises significant control over each action the machine performs, a high­ level language programmer focuses on the result desired and relies upon the compiler to control individual machine instructions. While this obviously results in greatly increased programmer productivity, a given program can and often does result in very inefficient object code that does not optimize machine resources or capabilities.

Transaction Processing

At the time PARS was developed there were several successful “high-level” programming languages in use[29]. Online transaction processing systems, similar to those required by airlines, had also been built or planned by this time. The architects of PARS were able to build on this body of experience, while tailoring a system to meet the specific needs of the airline planners that also participated in the project.

IBM’s Airline Control Program (ACP) was the first transaction processing tool applied PARS-type airline systems, beginning in 1965. The Transaction Processing Facility (TPF) was part of ACP, but developed more specialized applications and was used more independently as time progresses.

In 1984 IBM positioned TPF as a strategic product that would serve a different specific market segments together with MVS and VM, IBM’s state-of-the-art operating systems. IBM modified TPF by introducing numerous extensions that dramatically increased its power and attractiveness to users[30]. TPF now runs in native mode[31] on IBM 3090 machines, the system used by the majority of current generation airline systems, meaning that a host operating system is no longer required.

An early problem with TPF, indicative of the era when it was developed, was that only a uniprocessor[32] could be used.

This was corrected with releases subsequent to the 1984 announcement. An equally important enhancement was TPF support for compiled source code, whereas previous1y TPF programmers used assembly language.

Although TPF enhancements were successfully embraced by many airlines, some airlines, including American, continued to use a highly optimized form of ACP for years. American converted to TPF from a hybrid ACP/TPF environment completely in 1991. This is very significant, as it allowed American to operate directly on hardware and software that are part of IBM’s product mainstream, thereby reducing American’s long-term software maintenance expense and giving it a much more direct path to the latest technology[33].

Current Airline Automation

TWA PARS Snapshot

  • TWA’s system has about 50,000 communication terminals in the field worldwide. The size of the database is about 850gb (two to the thirtieth power; significant storage for its day) and is stored on 340 of 3380s DADS. It is fully duplicated for performance and availability reasons. About 8% of 850gb are occupied by passenger records.
  • A typical daily workload is about 20 million transactions. The peak performance rate is about 800 TPS.
  • The average rate is 552 TPS with a response time of 1.5 second.Reliability of these systems is also an important parameter. A partial database recovery (recovery from transaction or system program failure) takes about two minutes. A cold start (total failure or starting system from scratch) takes about eight hours. In 1972, TWA lost about $2 million and in 1976 about $250,000 due to system unavailability. However, the airline has improved its software reliability and did not have a major system outage for the last two years.
  • Initially, it had a 9083 CPR that ran on a 3083 uniprocessors. Since then, it has gone to multiprocessors (3390), four of which are dedicated to on-line processing. TWA can accommodate up to eight processors that might give it adequate power to reach 1,000 TPS

Some idea of the complexity that current airline reservation systems have achieved can be gleaned from the accompanying description of TWA’s PARS system[34].

By comparison, SABRE, the largest system in used by travel agents today, currently operates its Passenger Service System, one of five major divisions of SABRE, on 6 IBM 3090/200 series machines and 224 IBM 380 disk drives. Each drive has approximately 1,320 mb, giving the system a total storage capacity of 296 gb per volume. Current memory is 726 gb.

SABRE has announced figures of over 1,800 transactions processed per second during exceptional peak periods, and anticipates 2,200 or more in the future.

Airline systems share the same goals as other high-performance transaction processing environments:

  1. Reliability. As illustrated, downtime can significantly affect revenue and costs.
  2. Responsiveness. User productivity degrades rapidly with decreasing response times. In the highly competitive CRS area, small variances can result in a lasting competitive disadvantage.
  3. Price/Performance. An optimal environment must be in p1ace at an acceptab1e cost, both to the CRS and the user. This is a major challenge as CRS continue to develop and user appetites for new and more extensive applications increase.
  4. Enhancements. The CRS environment must be sufficiently adaptable to allow creation of the functionality users require. This has been difficult using old-style ACP/TPF systems, but will change as more standardizes and sophisticated operating environments are created.
  5. Process and System Integration. The most effective development path for the CRS will be to effectively integrate local, distributed systems and microcomputer­ based workstations into the CRS environment. This is an extraordinarily difficult task to do effectively, given pricing limitations[35] and the restrictive nature of CRS communications, architecture, and databases.

Transaction Processing Fundamentals

Today all airline systems operate as online transaction processing systems[36]. A transaction system allows operators to initiate queries and requests for specific applications, view the results of these applications in real-time, act upon the results of this process, initiate follow-on transactions that are determined by the results, and update files or records based upon information resident with the user rather than within the system. These events are driven by a series of independent interactions between the user and the system, known as transactions.

Transaction-based systems differ from other conventional computer applications (here applied specifically to airline systems which are among the largest and most complex transaction systems):

  • Transactions Discrete and Independent. While the information contained within one transaction may affect other transactions (perhaps because they use a common database where updates initiated by one transaction may be passed to a subsequent transaction), one transaction is usually not dependent upon another, nor connected with another as to time.
  • Batch-oriented systems, by contrast, manipulate data elements as a group. There may be extensive data dependencies in such systems, and omissions of some elements would require the entire application to be rerun.
  • Databases Updated and Queried as Needed. The transaction system works using information stored in a database and the results of online data entry are usually recorded in a database.
  • Geographic Separation. Transaction systems function without regard to where their users are located (as long as the requisite communication services are in place) Users working on the same system may be separated by thousands of miles.
  • Concurrent Sessions. Because transaction systems usually must support many (sometimes thousands) of users, they emphasize performance and throughput[37]. Since many system users may require access to the same databases simultaneously, transaction systems must support user sessions[38] running concurrently, while maintaining performance and the integrity of individual data elements.
  • Queries and Requests for Applications Responded to Real-time[39]. Thus the timeliness of data supporting an online system become critical, as old data cannot be used to make reliable decisions. Data must also be synchronized within the system, so that databases are consistent for all users and transactions initiated or updated by one user available to other users and do not conflict with updates made by other users.
  • Limited Functionality. Airline CRS attach functions to the database query and update portions of the online system. These functions act upon data retrieved online from the CRS, or input online at the user’s site, and produce specific responses, such as calculated prices for traveler itineraries, printed itineraries, or tickets. Online transaction systems offer few functions to their users, as compared to the range of functions that could be designed for the system, and these have limited, specific purposes.
  • Data Accessibility. Most required CRS data are available continuously while the system is operating[40] and can be accessed at random. This contrasts with batch­ oriented systems where data necessary for one application may not be available unless that application is running, and where offline storage (such as on magnetic tape),and sequential data access is required.
  • Event-Driven. Transaction systems must undergo regular maintenance, as must all complex systems. CRS also have scheduled database updates, as for schedule changes among other purposes. The CRS will also schedule certain applications affecting many transactions to be executed at specific times. This helps shift system load to “off­ hours”[41].

Most transaction-related processing, however, begins when users initiate transactions by entering or retrieving data and not based upon any particular schedule. It is impossible to predict when specific transaction-based functions will be required or from where on the network they will originate. This necessitates complex capacity planning and continuous availability of basic system functions at all times when the system is operative.

  • Point-of-Entry Editing. Online transaction systems reject incorrectly formatted instructions or data blocks as these data are entered and return an error message to the users real-time, so that a correction can be made and the entry reattempted. This must be done because interaction with the user is essential to the online transaction system’s successful operation, and only “good” data can be used for this purpose.
  • Transaction Size. Most batch or single application computer systems take several minutes to execute all but the simplest applications. Many larger programs can take hours or days to complete. The real-time, event-driven nature of the transaction system requires responses within seconds[42]. This means that systems must be designed and programmed for optimal speed[43] and transactions are kept small to achieve the best responsiveness. CRS reservations, for example, are compilations of many individual transactions, each executed individually but which, together, complete the traveler’s desired reservation.
  • Data Integrity. Each part of the CRS database, particularly passenger reservation records, must be current at all times and reflect the best possible condition of the data[44]; no delayed updates are acceptable. This is necessary because the random nature of transaction systems makes it impossible to assume that any available record or data element will not be the next required by some user on the system.[45] This applies only to CRS data that are used for transaction processing.
Understanding TPF

TPF is a proprietary development and processing environment specifically designed to facilitate entry and processing of a large number of simultaneous transactions from multiple network terminals, where extensive database queries must be managed against large quantities of data and where overall responsiveness is a critical factor. A transaction is defined as a single entry, or a series of entries that are assembled in an electronic work area and completed at one time and that, in turn, initiate other actions within the system.

Industry experts define airline transaction systems as high-performance, in that they typically process in excess of 500 transactions per second, with an efficiency rate of one second response time over approximately 95% of transactions processed.

TPF is based upon a variant of assembly language. The entire work environment is so unique that programmers skilled in its application and management are called “TPF Programmers” and are among the highest paid professionals in that aspect of the data processing industry.

Online transaction systems are becoming more common ways to access and update databases. As these applications grow in size and complexity, more powerful and fully-featured development environments and tools, such as TPF, become important. TPF development is very expensive and demanding, causing many potential users (who are not already committed to installations partly dependent upon TPF, as are the major CRS, or who simply cannot use a less powerful transaction processing facility) to explore more affordable alternatives.

It is important to define TPF’s role as a transaction management, rather than a database system. TPF controls terminal­based queries and responses, and disk access, while other programs are developed or purchased and integrated to handle true database functions. 1960s era software development defined database management as strings of commands directed at multiple files (as opposed to the hierarchical or relationship-based structures).

These older techniques are decidedly less flexible than is today’s database technology, but are efficient to access and store in an online environment while requiring relatively little (per transaction) machine overhead.

Because TPF, and the airline systems that use it, are reflections of this environment, they are not especially powerful, based upon today’s standards, as transaction processing tools. Effective airline systems rely upon other applications, independent of TPF, to support the database functions required of online systems.[46]

Role of PARS Today
PARS Characteristics 

  • Careful, intricate design, where many programs and applications work in concert to produce desired results.
  • Relatively inflexible structure that must incorporate modernized versions of very old programs with new applications to meet today’s business needs.
  • An ACP/TPF basis that is intrinsically expensive and difficult to maintain, very challenging to update,­ and that does not always have the latest technology tools available to it.Non-standard communication protocols and interfaces that are unique to the airline industry, are character rather than data-oriented, and are not efficient when used to support modern applications.

All this contributes to the characteristics that still shape PARS-based systems. The legacy and limitations of PARS-type systems have been very frustrating to CRS suppliers and users. Agents appreciate the basic commonality between PARS-type environments, that makes it fairly easy to learn a new system once one has been mastered, but do not appreciate the cryptic formats it uses (characteristic of 1960s-era systems). Suppliers like the high-reliability that has been achieved through constant refinement and diligent maintenance of their PARS environments, but, with agents, are dismayed at the difficulty (and expense) of introducing modern functionality.

Many people believe that PARS systems are nothing more than antiquated 1960s technology and must ultimately be replaced by current technology. Several years ago this was a common theme particularly among misinformed government regulators who criticized airline software development and management practices as not being in the public interest.

“… a good example would be the April 6, 1982 issue of The Travel Agent, which reported then CAB[47] Chairman Marvin Cohen as having ‘scored one point with travel agents’ by telling a Washington, D.C. ARTA[48] meeting that airline computer reservation systems should be criticized for being unable to track complex fare changes under deregulation.

“Mr. Cohen expressed numerous errors of fact. One was:

If an agent seeks one display of the lowest prices of all airlines (flying) between Washington and New York, he or she will fail’, which drew light applause.

“In reality, I believe all major systems did an acceptable job of providing this information …”

“His remarks also included a statement that the software used by United and American is 20 or more years old and that it would take as long as two years to “reprogram the computers”[49].

