"A FAST Way to Define System Requirements" | Gary Rush, IAF CPF

Article Published in Computerworld, October 1985

FAST logo

"A FAST Way to Define System Requirements", by Gary Rush, Computerworld, Volume 19 Number 40, In Depth pages ID/11 to ID/16 (pages 47 to 52), October 7, 1985.*

MIS has long applied productivity techniques to the problems of coding and testing computer application. Now DP planners also are trying to develop better application requirements and design specifications as a way to increase productivity.

Over the past eight years, several companies have developed specific techniques to address the problems of communication between data processors and end users. Facilitated application specification techniques (Fast) focus on the information-gathering stage of system design. Fast sessions bring DP staff and end users together, making design an interactive process.

Corporate clients are trained in the technique of their choice by one of several vendors and then conduct sessions in-house. The hope is that negotiations at the early stages will ensure smooth operations later on. If end-user requirements are spelled out in flow charts and brought face-to-face with DP considerations at this early stage, the final system design will suit the organization more precisely, boosting productivity over the system’s lifetime.

As word of their success spreads, these interactive design techniques are finally being accepted and implemented in various companies in the U.S. Major corporations are investing in visual aids and training for session leaders. For the first time since these techniques were developed, a variety of consulting and training assistance is also available.

Interactive design has a direct impact on DP professionals’ ability to deliver error-free applications. DP management is all too aware that reducing errors is one of the most effective ways to reduce the cost of computer systems. Now studies at ITT Corp., IBM, TRW, Inc. and Mitre Corp. indicate that error removal constitutes up to 40% of the cost of a system – and that between 45% and 65% of these errors are made in system design.

Numerous analytical methodologies, design methodologies, and programming techniques have been developed to address the error problem. While these work well for analysis and design, they have not always addressed the interviewing and information gathering process that must take place to provide input. Now Fast has made a science of information gathering.

The information gathering process bogs down for two major reasons. The first is the communication gap between DP and the business community. In An Information Systems Manifesto (Prentiss-Hall, 1984), James Martin says, “When the traditional systems analyst and potential end users first come face-to-face, they come from widely different cultures. It is rather like a Victorian missionary first entering an African village.”

This language difference brings about the second major problem: the invariable power struggles between the players involved. The end user demands systems faster than DP can deliver them, or else DP creates a sort of technological blackmail over the end user. Often these power struggles are not even intentional but result from widely different approaches to and views of the same problem.

Four major techniques have been developed since the late 1970’s to address the information gathering problem. They are Joint Application Design (JAD), Consensus, Wisdm, and The Method.

In 1977, IBM developed JAD to help extract requirements for distributed systems implementation. Boeing Computer Services Co. adapted a technique that was used to design the Boeing 747 aircraft and started Consensus. Wisdm was developed by Blair Burner at the Western Institute of Software Engineering (WISE). Performance Resources, Inc. took the JAD technique, modified it to work better for decision support systems and called it The Method.

Structured Agenda

All of these methods are geared toward the front end of the system design life cycle. They help users define an application from its first conception through the complete design. These techniques can be used for every new development of maintenance system project in a company.

Fast sessions range in length from one three-day workshop to more than 30 workshops to bring the design to a point where it can be turned over to technicians for programming. Any project that requires more than one person’s input on decisions can profit from interactive design techniques such as these.

Typically, the project manager will initiate using Fast in the company before the start of a new project. The project manager may contact a Fast vendor for training or licensing or else present the idea to top management to make the contact. Occasionally, other DP staff members or even end users will be the initiators. In any case, the project manager will not end up actually leading the sessions, as the session leaders’ success depends ultimately on their position as objective outsiders.

Diverse views

Each Fast method focuses on a slightly different aspect of the customer’s needs. JAD primarily addresses the detailed external business design problem. It has a structured agenda that follows the flow of work through a work area and details each stage: planning, receiving, tracking, assigning, processing, recording, sending, and evaluating the work.

Each JAD session is lead by an impartial session leader who is responsible for controlling the agenda and the numerous visual aids. The session leader, with the project manager’s help, prepare beforehand, tailoring each session with application-specific information. The leader must also ensure that management, users, and DP come to the session with the same purpose and objectives and that the scope of the session is clearly laid out.

JAD stresses the use of business rather than technical language to encourage participation and enhance group dynamics. Visual aids in the form of vinyl magnetics, slides and Vu-Graphs help drive the detail and quality of the design.

JAD sessions typically run three days and involve both technical and managerial personnel. Participants are key staff within the business area for which the system is being developed – whoever is in the best position to describe the business functions, information, and data needs, transactions, design screens, reports, and document changes to the workflow.

Companies such as American Airlines, Texas Instruments, Inc., IBM, Mutual Life Insurance Co. in New York, and Bell Canada have been using JAD successfully. Some recorded increases in productivity ranging from 20% to 60% during the requirements and design process.

