Systems analysis and design
The systems development life cycle (SDLC) is the process of understanding how an information system (IS) can support business needs, designing the system, building it, and delivering it to users. If you have taken a programming class or have programmed on your own, this probably sounds pretty simple. Unfortunately, it is not. A 2004 survey by the Standish Group found that just 28% of IT projects succeed these days. Outright failures—IT projects cancelled before completion—occur in
18% of all IT projects. Unfortunately, many of the systems that aren’t abandoned are delivered to the users significantly late, cost far more than planned, and have fewer features than originally planned.
Most of us would like to think that these problems only occur to “other” people or “other” organizations, but they happen in most companies. See Figure 1-1 for a sampling of significant IT project failures. Even Microsoft has a history of failures and overdue projects (e.g., Windows 1.0, Windows 95).1
The key person in the SDLC is the systems analyst who analyzes the business situation, identifies opportunities for improvements, and designs an information system to implement them. Being a systems analyst is one of the most interesting, exciting, and challenging jobs around. As a systems analyst, you will work with a variety of people and learn how they conduct business. Specifically, you will work with a team of systems analysts, programmers, and others on a common mission.
You will feel the satisfaction of seeing systems that you designed and developed make a significant business impact, while knowing that your unique skills helped make that happen. It is important to remember that the primary objective of the systems analyst is not to create a wonderful system. The primary goal is to create value for the organization, which for most companies means increasing profits (government agencies and not-for-profit organizations measure value differently). Many failed systems were abandoned because the analysts tried to build a wonderful system without clearly understanding how the system would support the organization’s goals, current business processes, and other information systems to provide value. An investment in an information system is like any other investment, such as a new machine tool. The goal is not to acquire the tool, because the tool is simply a means to an end; the goal is to enable the organization to perform work better so it can earn greater profits or serve its constituents more effectively.
In many ways, building an information system is similar to building a house. First, the house (or the information system) starts with a basic idea. Second, this idea is transformed into a simple drawing that is shown to the customer and refined (often through several drawings, each improving on the other) until the customer agrees that the picture depicts what he or she wants. Third, a set of blueprints is designed that presents much more detailed information about the house (e.g., the type of
water faucets, where the telephone jacks will be placed). Finally, the house is built following the blueprints—and often with some changes and decisions made by the customer as the house is erected.
The SDLC has a similar set of four fundamental phases: planning, analysis, design, and implementation. Different projects may emphasize different parts of the SDLC or approach the SDLC phases in different ways, but all projects have elements of these four phases. Each phase is itself composed of a series of steps, which rely on techniques that produce deliverables (specific documents and files that provide understanding about the project). For example, when you apply for admission to a university, there are several phases that all students go through: information gathering, applying, and accepting. Each of these phases has steps: information gathering includes steps like searching for schools, requesting information, and reading brochures. Students then use techniques (e.g., Internet searching) that can be applied to steps (e.g., requesting information)
to create deliverables (e.g., evaluations of different aspects of universities). Figure 1-2 suggests that the SDLC phases and steps proceed in a logical path from start to finish. In some projects, this is true, but in many projects, the project teams move through the steps consecutively, incrementally, iteratively, or in other patterns. In this section, we describe at a very high level the phases, steps, and some of the techniques that are used to accomplish the steps. We should emphasize that, in practice, an organization may follow one of many variations on the overall SDLC. For now, and there are two important points to understand about the SDLC. First, you should get a general sense of the phases and steps that IS projects move through and some of the techniques that produce certain deliverables. Second, it is important to understand that the SDLC is a process of gradual refinement. The deliverables produced in the analysis phase provide a general idea of the shape of the new system. These deliverables are used as input to the design phase, which then refines them to produce a set of deliverables that describes in much more detailed terms exactly how the system will be built. These deliverables in turn are used in the implementation phase to produce the actual system. Each phase refines and elaborates on the work done previously.
Planning
The planning phase is the fundamental process of understanding why an information system should be built and determining how the project team will go about building it. It has two steps:
1. During project initiation, the system’s business value to the organization is identified—how will it lower costs or increase revenues? Most ideas for new systems come from outside the IS area (from the marketing department, accounting department, etc.) in the form of a system request. A system request
Presents a brief summary of a business need, and it explains how a system that supports the need will create business value. The IS department works together with the person or department that generated the request (called the project sponsor) to conduct a feasibility analysis. The feasibility analysis examines key aspects of the proposed project:
- The technical feasibility (Can we build it?)
- The economic feasibility (Will it provide business value?)