“The technology to place remote computer terminals in almost every travel agency location did not exist 20 or even 10 years ago[50].

“No major airline today uses operating systems from 1961. Some of the basic formats and functions have been in use for a long time, but they are continually revised and updated. This is the nature of any complex data processing system. Programs for the major systems will not be re-written now or over the next two years. The systems will update and modify their current procedures, but will not start over.”[51]

All airline CRS use the most modern communications, storage, processor, and related technology available. PARS-based systems preserve the remnants of their beginnings, but each CRS has developed along different lines and operates and is maintained independently. Today’s CRS often struggle with significant expense and effort to free themselves of their inbred limitations, but each works with programs that are comparatively modern and bear the mark of their own unique development histories.


In the mid 1970s, prior to initial travel agency automation, ticket preparation and distribution were usually performed more cost-effectively by travel agencies than by carriers directly. Estimated average airline ticket office reservation processing costs were approximately 16.2%, whereas agency commissions were only 7%. Overall distribution costs could be reduced if a carrier’s agency volume increased, particularly regarding largely point-to-point business air tickets.[52]

Agency commissions during the regulated years were set by industry agreement, whereas today carriers determine compensation levels independently. General practice, with some exceptions, has held base commissions to 10% of net air ticket value (the total price less the 8% federal transportation tax), for an effective commission cost of 9.2% to the carrier[53]. At a 7% commission level, when the 8% tax was also in effect, the effective rate was 6.44%.

Airlines regarded travel agency distribution as cost-effective, without any fixed overhead costs, and, because travel agencies are found everywhere, a more efficient method of ticket distribution than was opening their own ticket sales offices.

Other strong incentives were present to direct more sales to agencies, among them the need to control steadily increasing costs which a regulated environment did not permit to be fully borne by the traveler. Distribution (or agency) costs can be influenced directly, unlike fuel, as an example, where costs increased dramatically during these years and over which a carrier has little direct control. The airline industry also expanded greatly during these years and needed a broad, relatively economical distribution system.

CRS as Travel Agent Productivity Tools

As manual processing methods do not permit great expansion of most agency transaction volumes, automation was made available as a tool to raise booking and document preparation productivity. The almost concurrent development of CRS and so-called “back room” (agency accounting systems)automation made large regional agencies and nationa1 chains a reality. A1though large agency networks existed before, mega-agencies with sales counted in the 10s and 100s of millions are impossible without automation, specifically CRS.

Value of Tools to the Agency

The primary reason the CRS suppliers are also interested in travel agency accounting automation relates to the overall effect such automation has upon agency operation and competitiveness. Because tools are so critical, particularly concerning large agencies, the vendors perceive significant opportunities to control agent loyalty through access to these tools. Where the vendor can provide comprehensive automation to the agency, front and back room, the value chain is strengthened regarding total services provided by the carrier as is the agency’s overall commitment.

Sophisticated tools are very expensive for a large agency, making conversions from one CRS to another much more difficult because of outright costs and retraining expense, were a conversion attempted. The vendor assumes that agencies receiving full automation support from a single source (accounting and CRS) are more committed, financially and practically, to the vendor than is an agency with partial recourse to an independent supplier.

This has caused most of the CRS suppliers to experiment (some less successfully than others) with supplying agency accounting system products. The aggressiveness with which the CRS vendors have pursued backroom agency automation as part of overall distribution strategies has largely decimated certain segments of the independent accounting system market.


As early as 1960, some major airlines began installing teleticketing machines in select travel agencies. This service was based upon standards implemented by the ATC[54]. Any airline could transmit tickets electronically to any travel teleticket machine by adhering to these standards.  Many, but not all, airlines offered the service.

The machines were modified teletypewriters that were accessed by the airline’s reservation center using normal telephone circuits. Once accessed, the machine could print a ticket, using special stock, in between one and three minutes. The most common (and for many years the only) supplier of teleticketing machines was RCA[55]. Agents purchased the machines and provided the telephone circuit necessary to operate it.

Agencies received teleticket machines under the sponsorship of a major airline, which took responsibility for ordering and installation of the machine, periodic maintenance[56], and informing other airlines of the telephone number necessary to access the agency’s machine[57]. Agencies were initially offered the machines based upon their sales volume, but the practice later became so popular that the smallest agencies used teleticketing.

Travel agencies paid small fees, of between 50¢ and $1.00 (there were several fee adjustments over the years) for each ticket issued.

The travel agency benefitted because laboriously writing tickets by hand (with not infrequent errors) could be reduced and because the airline issuing the ticket computed the price and assumed responsibility for its accuracy. The airline benefitted because agency ticketing errors were reduced and productivity improved, which was important for reasons just explained.

Teleticketing enjoyed widespread usage until well into the 1980s (past the beginning of the CRS era). Widespread agency computerization eventually rendered the machines obsolete, but many were still in place as late as 1985, although new installations had ceased.

Teleticketing was popular because it was the only semi­automated productivity tool available to most travel agencies. It was not without problems:

  1. Ticket issuance was not instantaneous. The earliest teletickets were composed using teletypewriter terminals designed for the purpose and transmitted once all relevant data had been input. Agents usually tried to give an airline 24 hours to complete this process, as even a “rush” same-day request could take hours.
  2. The more technologically sophisticated airlines eventually began transmitting teletickets directly from their computers but, as pricing software in particular was not fully developed, manual composition was still frequently necessary.
  3. As travel agency ticketing volumes increased, the burden of issuing tickets centrally became unacceptable for the airlines.
  4. While, overall, the teleticketing process was very reliable technologically, there were frequent mechanical problems, sometimes caused by using 9-part forms (necessary because airline tickets require this many copies) in the teletypewriter machines, resulting in misalignments and retransmissions.
  5. Not all airlines offered teleticketing services. This was particularly inconvenient where foreign airlines and frequently complex pricing and itineraries, were involved.
Airline Reservations in the Travel Agency

It was clear to agents and airlines that substantive agency productivity gains could only be achieved by making the same general type of automation used by airlines for reservations available to travel agencies. Initial projects were launched to explore technological and commercial options through industry trade groups. All these failed because of government regulatory difficulties, impractical business propositions, or poor technological vision.

Many airlines made reservation systems available to select agencies, chosen either because of their value and importance to the airline or simply as test sites, with the first systems installed by American Airlines in 1968. Trans World Airlines and United Airlines installed their own systems beginning in 1972[58]. These early systems were very limited functionally, as they displayed availability only for the host airline and were more airline reservation systems than travel agency CRS.

What is a CRS?

Several large airlines have created CRS entities that make reservation and related function systems available to travel agents. These are specialized online transaction processing systems and databases that are specially designed to meet the needs of travel agents. A CRS provides:

  • Airline and AMTRAK (rail) schedules[59].
  • Availability for transportation carriers that have agreed to pay booking fees (the CRS’ primary revenue source) for reservations made through the CRS.
  • Fares for services (air and rail) maintained in the CRS Storage of user and customer-specific databases.
  • Communication facilities necessary to support interline reservations and special service messages (such as seating and meal requests) that users may initiate on behalf of travelers whose reservations are made through the CRS.
  • Storage of traveler reservation files (PNRs) Applications, such as ticketing and itinerary issuance, that produce printed documentation based upon reservation files and other CRS data.

These services are adapted for travel agency needs. While the airline’s own reservation and ticket office operations require similar functions, they are not the same. For instance, there are very strict rules (federally mandated) specifying how the CRS may display flight availability, in order to give all airlines participating in the CRS equal access to potential customers, based upon the merits of their flights (departure, arrival, and elapsed time, stopovers, and origin and destination airports). An airline is under no such limitations for its own reservation system and will display availability and schedules for its own flights to the exclusion of competing services.

Thus the airline’s own reservation system is a different application, even if it shares computer resources with a travel agency CRS. The early attempts at travel agency automation preceded the development of the CRS[60] and were simply terminals for the airline’s system.

The First Travel Agency CRS

By 1975 the last of the “industry” system projects, this time an effort called MAARS[61], a switching system that would allow agency users to be connected directly to it and conduct sessions in individual airline reservation systems, as desired by the operator, was over unsuccessfully. A multi-access system called MARS[62] was eventually brought to market by ITT, with significant help from Eastern Airlines and a number of smaller carriers. It was never operationally or commercially successful and eventually folded after Eastern withdrew marketing support to pursue its own sales for its own CRS product, SODA[63], later SystemOne.

Figure 1: Relative CRS Market Share

Figure 1: Relative CRS Market Share


American Airlines and United Airlines began aggressively marketing SABRE and Apollo, their respective CRS, at this time. Their stated intentions were to invest in travel agency automation to the extent necessary to establish automated distribution networks nationally. Both corporations allocated millions of dollars to the project. TWA also actively marketed PARS at this time, but more on a regional basis.

These three systems were sophisticated travel agency-oriented reservation tools, for the time, and were free of many limitations that faced earlier efforts. A number of other airlines, among them Western Airlines (now merged with Delta Airlines), Allegheny Airlines (now US Air) and Alaska Airlines, offered reservation products to selected agencies within their primary markets. These were far less sophisticated than SABRE, Apollo, or PARS and were quickly eclipsed by the superior products.

Demand for CRS automation far outstripped the suppliers’ ability to install and support the systems, so the largest agencies were selected to receive systems first. Smaller agencies were frequently unable to automate for months after their larger competitors[64]. Once in p1ace, however, the CRS allowed agencies of all sizes and descriptions to achieve massive productivity gains and support sales volumes previously unreachable.

For example, using today’s CRS environment, good reservations agent productivity in a commercial (as opposed to leisure) sales environment is $1 million in air sales annually. Without a CRS, an agent would find it difficult to maintain commercial agent productivity much in excess of $250,000 annual air sales.

Later CRS Developments

CRS quickly proved itself as a good business and, more importantly, a distribution tool of unparalleled effectiveness in the travel industry. This motivated Eastern Airlines to market SODA (SystemOne) beginning in 1981 and Delta Airlines to market Datas II in 1983. Both vendors started late in the CRS game and found it difficult to build market share, particularly among the larger business travel agencies that are the most valuable distributors.

Eastern and Delta did create credible products and were able to protect their own primary markets from competing CRS to a large degree (the importance of this is discussed later). Delta merged its system with PARS in 1989, while SystemOne continued to be marketed independently into the 1990s.

Figure 2: CRS Market Share Trends

Figure 2: CRS Market Share Trends


Aggressive marketing and product sophistication allowed SABRE and Apollo, in that order, to establish the largest market share of any U.S. CRS, followed by PARS as a distant third. SystemOne and Datas II, when introduced, took market share from each of the earlier three, with a slightly greater percentage coming from PARS. SystemOne exceeded PARS in number of installed CRTs by 1987, although not in booking volume, due to the somewhat larger average size of PARS agencies. PARS installed base recovered rapidly, due to ongoing operational and business problems at SystemOne.

CRS Bias

After travel agency automation became an accepted business tool, the CRS recognized that manipulating screen displays results in incremental bookings, for one carrier at the expense of another, depending upon how the screen display is influenced. Between 70% and 90% of airline flights booked by a travel agent are reserved from the first CRS availability screen displayed, assuming the agent’s initial availability request was accurate, with 50% of flights being booked from the first line of the first screen. In a competitive industry, where product differentiation is often tenuous, some CRS e1ected to actively influence agent flight selection based solely upon screen management.

So-called screen bias in the Apollo and SABRE was well-documented, while PARS and SODA were less overt and DATAS II used its unbiased nature as a point of competitive differentiation.

Extensive user and consumer pressure eventually lead to the introduction of federa1 CRS standards that eliminated the most overt forms of screen bias. Some experts believe that displays can still be manipulated,[65] but certainly to a much lesser degree than in the past.