CNA Insurance Co. in Chicago conducted a trial of the JAD technique in 1983. CNA used the Function Point productivity measurement technique from IBM to evaluate JAD’s effectiveness. A control project that did not use JAD was measured along with the pilot project. The control averaged eight hours per function point during the requirements and external design phases of development, whereas the pilot, averaged 2.5 hours per point – more than a 200% increase in productivity.

High-level planning

Consensus covers four areas: strategic planning, management planning, requirements definition/analysis and preliminary design. Like JAD, Consensus primarily covers the higher level planning issues. Consensus does not delve into screen and report design. It does, however, cover business functions and information or data needs in detail.

The session leader in a Consensus session will not use numerous visual aids other than Vu-Graphs and flip charts. On the flip charts, the session leader will draw graphic representations of the system using data flow and information flow diagrams to aid the discussion.

The workshops generally run long hours (12 to 14 hours a day) and often for up to two weeks for larger systems, with a majority of the participants staying the entire time. The users of Consensus report up to an 80% reduction in elapsed time and a 50% reduction in costs for the requirements gathering process. Consensus is used by such companies as AMP, Inc., Boeing Computer Services Co., First Interstate Bank, and Blue Cross/Blue Shield.

Brainchild

Wisdm was conceived by Blair Burner while he was an employee at Boeing Computer Services in the mid-1970’s. In the late 1970’s, Burner formed Wise, where he perfected and began marketing Wisdm.

Wisdm is similar to consensus in that it is aimed primarily at front-end analysis. The first step in a Wisdm analysis is a thorough Problem Definition and Analysis workshop, generally consisting of key managers from the business area involved, which defines the business problem clearly. The second step is a Requirements Definition and Analysis workshop, followed by the final step, a Business System Design workshop. Wisdm, unlike Consensus, runs six to eight hours per day for three to 10 days.

In the last step, the requirements, business flow, data needs, and major processes are defined in a step-by-step process leading to a complete design:

  • Defining the external interfaces
  • Defining the external inputs and outputs
  • Building an interface model of the input and output flow
  • Defining internal functions
  • Building a graphic model of the functional flow
  • Defining the internal and stored data
  • Constructing a data flow model
  • Defining the general current and future requirements for the system.

Wisdm design workshops require two session leaders, one of whom sometimes takes the role of an apprentice in preparation for a future session. Wisdm makes heavy use of matrices and other visual aids to illustrate data elements. Unlike JAD, Wisdm does not extend into the detailed design portions of systems design. Wisdm is used extensively by the consulting firms of Wise and Comp-U Staff and has been used at the Hartford Insurance Co.

Hybrid technique

The Method is really a hybrid of the JAD technique. It was fashioned because decision support systems did not fit easily into the transaction flow agenda of JAD. Developed by Performance Resources, Inc. while working the AT&T, The Method is now used at CNA Insurance Co., AT&T, the Chase Manhattan Bank NA and some government agencies. The Method addresses decision support better than JAD in that it is more data driven: Data is identified earlier in the process and then is organized to help develop decision support.

The Method provides for planning sessions, a work analysis to set the scope of the sessions, a structured agenda, and an impartial session leader. In this case, session leaders are certified before running any sessions. Session leader training includes instruction in group dynamics plus specific software support to aid the session leader in planning workshops, scheduling time and tailoring the workshops to particular applications.

Implementing the techniques does incur costs. Implementation requires investing in training and materials, customizing the technique, securing workshop facilities, recruiting session leaders and conducting pilots. Training and implementation expenses can range from $3,000 to more than $100,000, depending on the amount of consulting support desired, the number of leaders trained, and the amount invested in workshop facilities.

But as a rule, this initial investment is easily recovered. On a typical 2,000-hour project, 20% or 400 hours, of the time is spent in requirements and design phase will require only 300 hours, saving 100 hours. Depending on the internal company billing rates, this 100 hours can be worth from $3,500 to $7,000.

Even with worst-case investment and return, the investment pays for itself after the 28th project – or much sooner when the technique is used on larger projects. Typical investments and returns are generally far from the worst case.

It is important to test or pilot the technique on one project with a good chance for success to determine how much customizing is necessary and how well the technique fits within the organizational climate. This initial pilot of the technique is the best way to sell the concept internally and gain acceptance.

Each of the four Fast methods has strengths and weaknesses. But boiled down, they are all structured meeting techniques designed to extract high-quality business system specifications from end users in a compressed time frame using a workshop environment. They are not replacements for analytical methodologies, but they can all work with and supplement any methodology.

Fast Elements

All of these techniques have certain common elements that both characterize them and determine their success. If customers modify or customize a technique, they should take care to preserve these key elements: a structure or formalized process, a dynamic workshop environment, an impartial session leader, a focus on information gathering and business system design and well-defined goals (producing system specifications and objectives and predefined documentation forms).