- The organizational feasibility (If we build it, will it be used?)
The system request and feasibility analysis are presented to an information systems approval committee (sometimes called a steering committee), which decides whether the project should be undertaken.
2. Once the project is approved, it enters project management. During project management, the project manager creates a work plan, staffs the projects, and puts techniques in place to help the project team control and direct the project through the entire SDLC. The deliverable for project management is a project plan that describes how the project team will go about developing the system.
Analysis
The analysis phase answers the questions of who will use the system, what the system will do, and where and when it will be used. See Figure 1-2. During this phase, the project team investigates any current system(s), identifies improvement opportunities, and develops a concept for the new system. This phase has three steps:
1. An analysis strategy is developed to guide the project team’s efforts. Such a strategy usually includes an analysis of the current system (called the as-is system) and its problems, and then ways to design a new system (called the to-be system).
2. The next step is requirements gathering (e.g., through interviews or questionnaires).
The analysis of this information—in conjunction with input from the project sponsor and many other people—leads to the development of a concept for a new system. The system concept is then used as a basis to develop a set of business analysis models that describes how the business will operate if the new system were developed. The set of models typically includes models that represent the data and processes necessary to support the underlying business process.
3. The analyses, system concept, and models are combined into a document called the system proposal, which is presented to the project sponsor and other key decision makers (e.g., members of the approval committee) that decide whether the project should continue to move forward.
The system proposal is the initial deliverable that describes what business requirements the new system should meet. Because it is really the first step in the design of the new system, some experts argue that it is inappropriate to use the term analysis as the name for this phase; some argue a better name would be analysis and initial design. Because most organizations continue to use the name analysis for this phase, we will use it in this book as well. It is important to remember, however,
that the deliverable from the analysis phase is both an analysis and a high-level initial design for the new system.
Design
The design phase decides how the system will operate, in terms of the hardware, software, and network infrastructure; the user interface, forms, and reports that will be used; and the specific programs, databases, and files that will be needed. Although most of the strategic decisions about the system were made in the development of the system concept during the analysis phase, the steps in the design phase determine exactly how the system will operate. The design phase has four steps:
1. The design strategy must be developed. This clarifies whether the system will be developed by the company’s own programmers, whether it will be outsourced to another firm (usually a consulting firm), or whether the company will buy an existing software package.
2. This leads to the development of the basic architecture design for the system that describes the hardware, software, and network infrastructure that will be used. In most cases, the system will add or change the infrastructure that already exists in the organization. The interface design specifies how the users will move through the system (e.g., navigation methods such as menus and on-screen buttons)
and the forms and reports that the system will use.
3. The database and file specifications are developed. These define exactly what data will be stored and where they will be stored.
4. The analyst team develops the program design, which defines the programs that need to be written and exactly what each program will do. This collection of deliverables (architecture design, interface design, database and file specifications, and program design) is the system specification that is handed to the programming team for implementation. At the end of the design phase, the feasibility analysis and project plan are reexamined and revised, and another decision is made by the project sponsor and approval committee about whether to terminate the project or continue.
Implementation
The final phase in the SDLC is the implementation phase, during which the system is actually built (or purchased, in the case of a packaged software design). This is the phase that usually gets the most attention, because for most systems it is the longest and most expensive single part of the development process. This phase has three steps:
1. System construction is the first step. The system is built and tested to ensure it performs as designed. Since the cost of bugs can be immense, testing is one of the most critical steps in implementation. Most organizations spend more time and attention on testing than on writing the programs in the first place.
2. The system is installed. Installation is the process by which the old system is turned off and the new one is turned on. It may include a direct cutover approach (in which the new system immediately replaces the old system), a parallel conversion approach (in which both the old and new systems are operated for a month or two until it is clear that there are no bugs in the new system), or a phased conversion strategy (in which the new system is installed in one part of the organization as an initial trial and then gradually installed in others). One of the most important aspects of conversion is the development of a training plan to teach users how to use the new system and help manage the changes caused by the new system.
3. The analyst team establishes a support plan for the system. This plan usually includes a formal or informal post-implementation review, as well as a systematic way for identifying major and minor changes needed for the system.
SYSTEMS DEVELOPMENT METHODOLOGIES
A methodology is a formalized approach to implementing the SDLC (i.e., it is a list of steps and deliverables). There are many different systems development methodologies and each one is unique because of its emphasis on processes versus data and the order and focus it places on each SDLC phase. Some methodologies are formal standards used by government agencies, while others have been developed by consulting firms to sell to clients. Many organizations have their own internal methodologies that have been refined over the years, and they explain exactly how each phase of the SDLC is to be performed in that company. All system development methodologies lead to a representation of the system concept in terms of processes and data; however, they vary in terms of whether the methodology places primary emphasis on business processes or on the data that supports the business. As an illustration, refer to the diagram shown in bellow picture depicting the activities and information used in producing the payroll for an organization.