Bias is discussed in greater detail on page 40 of this report.

Basic CRS Operation

The following sections illustrate basic CRS operation by following a typical transaction from beginning to end. The SABRE system was chosen for these illustrations; however the similarities between PARS-type systems will make the examples applicable to most other CRS. The examples also assume that a travel agency, and not a corporate travel manager or airline reservation center, is using the system, and that the most current hardware and software releases are available.


Most travel agencies do not “power-down” their CRS equipment overnight. Restarting the LAN, once an installation is completely shut down, can be difficult and requires some expertise. Most suppliers also believe that normal hardware “wear and tear” is reduced if restarts are minimized[66]. No CRS start-up procedure is usually needed.

The agent’s first step is to LOGON to the system. This is a two-step process which involves a unique operator identification code and a password. While this is the most obvious level of security, there are others.

Each CRS user location is assigned a unique identification code, termed a pseudo-city(PCC) because it is not a true “city” code, as far as an airline is concerned. A CRS city code is a unique identifier that is assigned to each airport location the carrier serves. A pseudo-city, therefore, identified a physical location that is not an airport.

A PCC is a 4-character alpha-numeric string[67]. Normally only a single PCC would be assigned to each agency location, but some installations requiring special services, such as multiple communication circuits, may have several. The PCC allows the CRS to route reservations (PNRs) to electronic files, known as queues that belong to the location. The PCC also identifies reservations made by agents working at the location so that agents from other PCCs cannot access the same PNRs, unless that access has specifically been authorized by the agencies involved and the CRS.

This process is known as cross-access or bridging, depending upon the CRS. There is normally a monthly charge for each cross­access privilege open.

Each reservation created or modified by an agent is marked by the PCC of that office. If a reservation was created by one agency and later transferred to another, as customers occasionally request, the “ownership” of the PNR follows the last agency to modify the reservation.


Travel agencies are concerned about security. They believe that their PNRs should not be accessible to agents they do not authorize. There is considerable evidence to support the claim that even non-malicious “snooping” can create financial or perhaps worse, customer relations problems for an agency. An unscrupulous competitor might try to solicit new customers based upon information gained from examining PNRs, for example. While CRS security is a real travel agency problem, its actual effects are probably overstated.

The CRS have developed security procedures, such as those outlined above, which are sufficient to prevent most forms of unauthorized access. Similar measures are in place at the sophisticated CRS, although these vary considerably in their composition and effectiveness. In most CRS travel agents cannot view:

  • Reservations created by another travel agency using the same CRS, where no cross-access is authorized.
  • Reservations created by another travel agency using another CRS
  • Reservations created by the host airline’s own general reservations center- although some CRS do allow agents to access these reservations.

For obvious reasons, airline reservation centers, ticket offices, and CRS support staffs can access any reservation created by any travel agency.

Transaction Synchronization

Once initial SIGN-ON is complete, the agent is working in what is known as an AAA[68]or Agent Assembly Area. “Agent” in this sense refers to any CRS operator as an “agent”. The AAA is a short-term electronic file where the elements of the PNR, the passenger’s actual reservation, will be “assembled.” This is a key concept in CRS operation. Each agent, when signed-on to the CRS properly, has several AAAs available for use. The agent can toggle between them at will, and have reservations in various stages of completion in each -all functioning totally independently. The agent cannot move reservations or information between the available AAAs.

All work that a CRS “agent” does in a session is completed only in the AAA until a specific CRS “event” is invoked. This means that the agent can reserve flights, seats, special meals, and enter all other required parts of the PNR prior to permanently modifying the CRS’ major databases. In other words, the work the agent does is temporary–it exists only within the AAA until it is made permanent.

The agent does this by ending the transaction, that is, closing the AAA and updating the databases that will be affected by work previously done in the AAA. Each CRS has an End Transact (ET) button or command string (as in SABRE) for this purpose. Nothing done in the AAA is permanent until the “ET” event takes place.

This CRS process is both a valuable feature and a significant inconvenience. In fact, however, the CRS could not operate effectively without it. One ongoing challenge for all online transaction systems is transaction synchronization. The temporary nature of CRS work helps to address this challenge.

Reservations that the agent desires to modify must be retrieved into the AAA. As with new reservations, no change made to the reservation is permanent until ET. If another agent, perhaps from the airline or from another authorized travel agency location, displays the same reservation and begins to modify it, a conflict must result- which one is working with the “real” reservation? This conflict can also be programmatic from within the CRS. If an agent is modifying a reservation while an SSR message is sent from an off-line carrier or while a schedule change comes through, that agent is no longer working with the latest version of the reservation–any changes made might be irrelevant, based upon the new information.

The CRS resolves this conflict by saying that the first agent (or internal CRS process) to remove a PNR where potential conflicts exist by ET has created the latest version of the reservation.

Other agents working with the same reservation can continue to make changes in their AAA, but the CRS inhibits ET by sending the “simultaneous changes” response.


Once this condition is created, nothing agents can do will allow the system to accept their changes, because these were made on a version of the PNR other than the “latest.” If a great deal of work has been done, this can be a frustrating experience. The agent must invoke another CRS command, “Ignore” (I). This clears the AAA and allows the agent to retrieve the reservation and begin the process over again.

All changes or reservations made in the AAA when “I” is invoked are “ignored”. Reservations and seats return to inventory, and nothing is stored from that session in any CRS database. This applies only to host seats and inventory. When working in the AAA on the host carrier’s flights, desired inventory is pulled on a provisional basis. The host’s availability listings are reduced by the number and nature of services reserved in the agent’s AAA. Thus, if there was 1 seat left on AA flight 1 for a certain date, 0 seats would be shown as available once a reservation for 1 person had been started in an agent’s AAA. If a reservation in progress is “ignored”, 1 seat is returned to inventory.

If an agent is forced to “ignore” prior to completing the transaction, there is a possibility that the same services (flights or perhaps individual seat assignments) may not be available when, the process is attempted again. This does not apply to “offline” inventory and seat assignments, as the CRS does not send sell messages to other systems until ET- unless certain forms of direct access are used (explained later).

On the other hand, “ignore” can be valuable. If a traveler requests cancellation of a flight so that alternate reservations may be made, the exact services desired frequently are not available and the traveler decides to “stick with what they have”. Since all AAA changes are provisional, the agents only has to “ignore the transaction” to return the PNR to its last state.

Manipulating the AAA

The AAA may be displayed by the agent at will. In SABRE this is done by using the *A command, where * is the “display” character and “A” the delimiter for “all”, meaning all data in the AAA. If the command is invoked without a reservation in progress occupying the AAA, the response “NO PNR IN AAA” is generated. Agents may also display parts of the data in the AAA, such as “name,” “itinerary,” “passenger data,” and so on, by using the appropriate code. These commands are used because lengthy itineraries may be too large to display on one CRS screen, and the ability to segment parts of the AAA for display as necessary makes the transaction easier to manage.

Other parts of the PNR are not normally displayed and must be specifically requested. Once a PNR is changed after its initial creation, a “history” file is added that contains a complete listing of all changes made to the reservation, in descending chronological order. The agent uses “history” where there are questions as to why a reservation was modified or who authorized the modification[70] by invoking *H.

Once the AAA is occupied by a reservation in progress, that reservation must be ended by ET or I if it is a new reservation or if changes of any type are made to an existing reservation. If the agent tries to display another existing reservation before this is done the response “FINISH OR IGN PNR” is generated. If no changes have been made to an existing reservation, that PNR is automatically “ignored” when a subsequent request to display an existing PNR is initiated.

Text Reference

Once logon is complete, SABRE displays one or more brief informational messages that may be of use to agents. These are usually changed daily, but may be changed at will. Their purpose is to bring information to the operator’s immediate attention that might otherwise be overlooked, such as a major air traffic system or weather delay, a labor dispute that will disrupt air service, or particularly in the age of deregulation, the bankruptcy or cessation of operation of an air carrier.

These specialized text messages are related to the CRS’ general or Direct Reference System (DRS), which may be invoked by the agent at will, even while transactions are in progress in the AAA. The DRS is a very simple text-based file system, which is used to store information that is supportive of CRS operation but that does not require frequent updates, such as availability or flight information (scheduled arrivals), which changes constantly. Some typical DRS examples include:

  • Marketing messages from airlines, hotels, car rental companies, and others that participate in the CRS and pay a fee for the privilege of including information on selling their products through the CRS. Information about the products may also be included.
  • Immigration and visa information.
  • Ground transportation from major airports to the cities they serve.
  • Local weather conditions[71].
  • Information on shows found on Broadway, in London, Atlantic City, Las Vegas, or Reno.
  • Information on the correct operation of the CRS itself.

DRS “pages” are displayed by referencing their location within the filing system by means of a unique character string. The CRS then copies text from that “page” to the screen originating the request. Data in the AAA are not affected. This is done so that agents may break their work flows to answer customer questions or obtain information required to complete a transaction in progress, then return to that transaction without losing continuity with whatever work was underway.

Help System

DRS is not the only way a modern, sophisticated CRS, such as SABRE, has to make information on its correct operation available to users. A text-based “help system” is often available for operator reference.

FOX, SABRE’s help system, is illustrative. FOX is not context-sensitive, in that it cannot reference work in progress in the AAA and suggest commands or procedures based upon the transaction presently before the operator. Individual FOX text “pages” must be referenced directly by the agent.

FOX differs from ordinary DRS in that more complex commands, employing keywords, are used to locate desired pages. The agent builds strings of related words so that searching for the desired page is greatly reduced. Once an initial FOX page is retrieved, the agent is directed to additional references containing more detail.

Reservation Process Flow

The CRS agent begins work on a transaction by requesting availability of flights, based upon the traveler’s particular needs. Usually air availability is requested prior to cars and hotels, with the air itinerary forming a shell around which added services may be deve1oped, a1though a PNR need not necessarily contain air reservations.

Cars and hotels are the other large availability systems within the CRS. They operate separately from the air availability system, although many of the same principles are employed.

In any major CRS, the ratio of car reservations to air reservations is very low, and the ratio of hotel to air reservations still lower. This is partly because the availability and reservation environments offered by the CRS for hotels and cars are very limited and imperfect, but it is too simplistic to ascribe limited CRS use in these areas to functionality alone. The simple fact is that travel agents, the primary users of CRS, do not routinely make (or even offer) hotel reservations to their customers- CRS-based or otherwise. This is a selling deficiency and predates the introduction of CRS in travel agencies.

The basic SABRE availability screen is displayed by forming a command string beginning with the 1 delimiter, followed by the date of travel, cities for which availability is requested and a desired departure time:


The CRS responds with an availability screen that is organized based upon algorithms and display criteria unique to that CRS, and that are designed to display schedules that will be more desirable to the passenger first. Thus non-stop flights are shown before flights with stops, and direct flights (non-stop or otherwise) before connections where p1ane changes are involved. Schedules with the least (scheduled) elapsed travel time are also shown first. The availability display also indicates what type of aircraft is used for each flight and a numeric grading, based upon U.S. Federal Department of Transportation data on flight timeliness. The number of stops made by individual flights, if any, is also shown.

In the accompanying illustration, American Airlines flights (the “host”) show 9 seats as available for each class of service offered on the flight. If less than 9 seats are shown, then that number represents the exact number of seats allotted by the airline to that class on a specific flight. Offline flights show either 4 seats (the maximum number that can be sold under the “sell/no-sell” conventions used by the CRS)[72], or 0 seats, if availability has been closed for that flight and date by the airline in question.

If no flights meeting the traveler’s requirements are displayed on the first availability screen, the agent may request numerous subsequent screens until the optimal flight is located, based upon the CRS’ internal availability logic, these screens are offered in generally descending order of flight desirability.