Each technique also stresses end-user input and ownership of the system and requires end-user commitment both at the outset and the completion of the sessions. Good visual aids and comprehensive documentation are also important.

Structure is important to these techniques because the structures chosen are built on proven methods. The sessions have an agenda, a purpose and objectives. The workshops are orchestrated, not ad hoc, which helps ensure that they do not turn into time-consuming brainstorming sessions or uncontrolled discussions.

Often the structure is as simple as following the flow of actions against a piece of work and asking end users the following questions at each stage:

  • What functions are you performing at this point?
  • What information do you need to complete these functions?
  • What do you want done with the information?
  • How do you want the information displayed?

These questions are repeated for each step of the work flow.

The workshop environment and dynamics of a group help session participants concentrate on idea sharing, avoid politics and ensure that the information provided is complete. Group dynamics help avoid politics because groups tend to police themselves. The participants develop into a team by working together, in such a group pettiness and politics are seen for what they are and quickly disappear.

A key to creating this group environment, however, is having the right people in the workshop. These people need to be knowledgeable about the business and have the authority to make decisions about the design. In the workshop, end users will be asked to describe their business functions, information needs, data elements used and how they want to interface with the system. They will describe how the screens and reports should look and how the system will affect their business.

The type and number of participants vary, based on the level of detail being discussed in a workshop. In high-level planning workshops, there are up to 25 participants. These participants are the decision makers and managers of an organization. In workshops concerned with details of screens, reports and workflow design, there are fewer participants (usually a maximum of 12), and these will be the lead technical people in the end-user departments, their supervisors, and perhaps a manager or two.

DP is represented by the project leader and one or two lead technical advisers, who assist in answering questions about other systems, feasibility of design items and some cost estimates. The advisers make alternative suggestions when cost becomes a deciding factor. The session leader and one or two people to capture the documentation round out the participants.

An impartial session leader can eliminate the power struggles and communication gaps. A 1984 study, “Management Science,” by DeBrabander and Thiers of the State University Centre Antwerp in Belgium found that “the presence of a third party which stimulates the user to neglect the possible implications of power asymmetry… nullifies [this] disturbing effect.” Their study noted that the most effective type of facilitator is an active question elicitor. This type of session leader induces the quieter or less assertive workshop members to ask more questions or to respond when someone else takes a position.

This type of facilitated discussion puts DP and end users on equal terms and sets them up as partners. The session leader assumes the role of a referee at times, arbitrating debates between DP and end users of between groups of end users. When an issue arises that cannot be decided in a workshop, the session leader notes down the issue, and the group assigns someone to be responsible for its resolution. The goal is to discuss ideas fully and to reach decisions as a group without delaying the process in undue haggling.

Most of the techniques rely on the use of visual aids to help with communication. Better communication is, after all, the primary reason for structure, session leadership and the workshop environment. Often problems arise because the participants define terms indifferent ways. Defining terminology and building pictures of the evolving application serve to avoid misunderstandings and boost the group’s productivity.

Magnetics, slides, Vu-Graphs, charts, and drawings are all used in Fast sessions. Figure 1 illustrates slides used to help clarify terms and workshop process. Understanding what “input,” “data flow” or “function” means eliminates ambiguities and confusion. Through charts, the workshop can build up an application step by step.

Well-defined documentation is important to the consistency of the methods. Documentation is usually specific information recorded during the sessions. The documentation itself actually becomes the requirements specifications, so recording the session thoroughly determines, to some degree, the workshop’s usefulness.

The level of documentation detail will depend upon the level of the session. Planning sessions document the business problems, general requirements, action plans and priorities. Detailed design sessions produce the documentation needed to develop programming specifications.

More than a meeting

New Fast customers will often ask, “What is the difference between these techniques and a well-run meeting?” There are several answers.

The Fast method can be repeated because they are internally consistent and have been tested. Leaders can be trained in any of the techniques. Each technique holds known, explicit objectives.

Well-run meetings, on the other hand, depend on the skills of the meeting leader, who has not been specifically trained for the task. In general, the results of even a well-run meeting are unpredictable. In organizations where meetings are well run, Fast methods provide additional benefits to the application design information-gathering activities.

Each technique has its strengths and weaknesses. No one is all encompassing. Documentation during the sessions has bee a problem, and there is not well-developed automated documentation tool. The techniques all have been modified when brought into a company other than the developing company, and since these are relatively new and implemented differently in each organization, help for customers has not been readily available until just recently.

Most corporate customers do not feel comfortable implementing these techniques without help from a consultant or the product vendor. Many vendors license individuals rather than companies to use their techniques and are reluctant to customize their techniques for a particular company. Those that will customize their technique often charge a hefty consulting fee to do so.