The open-ended rectangles in the diagram represent data storage containers; the rounded rectangles represent activities performed; and the arrows represent activity inputs and outputs. Process-centered methodologies focus first on defining the activities associated with the system, i.e., the processes. Process-centered methodologies utilize process models as the core of the system concept. Analysts concentrate initially on representing the system concept as a set of processes with information flowing into and out of the processes Data-centered methodologies focus first on defining the contents of the data storage containers and how the contents are organized. Data-centered methodologies utilize data models as the core of the system concept. For example, analysts concentrate initially on identifying the data that must be available to produce
the payroll and organizing that data into well-defined structures (e.g., employee work log, employee pay rates, payroll tax tables, employee pay history, etc.). Object-oriented methodologies (Chapter 15) attempt to balance the focus between processes and data. Object-oriented methodologies utilize the Unified Modeling Language (UML) to describe the system concept as a collection of objects incorporating both data and processes.3 Another important factor in categorizing methodologies is the sequencing of the SDLC phases and the amount of time and effort devoted to each.4 In the early days of computing, the need for formal and well-planned life cycle methodologies was not well understood. Programmers tended to move directly from a very simple planning phase right into the construction step of the implementation phase; in other words, they moved directly from a very fuzzy, not-well-thought-out system request into writing code. This is the same approach that you may sometimes use when writing programs for a programming class. It can work for small programs that require only
one programmer, but if the requirements are complex or unclear, you may miss important aspects of the problem and have to start all over again, throwing away part of the program (and the time and effort spent writing it). This approach also makes teamwork difficult because members have little idea about what needs to be accomplished and how to work together to produce a final product.
In the following sections, we describe three major categories of systems development methodologies that have evolved over time: Structured Design, Rapid Application Development (RAD), and Agile Development. Each category represents a collection of methodologies that attempts to improve on previous practice, and varies in terms of the progression through the SDLC phases and the emphasis placed on each phase.
Structured Design
The first category of systems development methodologies is called structured design. These methodologies became dominant in the 1980s, replacing the previous ad hoc and undisciplined approach. Structured design methodologies adopt a formal step-by-step approach to the SDLC that moves logically from one phase to the next. Structured design also introduced the use of formal modeling or diagramming techniques to describe a system’s basic business processes and the data that support them. Traditional structured design uses one set of diagrams to represent the processes (process models) and a separate set of diagrams to represent data (data models). Because two sets of models are used, the systems analyst must decide which set to develop first and use as the core of the system—process models or data models. Since each type of model is important to the system, there is much debate over whether to emphasize processes before data or vice versa. Numerous process centered and data-centered methodologies follow the basic approach of the two structured design categories outlined next.
Waterfall Development The original structured design methodology (that is still used today) is waterfall development. With waterfall development-based methodologies, the analysts and users proceed sequentially from one phase to the next (see Figure 1-4). The key deliverables for each phase are typically voluminous (often hundreds of pages in length) and are presented to the project sponsor for approval as the project moves from phase to phase. Once the sponsor approves the work that was conducted for a phase, the phase ends and the next one begins. This methodology is called waterfall development because it moves forward from phase to phase in the same manner as a waterfall. Although it is possible to go backward in the SDLC.
The two key advantages of waterfall development-based methodologies are that system requirements are identified long before programming begins and that changes to the requirements are minimized as the project proceeds. The two key disadvantages are that the design must be completely specified before programming
Begins and that a long time elapses between the completion of the system proposal in the analysis phase and the delivery of the system (usually many months or years). The deliverables are often a poor communication mechanism, so important requirements can be overlooked in the documentation. Users rarely are prepared for their introduction to the new system, which occurs long after the initial idea for the system was introduced. If the project team misses important requirements, expensive
Post-implementation programming may be needed (imagine yourself trying to design a car on paper; how likely would you be to remember to include interior lights that come on when the doors open or to specify the right number of valves on the engine?).
Today’s dynamic business world often imposes rapid environmental change on businesses. A system that meets existing environmental conditions during the analysis phase may need considerable rework to match the environment when it is implemented. This rework requires going back to the initial phase and making needed changes through each of the subsequent phases in turn.