The several classes of service are indicated by letters accompanying the number of seats available in that class. An individual airline determines what classes will be displayed for each flight. These “classes” are mostly for inventory control purposes, in order that the airline may allocate a given number of seats for discount or promotional fares. There are usually only two physical classes of service on an aircraft – first class and coach class;[73] all other discounts sit in the coach cabin. Some flights add a third “business class” to specific flights, which represents a differentiated physical service on the flight and normally is not open to discounts or promotions[74].

Agents access inventory in what we have described generally as an online, real-timeenvironment. Under such conditions, when seats are reserved, inventory is decremented and the same space cannot be sold to another system user unless a cancellation ensues.

Airline inventory is defined as seats available to be sold on any particular flight segment, as determined by the actual number of seats on-board the aircraft together with inventory management adjustments – upward or downward–that compensate for anticipated conditions such as necessary overbookings due to no-show experience on the route.

When reservations are requested for the CRS supplier or user host, inventory is instantaneously decreased by the required number of seats.

Off-line carriers, those with their own inventory systems that are not part of the CRS host, operate in a sell/no-sell environment. The CRS user may “sell” typically up to four offline seats in any one transaction. The “sale” is recorded in the agent’s AAA until an ET is initiated. A sell message is then transmitted to the computer system where the inventory is managed, and appears in that system in the same way as do reservations made directly in it. When a mathematically predetermined point, (which differs between flights and time periods) is reached, a no-sell condition is created and inventory is closed in the offline CRS system–no further sell messages are accepted from offline systems.

The airline manages no-sell messages to other carriers so as to permit as many bookings as possible, but not so many that the inventory management parameters for that flight are violated- in other words, to limit the possibility that the time lag between the instant a CRS generated sell message is created and the instant that message enters the reservation system where inventory for that flight is stored might cause too many seats to be sold.

Because inventory is usually closed to further sales from other systems prior to the point where all available seats are actually sold, there are some variances between CRS. In other words, SABRE might show space available in economy class on a British Airways flight from London to Stockholm, whereas British Airways’ own BABS CRS (where the airline actually maintains and manages its inventory) shows no space available on the same flight.

In practice, when a discrepancy of this type is identified, BABS would automatically send a message to SABRE instructing that inventory be closed to further economy class sales; given the huge number of airline flights daily, such discrepancies are fairly rare.

The agent reserves flights based upon conversations with the passenger, then enters data necessary to identify the traveler in the AAA:

  • Name
  • Telephone contact
  • Person making the reservation
  • Ticketing arrangements (date the traveler will pick-up the ticket)

This is the required information for any basic PNR. Other information, such as traveler or travel agency addresses may also be entered, as necessary. In SABRE, the agent can enter two addresses: one that will be electronically transmitted to car and hotel companies whose services form part of the PNR, and another that would be used for accounting purposes, if the CRS is interfaced to an external travel agency accounting system. Typically the travel agency enters its own address in the transmitted field and the passenger’s address in the accounting address field.

The agent can also request seating and special meals for travelers. These can be confirmed real-time for the host carrier, but require that electronic messages (usually teletypewriter) be dispatched for off-line carriers. The agent can also display actual seat maps for the host and certain off-line carriers where the required connectivity has been achieved (discussed under Direct Access).

Normally, however, off-line seating would be requested by general location (aisle or window; smoking or non-smoking), rather than by specific seat.

All U.S. CRS have special fields for recording frequent traveler identification numbers. These are marketing programs, sponsored by the major U.S. airlines, which regard participants based upon the number of flights they take. The number must be part of the traveler’s PNR in order for credit to accrue. If an offline carrier is involved, the agent enters the appropriate number which is then transmitted to the reservation system of the airline providing service[75].

The larger U.S. airlines offer their frequent travelers benefits that increase by levels based upon the number of flights taken. This affects the CRS reservation process because some CRS functions are made available only to the most frequent travelers, as identified by their frequent traveler numbers.

United Airline’s Mileage Plus frequent traveler program is the best example. United rewards its most frequent travelers with Premier status within the program. A premier member’s number is identified as such when entered into United’s Apollo CRS. Premier status entitles members to preferred check-in at the airport and special seating on some flights. The CRS recognizes a premier traveler when the number is entered and makes these seats avail­ able. They are usually located in the front of the coach cabin. Thus aisle seating in the forward part of the cabin may be available to United’s best customers–Premiere Mileage Plus members–when that seating is not offered to the public generally.

Car and Hotel Reservations

The agent may also access car and hotel availability systems and make these services available. In SABRE, a car availability screen is requested for the city where a car is desired. It is maintained much as is flight availability–in the sell/no sell environment – unless special connectivity (explained under Direct Access, page 39) is in place. Each car-type offered is represented by a unique coded identifier. The availability display, as shown in the accompanying illustration, is organized to offer a large number of cars from several vendors simultaneously[76].

Car Preferred Customer Programs

Hertz                     Number One

Avis                       Wizard

National                 Emerald Aisle

Budget                   Budget Express

The car transaction identifies the type of car and length of time required. Information is included to identify the agency making the reservation, so that the car company may remit commissions after the rental is complete[77]. Other optional information that can be included with the reservation identifies the traveler for any enhanced services the car company may offer.

The major U.S.-based car rental companies have developed preferred customer databases that reduce customer processing times when renting cars.

When the traveler includes a unique identification number with a reservation, all relevant data pertaining to a car rental is automatically made available to the station where the car is to be rented. This includes driver’s license number, credit card number, car preference, and type of optional insurance desired. The information can be pre-printed on the customer’s rental contract, thereby greatly reducing the amount of information that a rental agent must request from the traveler upon arrival.

This number is entered into a CRS-based car rental reservation, along with numbers identifying the traveler as eligible for any discounts that may be offered.

Since so many potential rate combinations are possible, based upon corporate, organization, or individual discount programs[78], optional insurance, taxes, as well as car type and date, it is extremely difficult to accurately quote car rental rates using a CRS–in advance of a final statement received from the vendor after the rental is complete. While all major CRS contain car rates, these are useful only as illustrations and not as definitive charges the customer will actually pay.

Hotels reservations follow the same basic process, except that there is considerably more differentiation in hotel products, less commonality of rates and room-types, and therefore more descriptive information required. All CRS-based hotel availability and reservation processes use the sell/no-sell methodology no enhanced connectivity has been achieved between any hotel vendor and CRS.

The agent must be careful to include a credit card number or other information (as specified by the hotel in question) to guarantee a reservation, if required under the hotel’s policy. Usually travelers arriving after 6:00pm are required to guarantee their reservations. In the event the traveler does not honor the reservation, a credit card charge for a one night stay would be processed. Reservations may also be guaranteed to an address (agency or business) in some cases, or by advance deposit. This information is accommodated by the CRS and transmitted to the hotel’s reservation system along with the sale message.


Most CRS reservation changes are simply cancel and rebook processes – a reservation segment is canceled and a substitute replaces it. There are no true changes or modifications to reservations already existing without canceling them in PARS-type systems.

The sole exception is when reservations containing multiple passengers require modifications that do not affect all members of the party. In this case a divide function is provided that separates one or more individuals from a reservation so that their file can be changed.

The agent initiates the divide function by specifying the individuals that require changes. A new file, showing only these names, is created in the AAA. Cancel and rebook is then performed on this reservation, as appropriate. Once this is complete, the agent uses a file command (F), which places the reservation in temporary suspense. The original reservation, less the individuals that were separated out, is then displayed. CRS procedures indicate that the agent cross-reference the two reservations using OSI messages, indicating that passengers previously traveling on a single reservation were now separated. Changes affecting only this group of travelers can also be initiated on the reservation, as appropriate.

Once both reservations are complete the ET command is invoked, which completes both files. Messages are then sent to offline airlines that may be involved in the reservation, indicating that the PNR should be divided and changes made. This is an obviously complex process that is somewhat unreliable because all airline reservation systems do not accept the complicated stream of messages necessary to properly divide reservations equally well. For this reason, many CRS users routinely create separate reservations for all passengers, regardless of whether they are traveling together or not, in order to avoid the potential necessity of dividing reservations later.


All major CRS have online training environments that allow simulated reservations to be made. Accessing this environment requires using a specialized sign-on code. Agents using the CRS that have logged-on under this code can access most functions of the system, but any reservations they make cannot be ended by ET – they are automatically “ignored” once completed[79].

The CRS also have online scripted lessons that are used as basic educational tools for the system. Again, specialized sign-in codes are used. Once in the lesson system the agent cannot create individual reservations but can access prescripted files that demonstrate the operation of the system and provide commentary on its correct use. The 1essons must be accesses sequentially and correct responses given before the trainee can move on to following pages[80].

CRS Connectivity

Communication between CRS is critical to the interchange of availability updates and sale messages. This communication must be reliable and quick, and must also support communications between varieties of reservation systems that vary considerably in their sophistication. While by far the majority of airline reservation communications are handles electronically, some smaller local carriers still rely upon voice or paper communications.

The most basic form of electronic CRS communications involves telegraphy. Here, character-based messages are formatted using accepted delimiters and character schemes agreed to by the majority of international airlines. Several types of messages are used depending upon whether reservation, availability, schedule, or other data are involved.

These messages may be transmitted between reservation systems using direct telegraphy circuits that connect two systems to each other or shared telegraphy circuits. There are several shared systems in common use, including ESS, SITA, and ARINC. These are essentially networks that facilitate connectivity between participating airlines. Messages may be addressed to any participant that is connected to the system. This eliminates the need for direct or dedicated circuits, but does not give equivalent performance.

Because shared system messages are slower and somewhat less reliable than are direct circuits, many airlines bypass the common networks and use more advanced system-to-system data communication channels. In recent years these have been built using modern protocols instead of character-based telegraphy. IBM’s SNA[81] and bisync[82]protocols are in common usage, as is a variant of the X.25 protocol.

Another form of enhanced connectivity is so-called “direct access”, which bypasses data messages entirely and uses actual sessions conducted in offline systems. In order to understand how this works, we must first explain how CRS “sessions” are managed.

Concurrent Session Theory

All CRS workstations are connected to the host and have more or less equal access on its resources as needed. They are not physically connected and exchanging data with the host at all times, however. The thousands of terminals supported by a CRS would require tremendous system resource to support if they were all active constantly. Instead, the CRS use a system of virtual connectivity.

When a CRS session is opened in the agent’s AAA the CRS recognizes the “address” of the terminal that is conducting the session. Whenever that terminal transmits data or character-based commands to the CRS it is directed to the same AAA location within the system’s memory. The AAA remains active (subject to an overriding, general time-out limitation) until the transaction is ended. Although the agent’s terminal remains physically connected to the CRS at all times, and although it is logged-in as active, the CRS does not interact with it until the next data transmission or request is received.

Since most terminals are sitting idle on the system at any given moment (even when a transaction is in progress, most of the time required to complete that transaction is idle time- such as while a conversation is taking place with the customer), the CRS can support thousands of terminals that are virtually connected, in that they appear to be connected and active to the user, whereas in reality they are only connected when it is necessary for them to be so.

Direct Access
Major International CRS

SABRE American Airlines
Covia United Airlines
  British Airways
  KLM – Royal Dutch Airlines
  Abacus Partnership
Worldspan Delta Airlines
  Northwest Airlines
  Trans World Airlines
SystemOne Continental Holdings
Gemini Canadian International
  Air Canada
Galileo British Airways
  Austrian Air
  Aer Lingus
Amadeus Air France
  Adria Airways
  Air Inter
  Braathens Safe
  Emirates Airlines
Abacus Cathay Pacific
  Singapore Air
  Malaysian Airways
  China Airlines
  Philippine Airlines
  Covia partners

Direct access uses virtual connectivity to interrelate sessions between CRS. An agent using direct access uses a specialized character string to invoke the process. The CRS recognizes the commands coming from the agent’s terminal and translates them into a form understandable to the system where direct access is in place. It then opens a virtual session in the offline system and processes the agent’s request. This can be a request for availability, a sell message, a fare request, or a DRS request.