There are six primary companies offering services relating to interactive design sessions. IBM Information Systems Services and JAtech Design Systems, Ltd. offer contract JAD session leaders and JAD training. Performance Resources, Inc. offers contract session leading plus training and certification in The Method. Wise offers contract session leading as well as training and licensing in Wisdm.

Boeing Computer Services offers contract session leading and training in Consensus. Gary Rush offers contract session leading and training in a combination and adaptation of various techniques. In addition, various consulting companies are beginning to offer contract session leaders as part of their services.

On the plus side, Fast methods have worked very well. End users are satisfied and generally become better allies of DP. In one case, the users emphatically told the vendor to provide more support for the process so that they could use it on all of their systems. At least four companies (AT&T Communications, American Airlines, CNA, and TI) have trained numerous session leaders and use interactive design for many of their projects.

The specifications developed from these methods have been more thorough, better documented and more consistent than with a less rigorous approach as well as being obtainable more quickly and at less expense. Using a facilitated technique helps reduce maintenance because specifications are more fully developed and users know their system better. These methods can also enhance other analytical and design methodologies, including prototyping.

To succeed, the initial project should have the following:

  • End users hungry to have a system developed for them.
  • Cooperative, committed DP and end-user personnel.
  • A project requiring no more than 12 to 18 months of development effort.
  • A project that is primarily an on-line, transaction-based system.
  • An application which the underlying business is well defined and not controversial.

The initial use of a technique in a company is more critical for a session leader than the first workshop. Companies will not give a technique a second chance if it fails on the first try. A session leader, on the other hand, can often recover if the first session is not a great success.

Tools are available to help. Diagramming aids such as Index Technology Corp.’s Excelerator, IBM’s System A, or Technology and Information Products Corp’s Extim facilitate thorough, well-organized documentation. Data dictionaries such as CGI Systems, Inc.’s Pacbase or IBM’s Data Dictionary can help make the documentation a living portion of the application.

The data dictionaries also can help in developing prototypes or feeding prototyping tools. Finally, word processors and specification languages can be used to capture documentation.

Good training has not been easy to find. Most developers of the techniques teach only the technique itself. But training must also cover group dynamics, the facilitator concept, the specific implementation of the technique, systems analysis, tools for use in the workshop and the documentation process.

Regardless of the technique, each session will be somewhat unique, and the session leader must be capable of adjusting to different situations. Knowing the technique alone will not be enough.

The only vendors or consultants that provide extensive training are Performance Resources and Gary Rush. Both provide training in the design technique as well as in facilitation and group dynamics.

Performance Resources trains students in The Method using effective presentation techniques and facilitation training to enhance the session leader skills while providing basic instruction in the design technique. Gary Rush trains students in a variety of design concepts using group development, adult learning and facilitation training to enhance the session leader’s skills. Some customers have developed their own in-house programs that do a good job of training session leaders.

Selecting session leaders is also not easy. These people need to be well respected and feel comfortable standing in front of a group of people. They need to be able to control controversies and stay flexible. Above all, the session leader must be confident and well prepared.

Almost all session leaders have come from the DP ranks. A few have come from marketing backgrounds. Most session leaders are capable technical people with a high degree of communication and people skill.

The DeBrabander and Thiers study found that the most effective facilitators “are characterized by an interpersonal, time and goal orientation. They must have high influence throughout the organization… derived from technical competence.”

Need for change

The future for these techniques is bright, but all will need to evolve. Vendors need to keep the techniques current with the changing business environments. For example, non-DP personnel are not as unfamiliar with computer as they were in the mid- to late 1970’s. the types of applications being developed are changing from heavy transaction systems to smaller decision support, information or specialized transaction systems.

Fast techniques are also being used for more than data processing. They have been used to design robots, grain elevators, strategic planning – even to improve staff meetings. Customers need to understand the techniques fully so that they can integrate the techniques into their existing ways of building applications. If a methodology is being used successfully in a company, then it should be integrated so that it augments rather than replaces the methodology.

Guidance in planning, understanding, documentation and training is needed and must come from vendors and practitioners. The techniques need to be customized and integrated fully with local methodologies without compromising either one.

The techniques do work. They work well when customers invest some effort in understanding them. Sessions should evolve from the proven techniques and build on the experiences of the practitioners.

Most DP professionals will agree that getting better requirements and more user involvement will ensure better systems and reduced maintenance. Clearly these techniques assist in this; furthermore, involving users in the process of designing their business systems makes the users feel that the system being developed really is their system. logo

Update January 2016: "Consensus, Wisdm, The Method no longer exist. Gary Rush IAF CPF now teaches FoCuSeD™ Facilitation - a revolution from FAST Facilitation."

*Content on this page (all or in part) may NOT be used elsewhere without expressed permission from Gary Rush, IAF CPF. Thank you for respecting the effort that we have put into our original content.

guy