Parallel Development
The parallel development-based methodologies attempt to address the long time interval between the analysis phase and the delivery of the system. Instead of doing the design and implementation in sequence, a general design for the whole system is performed, then the project is divided into a series
of distinct subprojects that can be designed and implemented in parallel. Once all subprojects are complete, there is a final integration of the separate pieces, and the system is delivered.
The primary advantage of these methodologies is that the schedule time required to deliver a system is shortened; thus, there is less chance of changes in the business environment causing rework. The approach still suffers from problems caused by lengthy deliverables. It also adds a new problem: sometimes the subprojects are not completely independent; design decisions made in one subproject
may affect another, and the end of the project may involve significant integration challenges.
Rapid Application Development (RAD)
The second system development methodology category includes rapid application development (RAD)-based methodologies. These are a newer class of system development methodologies that emerged in the 1990s in response to both structured design methodology weaknesses. RAD-based methodologies adjust the SDLC phases to get some part of the system developed quickly and into the hands of the users. In this way, the users can better understand the system and suggest revisions that bring the system close to what is needed.5 There are process-centered, data centered, and object- oriented methodologies that follow the basic approaches of the three RAD categories described in this section. Most RAD-based methodologies recommend that analysts use special techniques and computer tools to speed up the analysis, design, and implementation phases, such as CASE (computer-aided software engineering) tools (see Chapter 3), JAD (joint application design) sessions (see Chapter 5), fourth-generation/visual programming languages that simplify and speed up programming (e.g., Visual Basic.NET), and code generators that automatically produce programs from design specifications. It is the combination of the changed SDLC phases and the use of these tools and techniques that improves the speed and quality of systems development. One possible subtle problem with RAD-based methodologies, however, is managing user expectations. As systems are developed more rapidly and users gain a better understanding of information technology, user expectations may dramatically increase and system requirements expand during the project. This was less of a problem with structured design methodologies where the system requirements, once determined, were allowed only minimal change.
Phased Development
The phased development-based methodologies break the overall system into a series of versions that are developed sequentially. The analysis phase identifies the overall system concept, and the project team, users, and system sponsor then categorize the requirements into a series of versions. The most important and fundamental requirements are bundled into the first version of the system. The analysis phase then leads into design and implementation, but only with the set of requirements identified for version 1. Once version 1 is implemented, work begins on version 2. Additional analysis is performed on the basis of the previously identified requirements and combined with new ideas and issues that arose from users’ experience with version 1. Version 2 then is designed and implemented, and work immediately begins on the next version.This process continues until the system is complete or is no longer in use. Phased development-based methodologies have the advantage of quickly getting
a useful system into the hands of the users. Although it does not perform all the functions the users need at first, it begins to provide business value sooner than if the system were delivered only after all requirements are completed, as is the case with the waterfall or parallel methodologies. Likewise, because users begin to work with the system sooner, they are more likely to identify important additional requirements sooner than with structured design situations.
The major drawback to phased development is that users begin to work with systems that are intentionally incomplete. It is critical to identify the most important and useful features and include them in the first version while managing users’ expectations along the way.
Prototyping The prototyping-based methodologies perform the analysis, design, and implementation phases concurrently, and all three phases are performed repeatedly in a cycle until the system is completed. With these methodologies, a basic analysis and design are performed, and work immediately begins on a systemprototype, a “quick-and-dirty” program that provides a minimal amount of features. The first prototype is usually the first part of the system that the user will use. This is shown to the users and the project sponsor, who provide reaction and comments. This feedback is used to reanalyze, redesign, and reimplement a second prototype that provides a few more features. This process continues in a cycle until the analysts, users, and sponsor agree that the prototype provides enough functionality to be installed and used in the organization. After the prototype (now called the “system”) is installed, refinement occurs until it is accepted as the new system (see Figure 1-7). The key advantage of a prototyping-based methodology is that it very quickly provides a system for the users to interact with, even if it is not initially ready for widespread organizational use. Prototyping reassures the users that the project team is working on the system (there are no long time intervals in which the users perceive little progress), and the approach helps to more quickly refine real require- ments. Rather than attempting to understand system specification materials, the users can interact with the prototype to better understand what it can and cannot do. The major problem with prototyping is that its fast-paced system releases challenge attempts to conduct careful, methodical analysis. Often the prototype undergoes such significant changes that many initial design decisions prove to be poor ones. This can cause problems in the development of complex systems because fundamental issues and problems are not recognized until well into the development process. Imagine building a car and discovering late in the prototyping process that you have to take the whole engine out to change the oil (because no one thought about the need to change the oil until after the car had been driven 10,000 miles).