The offline system responds to the CRS and includes the address of the agent terminal where the request initiated. The response is then passed directly to the agent’s screen, where action is taken, as indicated. Thus the agent is able to gain the benefit of virtual access into multiple CRS, without the need to learn a variety of commands as would be required if a multi-access system or gateway were used (whereby the agent’s terminal actually became as if it were connected to various CRS, as specified by the agent). Because the database of an individual carrier’s reservation system normally has better data concerning that carrier, the agent can work from a more complete database and provide more accurate information to travelers.

A variant of direct access is used by several CRS to enhance their performance. Sometimes the CRS programmatically opens direct access sessions in offline CRS, without specifically being commanded to do so by the agent, in order gain access to the vendor’s complete availability.

This process is used by some CRS to enhance car reservations also. Here a direct access session is opened in a car vendor’s system. When a sell message is generated, the CRS takes traveler name, phone number, and address data directly from the agent’s AAA, without the need to end the transaction, and submits the sell message to the car vendor’s system[83]. A confirmation number is received and returned real-time to the agent’s AAA, so that a car can be confirmed and assured through a confirmation number immediately.

CRS as Marketing Tools

The principal U.S. CRS vendors have individually invested hundreds of millions of dollars to automate travel agencies. This was done for purely economic reasons, initially unrelated to computer profits. Only recently have the CRS vendors viewed data processing and automation as revenue centers independent of airline marketing benefits accruing to the sponsoring carrier. All the U.S. CRS have created independent computer subsidiaries not only to provide automation services to the airline but also to exploit data processing opportunities within other industry segments as well as in unrelated fields.

Airline Economics and CRS

Incremental volume is critical to any discussion of airline distribution, as in any high fixed-cost business. Airline fixed costs average between 80% and 85% of total costs.

This simple economic principle is important to understanding how drastically incremental passengers affect overall carrier profitability, as this accounts for a range of activities which, taken in isolation, may be perceived to be economically questionable.

The cost/revenue structure is such that relatively small negative passenger shifts result in large losses, most of which cannot be compensated for through cost-cutting. Small positive passenger shifts, so-called incremental passengers gained at the expense of a competitor- produce substantial profits, so much that the carrier may be willing to make significant investments in agency “loyalty” programs, tied to CRS or otherwise.

Screen Bias

Overt manipulation of CRS displays (termed screen “bias”) was practiced for several years, to the benefit of the CRS vendor­ carrier (the “host”) and, to a lesser degree, other carriers willing to pay fees in return for preferential displays (termed at the time “co-hosts”). The existence of bias has been thoroughly documented and described by governmental and independent private researchers. Although precise financial impact is difficult to measure, it is generally accepted that the “have-not” carriers (those without travel agency CRS) lost tens of millions of dollars in incremental passengers to the “Hosts” and “co-hosts”.

The co-host issue was heavily debated during this period. It is generally understood that co-host fees secured general parity in screen display but that overall benefits remained greater for the Host. Each CRS vendor had unique co-host programs that changed from year to year.

Many airlines objected strenuously to the degree of control the CRS vendors exercised over the distribution system, as did some (but not all) travel agency groups.

These culminated in a 1984 U.S. Federal Government regulatory decision that overt CRS bias should be eliminated and that certain standard practices guaranteeing “have-not” airlines generally “free access” to the major CRS networks should be adopted.

Table 1: Estimated Internal Rates of Return (IRR)
Earned by
Airline-Owned Computer Reservation Systems
Contribution to
Airline Earnings
BO% of bias &

halo revenue (a)









1986 108.9% 129.5% 75.2% 137.2%
1992 109.6% 129.7% 91.9% 145.0%
40% of bias &

halo revenue (a)

1984 43.6% 63.9% n/a n/a
1986 53.4% 68.7% 19.8% 54.2%
1992 58.1% 70.5% 42.1% 74.1%
Actual % of tkt revenue(a)(b) 1984 n/a 5.7% n/a n/a
1986 13.3% 22.8% n/a 9.6%
1992 25.5% 31.5% 24.7% 37.8%
Source Computed from historical and projected cash flows and residual asset values reported by the entities in question.
Note n/a indicates that no positive cash flows were reported during the period and thus no return was earned on the investment over that period.
(a) Including indicated percent of vendors’ estimates of CRS contributions to host airline revenues (incremental revenues) from bias and halo effect through 1984, and from the halo effect only after the CAB rules went into effect.
(b) Including same percent of incremental revenues as the percent of domestic passenger revenue actually passed through to airline operating profit during the period of commercial operation for each CRS. These figures are 5.1% Apollo; 5.6% SABRE; 0%/2.5% PARS/Worldspan; 3.2% SystemOne.

Although there continue to be complaints regarding CRS displays and timeliness of system updates, the focus has largely shifted from overt screen favoritism to a more subtle process of agency incentives, market support, and strategic affiliations, which are tied to CRS practices only in part. The principal means of direct CRS involvement in agency dealership maintenance is the so-called Halo Effect by which incremental bookings accrue to the CRS sponsor.

Understanding the Incremental CRS Halo Effect

A travel agent, using any CRS, has a natural tendency to book the flights of the carrier providing the system. This behavior has been well-documented and is routinely accepted as fact by CRS, although the precise mechanism causing incremental bookings to accrue is complex and not dependent upon any single factor.

The travel agent has greatly increased confidence that the information available about the CRS vendor’s services, fares, and availability are accurate, as opposed to those of off-line carriers that may participate in the CRS. Since most agents can cite numerous instances where inaccurate information has created customer dissatisfaction or otherwise adversely affected business, the agent tends to favor a carrier where the best information is available.

This is partly because the nature of airline computers does not encourage timely updates or accurate information. Availability is usually not real-time, with “OK” or “Not OK to sell” status being employed; true seat inventory is not maintained for off-line carriers. Tariff updates are also inconsistent as most CRS use limited sources for these data, such as ATPCO and OAG and not the off-line carrier itself, with changes often lagging severely from the time they become effective until the updated information is available to a CRS user.

Figure 3: CRS Cash Contribution  (Assuming 80% Halo Revenue)

Figure 3: CRS Cash Contribution
(Assuming 80% Halo Revenue)

Part of a successful CRS-based dealership strategy involves consistent contact with the agent, so that the confidence stemming from information access and reliability is reinforced. CRS gives the vendor a cost-effective vehicle to maintain agent contact both on a personal level and through electronic or informational messages.

The computer’s functional range is generally greater respecting the CRS vendor because no computer-to-computer communications are involved and no nuance of interpretation between the command structure of differing machines. Customer requests such as special meal, seating, or boarding passes are easier for the agent to provide.

This is because airline computer communications, whether using ARINC, SITA (ALC or SLC command structures) or through direct communications links (which may employ industry-standard protocols or the same ALC/SLC technology) are typically primitive by current communications standards and do not permit advanced queries or information exchanges. There are several extensive modernization efforts underway, principally by SABRE and Covia, that will enhance communications between airlines as well as between other vendors (such as hotels) also using CRS.

For many reasons agents often feel a sense of loyalty to their computer supplier. This is much like a club or an Alma Mater, a relationship encouraged by most CRS vendors. In essence, many agents believe a CRS vendor somehow deserves booking loyalty when other more immediate factors, such as client preference or cost, do not interfere.

Successfully quantifying the Halo Effect is limited due to the number of factors that may influence agency carrier selection and the inability to isolate CRS placement from among them. Research conducted by the author, on behalf of several U.S. CRS, shows approximately a 10% incremental effect is typical (although some estimates go much higher), namely:

  1. That the agency uses one CRS as its principle reservations booking tool. Incremental volume is far more difficult to realize where bookings are split between two or more systems.[84]
  2. That the incremental effect is to be realized by re placing one CRS with another. Introducing CRS into an agency for the first time brings a wholly different set of booking dynamics into play which may or may not benefit the CRS vendor.
  3. That the CRS being placed in the agency enjoys general functional equivalency with the one it is replacing. It is not necessary, contrary to popular belief, to replace one CRS only with a functionally superior one. Precisely the opposite will be acceptable to the agent, given the right circumstances and a sufficiently lucrative support program apart from purely CRS considerations. Any system will generate incremental bookings provided the basic elements necessary for the agency to do business are present.
  4. That the agency does not book trips exclusively or largely over routes the CRS vendor does not fly. This does not necessarily limit productive CRS conversions to agencies located in the CRS vendor’s on-line cities, as large commercial agencies will do business over many routes.

Figure 4: CRS Cash Contribution  (Assuming 80% of Halo Revenue)

Figure 4: CRS Cash Contribution
(Assuming 80% of Halo Revenue)


There are other important benefits to a CRS vendor, apart from incremental passengers. Under the CRS regulations, vendors are permitted to charge participating carriers for reservations made through the systems. These fees are not dependent upon actual passengers flown.

This creates a financial incentive for the vendor simply to place CRS terminals as widely as possible, irrespective of the Halo Effect. Installing a CRS in a substantial agency in a large market where the CRS vendor does not enjoy a large share of passenger traffic can be more directly profitable (from a data processing perspective) than placing the same system in one of the carrier’s own best agencies.

Booking fees in such a case could amount to hundreds of thousands of dollars annually for the CRS  (and indirectly for its owners), at the direct expense of competitors- general estimates place the benefit of having a travel agent using a particular CRS at approximately U.S. $35,000 per agent reservation terminal, in booking fees alone.


The development of modern CRS and automation-related issues is important because computers are the tangible representation of most carrier dealerships in the United States, and are inseparably tied to the most successful agency management strategies now employed. It would be a mistake to consider, as some carriers have done, that computers define the full extent of the dealership. Automation is but one of several of interdependencies that together create such a strong tie between carrier and agency that the influence exerted over the distribution system greatly exceeds that possible when CRS bias was practiced or under other strategies previously employed.

In manufacturing, the concept of “captive” distributors has been employed since the beginning of industrialization. Airline dealerships are more subtle and, for the carrier, more effective, as they are not exclusive distributors but rather remain representative agents for a number of agents while enfranchised by one primary vendor. This gave dealership-creating carriers the ability to effectively redirect passengers from their competitors, by accessing the competitor’s own distribution network.

Automation serves as the first and most visible contact point between dealership-creating carriers and agencies both by design and by perception. Often these ties are very informal, such as the common travel agent perception that an agency using Apollo is a / “United” agency. The more pervasive relationships are those wherein the CRS vendor has supplemented this foundation with programs and commitments going far beyond perception.

The two largest CRS collectively enfranchise roughly 75% of the market. This position is one of general dominance, rather than numeric superiority only, as this comprises the largest and most important agencies in the United States. The other major systems are almost exclusively comprised of “secondary” agencies.

While not lessening the general value of maintaining a CRS network, the composition of that network greatly enhances both the incremental revenue available through dealership creation and the amount of gross booking fees available. This is evidenced by the intense competition for system “conversions” particularly between the major CRS.

Understanding Agency Segmentation

Agencies may be grouped into several broad categories, almost exclusively characterized by size but where this is not necessarily the only index of business efficiency. Size is somewhat reflective overall sophistication, as larger agencies are assumed to be better managed (although there are frequent and notable exceptions), and also comprise the general parameters within which airlines define the agency’s relative importance to a dealership strategy.

These segmentations may be drawn using U.S. Dollar ranges as illustrated in the figure below. The actual number of locations within each segmentation, based upon 1985 ARC data, is shown below. These data indicate numbers of true ownership entities, as distinct from locations; thus one national agency system with hundreds of outlets would appear only once in the tabulation. This methodology is based upon the premise that common ownership denotes common intent and management, which are key considerations for a dealership program.

Table 2: ARC Travel Agency Statistics
($ MM)
% Entity
Size ($ M)
($ M)
% Total
0 – 5 20,427 96.50 927 18,945,198 62.31
5 – 15 587 2.77 7,918 4,647,731 15.29
15 – 40 111 0.52 22,295 2,474,778 8.14
40+ 42 0.20 103,247 4,336,354 14.26
Source: 1985 Airline Reporting Corporation (ARC) Data Tabulations
Note: 1985 is the latest year data are available for this publication. Later years would show some concentration in the 40+ MM range, fewer entities in the 15-40 MM range, and more entities in the lower two ranges, especially the 0 – 5 MM. Otherwise, the chart continues to illustrate the principles discussed, which remain applicable.

Airline dealership and distribution strategies are greatly complicated by the fact that standard motivational and control methods do not apply. Not only do agency dynamics not lend themselves to most traditional structures, but, as the figure illustrates, no 80%/20% seller-volume ratio, common to so many industries, applies.

By far the majority of agencies (96.50%) generate the bulk of agency-based ticket sales in the U.S. (62.31%).

The development of CRS has signified a developmental condition far more significant than computer technology alone. Although technology is the focal point, and there are certain rationale for maintaining access and domination over as large a block of data as possible for planning and intelligence-gathering purposes, major carrier marketing strategies are definitively stratified in their approach to agency marketing.

This permits the carrier to dominate and reap incremental benefits not only from the broad mass of smaller agencies through the “halo” effects associated with CRS, but also to exercise tighter and more specific control over key distributors in major markets.

A critical part of this strategy is the overall marketing and competitive profile of the CRS vendor. Dealership strategies, particularly those focusing upon incremental benefits that accrue from many agencies, are successful because the five CRS vendors individually constitute the largest carriers in the United States as well as the most aggressive marketers. They are impossible to avoid, should an agency want to sell away from them, represent major presences in most markets, even those where they are not dominant.

It is strategically acceptable for an agency to align itself with a any of these major vendors (although some are preferable to others under given circumstances), because of their size and economic power, whereas smaller carriers could not offer similar benefits in any case. This would be true irrespective of technology in any form. The financial and market-support programs available through association with these major airlines are too great to be disregarded.

Major market presence is not the only ingredient to a successful dealership program. Successfully motivating and controlling critical agencies or groups of agencies within the market is far more significant. The extent to which this control may be exercised is limited only by individual market conditions and is often extensive in scope.

The task is also not as difficult as would initially appear. Since agency behavior can be categorized for marketing purposes within general size ranges, and the small number of exceptions dealt with individually, programs may be structured that are attractive to many agencies and are also financially effective for the carrier as well as strategically efficient to administer and control.

Current Agency Dealership Programs

By targeting major agencies (those positioned to advance significant numbers of incremental passengers) the dealership-­creating carrier is able to motivate vendor preference and success of other program elements to a surprising degree. The vendor’s objective is to extend as much control as is possible over the “accepted” distribution process, thereby gaining preferential treatment at the expense of competitors.

This provides point-of-sale leverage without the risk of direct distribution channel ownership and associated expense. Assuming the agency’s commitment is sufficient, ownership becomes irrelevant while all the benefits of a “captive” distributor are realized.

The agency’s general sales stream may therefore be tapped without inviting overt retaliation by competitors who might refuse to deal with a vendor-exclusive distribution system.[85]

Even so, dealerships, as they now exist, are a phase of an evolutionary process that will see distribution control migrate fully from an agency to a vendor-centered focus. Although precise timing varies with individual carriers, with “have-not” airlines lagging significantly behind the more technologically sophisticated CRS vendors, dealerships in their present form are not long-term phenomena.

Understanding Travel Agency Economics

Many travel agencies are financially precarious at best. Inflexibility in adapting to the needs of a deregulated environment, where price instability and rising costs coexist, has seriously weakened many and brought some to the point of bankruptcy. A well-financed carrier, where air sales will account for 60% or more of the agency’s total volume, can offer a dealership relationship that is both pervasive and irresistible.

Through front and backroom automation, the carrier can create dependency in all ways and reinforce this position wherever possible. Long-term CRS and other automation contracts, now usually specifying minimum usage levels, lock agencies in for years, often with large “liquidated damage” clauses that impose penalties if a systems conversion were undertaken, to the degree revenue would be lost by the vendor. Dependency is also fostered through financial measures tied to bookings or other incentives. Instability in some large agencies and rapid growth in others often creates a cash crisis which the agency is unable to correct.

Although the entrepreneurial nature of most travel agencies permits reasonable adaptation to some growth, many are overextended by debt incurred through acquisitions and unnecessary operational inefficiencies. Others have lost business to increased competition and consolidation of their major accounts with larger national agencies.

Business contraction is contrary to the entrepreneurial tendency of most small businesses and requires specific skills most lack if it is to be done effectively. The agency is therefore greatly receptive to financial propositions that will relieve its economic difficulties, if only temporarily. Many dealership initiatives capitalize on this agency “cash crisis” and create such a financial dependency that the agency may have no choice but to support its sponsor (many, if not most, large agencies have costs in excess of their basic per-ticket revenue (less overrides or “incentive” commissions) and are therefore unprofitable without some vendor support apart from normal commissions.

Financial support is provided in two direct and many indirect manners, each fulfilling specific needs. Dealers receive override commissions which, although not unknown to all agencies, are far more liberal for the dealer.

While these can be nothing more than percentage payments on gross sales volume, the more sophisticated dealership-creators structure complex reward matrixes wherein compensation is tied to exceeding a minimum production target that is driven by the carrier’s overall share of market as well as the respective shares of market represented by all dealers in the market.

This makes extraordinary commitment sometimes necessary in order for the agency to realize any substantive payment. Agencies are often sufficiently dependent upon their sponsor that they will redirect passengers using whatever means are necessary in order to achieve their revenue targets. Sometimes this includes representing that a competing carrier’s flights are sold out or more expensive when they really are not.

Good market intelligence and planning, sometimes evident in dealership creation, permits a high degree of dependency to be built without sufficient support being offered so that the strength of the carrier relationship could be lessened. This is especially true when considering outright agency cash payments, usually associated with CRS contracts, which infuse large sums into the business but do nothing to improve its overall position.

Frequently a CRS sponsor will pay a large agency hundreds of thousands of dollars to change system affiliation and sign a restrictive long-term contract. While occasionally these payments accompany a long-term dealership commitments and sales support programs, often the carrier has no long-term interest in the success of the agency, as a competing agency will fulfill the same purpose once the initial payment is recovered.

Agencies assume “golden handcuffs” provided by their sponsors to a degree not experienced in most industries. If financial and market support is forthcoming, which often results in the vendor making one agency more successful at the direct expense of competing agencies in the same market (which are also “agents” for the carrier), the agent may enter into the restrictive CRS contracts previously described, undertake various other covenants regarding how the business will be operated and how business will be conducted, and forego relationships with other vendors on even the most casual of terms. Most agencies fail to recognize or accept the implications of airline-type dealerships by continuing to assert their independence and customer orientation while accepting the most restrictive of business environments.

Role of CRS In Dealership Maintenance

CRS permits the dealership-creator to implicitly encourage redirection of agency bookings from dealers both formal and informal. Essentially any agency using the system will be partly influenced by the “halo effect” and can be relied upon to produce passengers disproportionate to the carrier’s “true” share of the agency’s business, because of CRS. Other elements of dealership strategies serve to strengthen and accentuate this phenomenon.

Overt screen manipulation, although alleged still to exist by many, need not be a fundamental part of the incremental booking process. With the opportunity to affect agency carrier selection in so many powerful ways, and with computer services profitable ventures overall in their own right, improper display tactics make little sense for a sophisticated vendor.

The most successful agency distribution programs, contrary to the great body of academic speculation, are not currently technology-driven. Technology is the cohesive element that ties the relationship between carrier and agency together and creates the “halo effect” by its existence, but CRS could not, in isolation, create the dominant market positions enjoyed by the most successful vendors. Only a combination of well-designed programs, together with an airline that is naturally dominant by virtue of its structure, can occupy the preeminent positions enjoyed by the CRS vendors in the U.S.

Internationally, where most of the CRS vendors have strategic expansion objects, this presents both a challenge and an opportunity for competing carriers:

  • Technology creates an effective competitive barrier because it is time-consuming, difficult, and expensive to replicate.

Given sufficient resources, however, technological advantage will be insufficient to unseat many European or Asian carriers from the preeminent market positions they enjoy, if effective steps are taken in the near-term to compete directly in a dealership environment.

As the foregoing ARC statistical illustrations show, dealership maintenance in the United States must be accomplished on two levels:

  • The largest portion of the market may be influenced indirectly, using the “halo effect,” and through CRS as a principal contact point between carrier and agency.

The size of these agencies (and therefore their overall sophistication) precludes the carrier from an economic strong dealership program, while limiting their ability to affect carrier market share positively or negatively in any dramatic fashion. The best and most cost-effective programs for these agencies are probably those informal preference strategies based upon CRS placement.

  • Segmentation of the market permits the vendor to effectively address the larger and more sophisticated agencies with more advanced programs.

As only a few businesses are involved, which collectively produce significant volume, these can be approached and managed individually, with CRS serving as one tool among many.

Dealership-creators also approach agencies within this group differently, often offering better programs to regionally-dominant agencies with a significant but finite market presence rather than to so-called nationals, which cannot concentrate their sales or management efforts in specific areas to materially impact the vendor’s business.

Strong Dealership Agency Type Classifications

National Agency Systems are businesses wherein a common owner­ ship and management structure controls outlets in numerous (sometimes hundreds) of locations. These well-known companies collectively process significant portions of a vendor’s business, but usually (with some exceptions) do not dominate specific city markets. This creates booking fragmentation within the national, its principal point of weakness, which, together with often unfocused and inconsistent management, renders it unable to significantly impact most vendors positively or negatively.

Further, because the national has interests in a number of markets, one overall vendor preference strategy or dealership alignment may be impractical. In order to grant support in areas vital to the agency, the vendor may insist upon concessions in other areas where the agency may enjoy leverage contrary to the vendor’s best interests. When considering a large-scale strategic alignment, the agency’s negotiating position if often weakened beyond repair due specifically to size and coverage, not strengthened by it as many believe.

Regional Dominant Agencies control a large market share within a geographic area, usually to the extent that they represent the most significant booking source and distribution vehicle within the area and can materially influence the thin margin of profitability realized by any vendor for the market. They are often among the better managed agencies (a characteristic arising partly as a result of their regional focus and partly responsible for their success in implementing a strategy based upon that focus), and enjoy far more leverage with any single vendor (CRS or otherwise) than any other agencies.

Large Local Agencies have grown usually through business development (a large-scale acquisitions strategy is usually implemented regionally), and are typically aggressive entrepreneurs. They build business by competing successfully within a small geographic market and by capturing accounts from national and other local competitors wherever possible. Among the most attractive of acquisition candidates, they cannot command enough market leverage to influence carrier margins and are employed opportunistically by dealership-creating carriers because they comprise blocks of volume that can be influenced through general dealership management techniques.

Future of Airline Dealerships

Under regulation, competition within the distribution system was circumscribed by legal mandate, which guaranteed market orderliness irrespective of specific carrier intervention. Deregulation permitted some vendors to create CRS which were difficult for others to emulate, and seize control from uncertainty. Many carriers still have not grasped the implications of their reliance upon a distribution system dominated by a few principal carriers – irrespective of the role CRS plays in the process. Agencies have formed alignments with carriers from competitive necessity, as others in their markets would otherwise receive benefits they would not enjoy, and also because of financial imperatives, many of which were created because of deregulation.