Throwaway Prototyping Throwaway prototyping-based methodologies are similar to the prototyping-based methodologies in that they include the development of prototypes; however, throwaway prototypes are done at a different point in the SDLC. These prototypes are used for a very different purpose than ones previously discussed, and they have a very different appearance6. The throwaway prototyping-based methodologies have a relatively thorough analysis phase that is used to gather information and to develop ideas for the system concept. Many of the features suggested by the users may not be well understood, however, and there may be challenging technical issues to be solved. Each of these issues is examined by analyzing, designing, and building a design prototype.
A design prototype is not a working system; it is a product that represents a part of the system that needs additional refinement, and it contains only enough detail to enable users to understand the issues under consideration. For example, suppose users are not completely clear on how an order entry system should work. The analyst team might build a series of HTML pages viewed using a Web browser to help the users visualize such a system. In this case, a series of mock-up screens appear to be a system, but they really do nothing. Or, suppose that the project team needs to develop a sophisticated graphics program in Java. The team could write a portion of the program with artificial data to ensure that they could create a full-blown program successfully. A system that is developed using this type of methodology probably uses several design prototypes during the analysis and design phases. Each of the prototypes is used to minimize the risk associated with the system by confirming that important issues are understood before the real system is built. Once the issues are resolved, the project moves into design and implementation. At this point, the design prototypes are thrown away, which is an important difference between this approach and prototyping, in which the prototypes evolve into the final system.
Throwaway prototyping-based methodologies balance the benefits of well thought- out analysis and design phases with the advantages of using prototypes to refine key issues before a system is built. It may take longer to deliver the final system as compared with prototyping-based methodologies (because the prototypes do not become the final system), but the approach usually produces more stable and reliable systems.
Agile Development
A third category of systems development methodologies is still emerging today: Agile Development.7 these programming-centric methodologies have few rules and practices, all of which are fairly easy to follow. They focus on streamlining the SDLC by eliminating much of the modeling and documentation overhead and the time spent on those tasks. Instead, projects emphasize simple, iterative application development. Examples of Agile Development methodologies include extreme programming, 8 Scrum,9 and the Dynamic Systems Development Method (DSDM).10
To illustrate an agile development methodology, we describe extreme programming in the next section. Typically, extreme programming is used in conjunction with object-oriented programming languages.
Extreme Programming Extreme programming (XP)11 is founded on four core values:
Communication, simplicity, feedback, and courage. These four values provide a foundation XP developers use to create any system. First, the developers must provide rapid feedback to the end users on a continuous basis. Second, XP requires developers to follow the KISS (Keep It Simple, Stupid) principle. Third, developers must make incremental changes to grow the system and they must embrace change, not merely accept it. Fourth, developers must have a quality first mentality. XP also supports team members in developing their own skills. Three of the key principles that XP uses to create successful systems are continuous testing, simple coding performed by pairs of developers, and close interactions with end users to build systems very quickly. After a superficial planning
process, project teams perform analysis, design, and implementation phases iteratively Testing and efficient coding practices are core to XP. In fact, each day code is tested and placed into an integrative testing environment. If bugs exist, the code is backed out until it is completely free of errors. XP relies heavily on refactoring, which is a disciplined way to restructure code to keep it simple. An XP project begins with user stories that describe what the system needs to do. Then, programmers code in small, simple modules and test to meet those needs. Users are required to be available to clear up questions and issues as they arise. Standards are very important to minimize confusion, so XP teams use a common set of names, descriptions, and coding practices. XP projects deliver results sooner
than even the RAD approaches, and they rarely get bogged down in gathering requirements for the system. For small projects with highly motivated, cohesive, stable, and experienced teams, XP should work just fine. However, if the project is not small or the teams aren’t jelled12 then the likelihood of a successful XP project is reduced. Consequently, the use of XP in combination with outside contractors produces a highly questionable outcome, since the outside contractors may never “jell” with insiders.
13 XP requires a great deal of discipline to prevent projects from becoming unfocused and chaotic. Furthermore, it is only recommended for small groups of developers (not more than ten), and it is not advised for mission-critical Applications. Since little analysis and design documentation is produced with XP there is only code documentation; therefore, maintenance of large systems developed using
XP may be impossible. Also, since mission-critical business information systems tend to exist for a long time, the utility of XP as a business information system development methodology is in doubt. Finally, the methodology requires considerable on-site user input, something that is frequently difficult to obtain.