Dealership-creating carriers will continue to strengthen their dominance over agencies accepting dealership status through the means previously described. On an opportunistic basis, agency ownership will be explored, over a period of years. In general, however, dealerships render ownership meaningless for vendors able to support a dealership strategy.

As systems become available, carriers will begin to actively circumvent the agency channel. With implied agency retaliation minimized through domination of the distribution channel, and with the support of complex computer systems, vehicles such as automated ticket machines, cooperation with other electronic vendors, and so-called videotext services (in a form greatly expanded from that known today in the U.S. and modified technologically to support airline-specific needs), agency services will become less cost­ effective for certain types of tickets.

With reference to commercial or business travel, the agency serves few purposes apart from order-taker and document-preparer. A large carrier reservation center can process orders for only a few dollars each, far less than agency commissions. Technology will provide more opportunities to distribute documents (among the most effectively automated of functions and already highly automated within the travel industry) directly to the customer, thus avoiding the agency as an intermediary. Future developments may render tickets per se, and therefore the ticket-issuer, unnecessary – a concept a1 ready tried experimentally by some carriers.

Although agency services of some type will always be desirable, for certain transactions agency processing is already superfluous. The essence of dealership creation, that being control of the distribution channel by a vendor, provides the carrier with significant opportunities to eliminate the dealer where appropriate- something the agency can do little to mitigate.

Non-dealership carriers will find dependence upon the dealership- controlled distribution increasingly unacceptable, driving them much more rapidly toward non-agency distribution systems wherever possible.

While many of these will be technology-driven, “direct dealing” (discounting to major purchasers) will be implemented widely within the airline industry first from within this group. Dealership-creators prefer to use the dealership system to pass along customer discounts for a variety of strategic reasons, among them the desire to mask discounting (and therefore minimize it) as much as possible. Without the dealership option, other carriers will prefer to lock-in distribution using this method that the dealership-creators will be reluctant to employ.

Non-U.S. airlines may choose to use their own market leverage to create dealerships they are positioned to support and thus circumvent not only intrusion by aggressive U.S. CRS vendors but also influences by other local or regional competitors.

The dea1ershi p strategy has been shown to be effective in achieving most carrier marketing goals using systems currently in place and, in many cases, a distribution system that is readily receptive to this type of support.

The significant point of this discussion is that CRS is neither the key to dealership activity nor the insurmountable obstacle it is frequently portrayed to be. On the other hand, non-U.S. airlines must avoid the Trojan Horse of relinquishing data control to successful U.S. CRS vendors having internal agendas that have always included domination of distribution, tied to computerization.

The evolutionary processes set in motion by deregulation in the U.S. have shown that non-agency avenues must eventually be explored, and that systems do comprise the lever that will make this practical. Dealerships are one phase of market control but cannot be considered an end in themselves, just as CRS contributes to distribution options but does not constitute their fullest extent. Distribution control can be managed, and dependence upon the systems and programs of others avoided, but only if appropriate pro-active steps are taken before an unbreakable wall of agency dependence is built.

Works Cited

(1989, February 4). The Economist , p. 16.

(1989, February). Software Magazine , 9, p. 57.

Buckley, N. (1989). private communication. TWA .

Desmond, J. (1989, February). Max Hopper Knows Sabre’s Guts. Software Magazine , p. 57.

Gilford, D., & Spector, A. (1984, July). The TWA Reservation System. CACM , 27.

Kumar, V. (1990, January). Current Trends in Transaction Processing Systems. Journal of Systems Management , 41, p. 33.

Wardell, D. J. (1982, July 12). The Right Decision for the Right Reason. The Travel Agent, p. 70.


[1] Pricing in this sense refers to the application of fares, as contained in a fare database and as specified by a rule database to specific itineraries as reserved by a traveler. Thus a travel agent will make a reservation and price the itinerary–yielding a total fare, or price, for all segments within the passenger’s itinerary, when considered as a whole.

This is significant as airline reservations are frequently flight or itinerary-specific. For example, fares between desti­nations A and B maybe lower if the passenger travels round-trip, if a certain flight is used that the airlines desires to promote, or if flights with intermediate stops (as opposed to non-stop flights) are used.

Fare databases are used to quote rates based upon passenger inquiries, but only a complete itinerary price is used to issue a ticket and is guaranteed to the traveler.

[2] Dealerships as understood herein are distributors that represent the primary interests of a single principal or small group of principals (vendors) to the overall detriment of other vendors. The dealership may or may not effectively meet the needs of its customers, depending upon the precise nature of its business practices.

Dealerships are typically thought of as automobile distributors or appliance stores. Travel dealerships are looser are are rarely, if ever, identified as such publicly. The airline vendor’s relationship with its dealers allows significant benefits to accrue to the enfranchising principal, regardless of how the relationship is described.

[3] KLM is the world’s oldest airline still operating under its own name, and the Amsterdam – London route is the world’s oldest air route.

[4] Aviation “firsts” are always subjects of considerable discussion. While some airlines put their beginnings prior to 1920, they carried mail, and not passengers. May 17, 1920 is used by most sources as the beginning of scheduled, passenger transportation.

[5] Fares are charges that apply for passengers between a specific origin and destination point, using a specific class of service and a particular routing, as determined by the carrier.

In earlier times, fares were tied to distances traveled more closely than they are today. Now, promotional and yield management considerations make it very difficult to project or establish fares based upon distance alone.  Outside the U.S., airlines often operate in regulated environments, where prices are fixed by international agreements or government decision, rather than by competition.

This occasionally results in considerable consternation on the part of travelers and consumer groups, as these fare and pricing inequities are difficult to justify when examined outside the larger airline pricing context.

Two of the world’s busiest air routes – London to Paris, and New York to Washington, DC- are both around 220 miles. An unrestricted London-Paris ticket as of this writing (Fall, 1990) is £88 ($172), whether the customer flies British Airways or Air France, whereas the price of a New York­ Washington shuttle ticket is $129 (The Economist, 1989).

[6] Some travel agencies perform a financial function, in that they organize payments on behalf of business travel purchasers and issue consolidated statements and invoices for tickets purchased. This can also be a credit or cash management function, if the travel agency does not require either payment on delivery or payment through a charge card with ticket issuance.

In either case, this is a minor part of any value-added by the travel agency to its business travel customers.

[7] Although “hard-copy” reservations management is more common in the hotel industry, many substantive airline operations, operate in a primarily manual mode. These often use CRS inventory displays for communicating sales made to a central facility, where the actual database resides in manual form. This is one illustration of the absolute necessity of CRS as distribution tools, if not inventory management tools.

[8] BOAC (later consolidated with British Airways) began transatlantic service from New York to London October 4, 1958. National Airlines (later purchased by Pan American World Airways) began domestic U.S. service from New York to Miami December 10, 1958.

[9] Telegraphy has a long history in the travel industry and is still an integral part of data communications for many suppliers (as discussed later in this chapter). The term telegraphy describes a telecommunication system whereby graphic symbols (using a small character set of letters, numerals, and a few special control characters) are transmitted using a signal code that describes each character so transmitted individually.

Frequently the 5-bit Baudot encoding system (from which the term “baud” is derived) is used, particularly in print-based teletypewriter machines.  Data processing (non-print media) telegraphy can use any number of encoding systems, the essential quality being representation of discrete characters by specific codes without intermediate states or gradation.

Teletype is the trademark of the Teletype Corporation for their teletypewriter devices.

[10] (Software Magazine, 1989)

[11] These included electrostatic storage, which was insufficiently reliable for most applications; mercury delay line storage, a very slow (on the order of 500ms, or about 50 times greater than 1990-era storage media) system used only by UNIVAC; magnetic drum storage, widely used but also slow; and magnetic core, the most successful process for both commercial and scientific applications.

[12] During these years, LSI or Large-Scale Integration devices were produced. These are similar to the VLSI or Very Large-Scale Integration devices (broadly defined as a device containing more than 10,000 transistors on a single circuit) used in today’s circuits, but on a less sophisticated level.

[13] It is interesting to note that integrated circuitry, although pioneered in the late 1950s and commercially available as early as 1961, was helped immeasurably on the road to wide acceptance and practicality by the early U.S. space program.

[14] International Business Machines, then, as today, manufacturers of the most successful, highly reliable, commercial data processing devices and operating systems.

[15] Now merged with Burroughs to form Unisys.

[16] Semi-Automated Business Research Environment

[17] (Software Magazine, op. cit.)

[18] Eastern PARS is distinct from the proprietary airline CRS developed and market by TWA, and which is now part of Worldspan, in partnership with Northwest Airlines and Delta Airlines. Although the platform used by TWA developed from IBM/Eastern PARS, the CRS is a distinct product with its own pedigree.

[19] Trans World Airlines (TWA) has a similar development project with Burroughs, and United Airlines worked with Univac (both Burroughs and Univac are now part of Unisys). Although Univac did finally implement a somewhat successful reservation system, which was used into the 1980s by airlines such as Northwest Orient, the initial United Airlines effort was abandoned.

[20] The era where long distance telephone communication was routine really did not arrive until the 1970s. Although nationwide direct dial service had existed for many years previously, line conditions were poor and rates high by today’s standards. The public generally limited long distance telephone conversations to unusual circumstances. Although reliance upon the telephone was more common in business, written correspondence still played a more prominent role in the conduct of long distance business than is true today.

The extensive commoditization of telephone communications was brought about by the gradual deregulation of long distance services, and subsequent lowering of rates, beginning in the early

1980s, and the introduction of more advanced telecommunication technology, such as facsimile machines (FAX) in the mid 1980s.

[21] Interline refers to tickets (or ticket-related transactions such as reservations) issued by one airline on beha1f of another. This is done because airline schedules often require coordination with other carriers in order to fulfill customer needs. Thus, a traveler flying from Portland, Oregon to London, UK must change planes at some intermediate city because there is no direct service.

The traveler’s itinerary might involve United Airlines between Portland and Seattle, Washington, then British Airways from Seattle, non-stop to London. Only a single airline ticket would be issued that would contain two valid flight coupons–one  for the United flight and another for British Airways. The traveler would also probably pay a single price, which would be divided between the two airlines based upon formulas agreed to by the airlines involved. The passenger’s reservation and ticket could be issued by either United Airlines or British Airways, because of the interline agreements in place between them.

The overall effect is to simplify the transaction for the traveler and any travel agent that would issue the ticket on behalf of the airlines involved.

[22] Telephoning is very common in the travel industry, as some airlines do not make reliable inventory information available to “offline” systems; others are not automated at all. Telephone calls for reservations routinely take several minutes at best, and can take much longer, if the reservation center called is busy or telephone communications are unreliable. This contrasts with electronic inventory sale messages which the operator completes in a few seconds.

[23] These are detailed in the publication: Reservations Interline Message Procedures — Passenger (ARIMP), current edition, Montreal: International Air Transport Association.

[24] Univac played a large role in this development.

[25] Partly this was doubtless because a comparatively “low-tech” approach to meeting the communications needs of the embryonic worldwide airline systems industry, where all phases of communications technology were underdeveloped and not completely reliable, was deemed more appropriate.

It is also partly true that the right applications and designs were simply not made available to the right airline system designers at the right time.

[26] There are numerous other programmer tools that make the production of machine instructions easier and more reliable. Most rely upon close interaction with a processor of some type, so that the programmer gains real experience as the work progresses.

An illustration is Microsoft’s excellent PC-based Quick-BASIC programming environment.  Here programmers use an online screen editor to create their code, assisted by online help files and self-editing features that identify potential errors as they are made. Programs, parts of programs, or even individual steps can be tested at will through an emulation feature that avoids the “compile” step and demonstrates what the final product will actually do on the host computer.

[27] To illustrate, the first computer the author, used was an IBM 360/20 that had 48k of magnetic core memory and very few peripheral devices. Contrast this with today’s PC which is routinely equipped with at least 640k of semiconductor memory, a CRT, and numerous peripherals.

[28] Sometimes, depending upon the type of assembly language used, a program called an assembler is used to translate certain parts of the program into a form suitable for the computer.

[29] Fortran (1958), Cobol (1960), and Algol (1960) are examples of several that are still used (with ongoing revisions and refinements) today.

[30] This, in part, accounts for American’s decision to migrate to a “pure” TPF environment–a project undertaken with significant expense and risk. SABRE’s access to product enhancements and IBM support is thereby significantly enhanced. Further, the programming upgrades introduced into TPF promise to make it a much more productive environment. These advantages were not available to hybrid systems.

[31] Native Mode usually refers to a computer operating environment that is specific to a particular type of machine, within which applications can be run without the need for interpreters or specialized operating environments that are foreign to the machine in question.

[32] Uniprocessing means that instructions are executed by the processor sequentially, as opposed to multiprocessing, where several processors function in parallel, executing coordinated instructions.

[33] Although American’s implementation of mainstream technology is a significant and correct strategic step, it is debatable whether the new environment provides as much high-volume transaction processing flexibility and capability as did American’s hybrid ACP/TPF system

While a detailed discussion of this question is beyond the scope of this text, it is well to remember that American’s decision was a strategic move, undertaken for reasons apart from pure transaction processing capability.

[34] Per Vijay Kumar (Kumar, 1990). Vijay Kumar is an assistant professor of computer science at the University of Missouri–Kansas City. His research areas are distributed systems, object oriented and expert database systems. Copyright Association for Systems Management 1990, used by permission. This material is based upon (Gilford & Spector, 1984) and (Buckley, 1989).

[35] Most CRS highly subsidize user (travel agency in the main) hardware and software. Rea1 CRS revenue is obtained through transaction fees charged to vendors. There is a complex relationship management challenge facing not only the CRS but others who would use the CRS as a value-added network. This is a fundamental premise that is usually overlooked by would-be CRS application developers.

[36] By far the majority of transaction-based systems produced today, regardless, of application, are online systems.

[37] Essentially meaning the time a transaction takes from its initiation to completion–including manual and computerized steps and processes.

[38] In this sense, an individual user’s interaction with the system in order to accomplish a specific purpose (such as make a reservation for a passenger) is termed a session. A session might include dozens of transactions and database queries, all of which would come together in the completed passenger’s reservation.

[39] A real-time application delivers a response during the time that the application is physically running on the computer, so that the process can receive interaction from the user and be directed accordingly. In airline reservation applications, for instance, requesting flight availability is an interactive, real­ time application because the user determines specific parameters and modifiers that are used to deliver “correct” flights, making the decisions as to what modifiers to use while interacting with the system.

[40] There are CRS databases that are used for internal MIS or other purposes not essential to system operation. These may not have continuous availability.

[41] A good example is the electronic filing of passenger reservations that have been coded to be ticketed on a certain date. This typically occurs overnight, when system usage is at a minimum.

[42] 1.5 seconds is the standard for most responses using SABRE. In other words, a user receives a response to any entry made within this time. The sole exception is the complex “best fare” pricing program, designed to locate the lowest air fare for a particular itinerary, which can take approximately 5 seconds.

Users are very conscious of system responsiveness and routinely complain if even a few seconds are added to familiar response times.

[43] This is part of the reason assembly language programming, which is intrinsically faster than many (not all) high-level compiled object programs, is used extensively in airline CRS.

[44] This means that there may be inaccuracies in the database, but these must not be planned or intentional. The system cannot assume that some inaccuracy or old data is tolerable.

[45] CRS do contain “non-essential” data that do not fit this definition. For instance, there are text-based reference pages in every PARS-type CRS that contain general information of definite value to system users but not essential to, or directly involved with, transaction creation or update.

[46] “The Bank of America plans to use TPF as the basis for a high-volume transaction processor front-end using IMS as the back-end system, and is also planning to enhance the power and capability of TPF.”

(Kumar, 1990). Note: IMS or Information Management System is an IBM database support product.)

In this same article, the author expresses the opinion, “In its present form, however, it (TPF) does not offer a solution to transaction processing problems.” It is correct that TPF does not offer a comprehensive solution to the range of difficulties associated with transaction processing, and its longevity (at least in present forms) may be questioned, but the large, high-volume, successful TPF operations in place today are sufficient testimony of its value as a transaction processing tool.

[47] The Civil Aeronautics Board, a U.S. federal governmental airline regulatory body dissolved as part of the Airline Deregulation Act of 1978. Many of its functions are now assumed by the Department of Transportation; others by the Justice Department.

[48] The Association of Retail Travel Agents, an industry group that limits its membership to businesses actively involved with the sale of travel to the public at a retail level.

[49] In the same talk, Cohen stated:

“The technical capabilities of today’s computerized systems reflects the old regulated industry and they (the airlines/CRS) must adjust to the new, competitive regime.”

What, at the time, was mistaken for technological limitations was really an unavoidable consequence of deregulation. This is so because the most challenging CRS consequence of deregulation is database maintenance, which is hampered by significant, non­technological, limitations. There have been significant pricing and fare-search tools developed for the major CRS, but the fundamental database problem remains and has not changed significantly, although the trave1 agent’s and the public’s ability to deal with the problem have improved greatly.

[50] This is because the transaction load experienced by today’s CRS, based upon their broad travel agency user bases, could not be accommodated by systems of that vintage.

[51] Excerpted from a four-part series, The Right Decision for the Right Reason(Wardell, 1982), published in The Travel Agent. Copyright 1982 by David J. Wardell. All rights reserved.

[52] This is costs refers specifically to airline City Ticket Offices (CTOs), which are storefront 1ocations handling ticket sales much as a retail travel agency does. Airport processing costs were somewhat less. Ticket-by-mail programs, never universally adopted by airlines, have consistently represented the most cost-effective distribution mechanism, but cannot meet the needs of high-volume business accounts and are also unattractive as some risk must be assumed where tickets are sent through the mail.

[53] The federal transportation tax is currently 10%, making the airline’s effective commission rate 9%.

[54] Air Traffic Conference of America. An airline industry cartel that set standards (both commercial and technological) for relationships between airlines and travel agencies, and policed these relationships.

[55] Radio Corporation of America, a pioneer in the electronics industry; merged with General Electric in 1986.

[56] Which was minimal because the teletypewriter machines used by RCA were extremely rugged and reliable; later machines from other suppliers had more problems.

[57] Airlines sponsoring teleticket machines varied based upon the carriers represented in various markets. Where one airline was stronger than others it usually took responsibility for installing the machines in select agencies.

[58] The Trans World Airlines and United Airlines reservation system experiments were discontinued in 1973, while American Airlines continued to maintain equipment in some large agencies.

[59] AMTRAK schedules, availability, reservations, and ticketing are offered by the major U.S. CRS. There are projects underway to make similar services available for European rail services and ferries, but none has thus far been finalized.

According to John Desmond  (Desmond, 1989), Writing in Software Magazine and quoting Max Hopper, Senior Vice President for information systems at American Airlines:

“The effort to include the trains and ferries of Europe into the reservation systems is a massive, continuing project. But it is essential to compete in Europe, because so much travel in that market uses multiple modes. He (Hopper) indicated the situation is analogous to the 1960s in the U.S., when airline transfers prolonged the minimum two hours necessary to make a reservation.”

[60] Indeed, the concept of a travel agency CRS was not well-defined at that time.

[61] Multi-Access Airline Reservation System. Multi-access is discussed in more detail in the connectivity section of this chapter.

[62] Multi-Access Reservation System.

[63] SystemOne Direct Access.

[64] It is interesting to note that some of the first CRS installed nationally were placed in commercial accounts (corporations with large travel volumes and in-house travel arrangers), rather than in travel agencies. This was also true for many of the first teleticket machines.

[65] One way this is said to be done is through manipulation of display algorithms and flight times.

[66] This fact is clearly demonstrable in medium and larger­sized computer installations. A complete power-down for many large computer centers would create problems of near disaster proportions. The effect is far less clear with small systems. In the authors opinion, regular power-downs of small computers is not harmful and probably prolongs equipment life for two main reasons: (1) Small computer mechanical devices, such as hard disk drives in particular, are far less robust than those designed for larger systems and suffer due to extended wear; (2) Office (and home) power conditions are not nearly as well conditioned as are large data centers, which makes office systems much more vulnerable to power surges and other anomalies if they are left running unattended for long periods.

[67] SABRE uses 4-character PCCs; other CRS use three.

[68] Termed “The Triple A.”

[69] “Simultaneous changes to PNR.”

[70] Passengers may be unhappy if reservations during peak periods of demand are canceled, for instance. A complete “history” file makes it possible for the agent to inform the passenger of the date, time, and individual who authorized the change, thus relieving the airline or other CRS user from responsibility for unsatisfactory reservation changes.

[71] These displays are updated frequently–usually several times daily. Some CRS have weather information systems that operate independently from DRS.

[72] More than 4 seats could be “requested.” This means that the agent could initiate the transaction and wait for a teletype­ writer message from the airline in question to verify that the number of seats required had been reserved. In such cases, reservation confirmation is not automatic.

[73] Outside the U.S., coach class is referred to as economy class.

[74] Business Class is a compromise between first class and coach or economy class. It lacks the full set of in-flight amenities found in first class, but is still a grade above coach. It is intended to appear to business travelers where budgets may not tolerate first class fares, but where a comparatively modest fare differential for business class may be acceptable.

Business class seating is very popular on international flights, where long distances make the extra amenities attractive.

[75] This process is fairly imperfect at present and, even where properly formatted and entered, frequent traveler number transmission is notoriously flawed.

[76] This reflects the commodity nature of the car rental product, as perceived by the CRS. In other words, one car is pretty much like another, there is little substantive information that can be passed by the CRS to the user about specific car-types, so the important information becomes type and availability only.

[77] Agencies issue airline tickets and deduct their commissions when tickets are issued. Car rentals are not generally pre­ paid, hence this system is impractical and commissions must be remitted once final charges have been established.

Commission payments are a major source of travel frustration, as many vendors are extremely careless about and accurate commission payment, while others simply ignore agency commissions due entirely.

[78] These programs are quite pervasive in the industry, to the effect that most car rental business is covered by a discount program of some sort.

[79] The main limitation of the training environment for travel agents is that tickets cannot be printed unless a completed PNR is on file. Reservations created by the training environment cannot be used to simulate ticketing because they never leave the AAA.

[80] U.S. airline CRS were among the first extensive users of online simulation and training systems.

[81] Systems Network Architecture. The IBM data communication standard, widely implemented in the computer industry, which describes the relationship between IBM’s virtual telecommunication access method and the network control program.

[82] Bisync or Bisynchronous transmission is a standard used for controlling synchronous data transmissions. The Bisync standard specifies message format and line protocol.

[83] This process obviously requires that complete passenger name and other data be present in the AAA before a car reservation is initiated. If these data are not present the direct access connection fails and the system reverts to standard communication methods.

[84] Some travel agencies elect to use two (sometimes more) CRS within their organizations in an attempt to preserve optimal relationships with several airline marketing organizations. While the commercial reasons for such a course can be appreciated, the agency suffers obvious lapses in productivity and management control (which most can ill-afford), while the CRS receives far less benefit from placing the system that would otherwise be the case.

[85] No agency could operate in the United States without the ability to represent all carriers.


This document may not be copied or reproduced either photographically, photomechanically, or electronically. It may also not be quoted or referenced in other documents, presentations, studies, papers, or symposia without express, prior, written consent.