An old fantasy permeates across some companies that Information Technology is some magical entity that can produce instant software in a vacuum. This has never been true, of course. It takes time to code programs--as well as time to figure out what needs to be coded in the first place. This fact is frequently ignored by a company's internal business units, each demanding the full attention of an overburdened IT department. The history of business throughout the existence of computing resources is wrought with examples of unreasonable deadlines with no allowance for developer availability and scope. For most companies, IT is a cost center and often must fight for appropriate resources or tools with which to work. With companies tightening their belts and placing increased scrutiny on capital expenditures, the fantasy of the IT wizards is perpetuated. IT is integral to the success of any modern business, and, though it may not produce revenue, it is vital that the product produced by applications programmers be relevant and reliable. Shops need to find ways to run efficient operations, and one way to help reach that goal is to better coordinate and control the software development process.
With no external customer to market a product to, the goal of an application development team must be to provide the user with a timely and reliable solution that has a positive impact on the bottom line. These days, the role of the IT employee is as much business analyst and customer advocate as it is technical expert. In addition to software development skills, applications programming today requires a well-rounded knowledge of the business, of project management, and of techniques that can be used to competently produce low-cost solutions.
Certainly, nothing in life is without risk, and some efforts are best left to chance, but in the IT world, as much or more so than any other arena, it is best to approach everything with a solid plan. The goal of every organization should be to bring a project in on time and within budget while keeping "scope creep" to a minimum. The fundamentals of project management are well beyond the range of this article and are thoughtfully described in a recent article by Steve Collins. They are mentioned here because of their importance, but even proven management techniques are useless unless applied to some development model. Too often, the route of choice for developers is to take a request for a new module or system and hammer out code extracted from their imaginations. This approach can be fun, but the results are shoddy, bug-infested systems that can end up costing a company more money for patches than it would have cost to develop a sound system to begin with. We often joke about how there is never time to do something right to begin with but always time to fix it. Unfortunately, this is a true statement in many companies in many industries.
The standard operating procedure in many shops is the same. A user or a group of users will request a new system or extensive modifications to old systems and will allow no time for IT to lay the foundation for a proper development effort. With IT as an afterthought in many industries, there is little understanding of the fundamental elements that should be in place for solid application development. IT personnel may have an understanding of how to structure their efforts, but to accommodate the request they are compelled to cut corners.
Several standard development methodologies can be adopted or adapted. They are not difficult to comprehend, and the steps, once they are laid out and understood, are easy to explain and follow. The three approaches I will discuss here are not the only three available, but they are among the most common and can be applied to most development needs.
The Waterfall Model
The top-down, or waterfall, approach (Figure 1) is a sequential process in which each step of the methodology is dependent upon the one before.
Figure 1: The waterfall model of software development
There are variations to the sequences and necessary documented artifacts, but essentially the methodology begins, as all software development should, with a period of requirements gathering. This is the most necessary--and frequently the most ignored or abused--element of software development. It involves a heightened level of contact with the users and mandates that a software developer possess a certain level of analytical skill in order to take a user's request and turn into a coherent concept. After the functional requirements are determined, the individual software modules are designed based on those specifications and then built and unit tested by the developer. Following that, the software is tested as a system. Once that level of testing is complete, the system is turned over to the users in a dummy environment, and they test the software as if it were live. If all goes well, the new modules are moved to a production environment, and any necessary documentation is rolled out to the users.
The waterfall methodology provides excellent structure to the endeavor and makes it very easy to gauge progress toward a specified deadline. You can easily determine how far off schedule you may be because the model is linear: One phase must be completed before the next one begins. For instance, you cannot test a module without building it, you cannot build it before it is designed, and you cannot design without first determining what the user really needs. The top-down approach does presume that all elements can be determined up front and that a comprehensive design can be created prior to construction. This is not, however, always the case, and this represents perhaps the single biggest drawback of the model. Change in any system development undertaking is inevitable, but the waterfall model makes it extremely difficult to implement. A strategy for handling requirements changes is a must for any project. Though this method is easy to follow, you may not want to implement it if your requirements tend to be dynamic.
The Spiral Model
The spiral model (Figure 2) is less rigid than the waterfall during the requirements gathering phase. Essentially, the initial requirements are followed by a prototype, which is then presented for review. Feedback is incorporated into revised requirements and a modified prototype, which is presented again for review. This process continues until there is a clear understanding of requirements, scope, and associated risks. From this point forward, the phases of coding, testing, and implementing are not fundamentally different from the waterfall approach.
Figure 2: Basic elements of the spiral method
While the spiral model may facilitate a more comprehensive requirements gathering effort, there is a danger that IT and the users can get caught up in the prototyping cycle and ignore or undercut the rest of the process. The scope of the requirements gathering using the spiral approach is generally greater than the time allotted for completion.
The Iterative Model
The third approach, the iterative model (Figure 3), is a relatively recent development that has become very popular because of its simplicity and its flexibility. It is a departure from the other two models both in execution and nomenclature. The iterative approach consists of an inception phase, an elaboration phase, a construction phase, and a transition phase. These phases are executed in that order, but individual tasks are not reserved for a particular phase.
| Inception | Elaboration | Construction | Transition | ||||
Requirements | X | X | X | | | | | |
Design | | X | X | X | X | | | |
Development | | | | X | X | X | X | |
Testing | | | | | X | X | X | |
Implementation | | | | | | X | X | |
Change Management | | | | X | X | X | X | |
Project Management | X | X | X | X | X | X | X | X |
Figure 3: Overlapping tasks in the iterative model
This approach operates in many ways like a combination of the spiral and waterfall approaches except that within each phase, as illustrated, the activities could overlap. During the inception phase, for instance, requirements gathering and design will consume the bulk of the effort over development and testing. The opposite would be true of the construction phase. Like the waterfall approach, the steps to address within each phase, as much as that phase requires them, are requirements gathering, design, development, testing, implementation, change management, and project management. In the iterative model, these steps are known as "workflows."
Test, Test, and Test Again
Regardless of the model chosen for software development, no software module or system should be promoted from development into a production environment without sufficient testing. While this may sound like common sense, the fact is that many testing efforts fall well short of what is needed to ensure the development of a quality product. A comprehensive test script should be developed for each processing scenario. These documents should be developed from the accepted requirements. A test script can be very elaborate if you choose, or it could simply be three basic columns that represent a sequential list of steps, the results expected from the steps, and a record of the results observed. If one step fails, the whole test fails.
Once the developer has tested the individual modules and deemed them operational, it is time to involve users. They should have been intimately involved in the requirements gathering phase in the first place and, therefore, have a concrete understanding of what they should be looking for. They will also have less bias toward a specific program object than the programmer who has just spent days, weeks, or months working on it. Get the users involved in developing the test scripts as well as validating the system.
Document (Almost) Everything
No matter which style of software development you use, it is good idea to provide a solid set of documentation throughout the development lifecycle. This documentation generally begins with a high-level project charter that describes the reasons for the development effort and the business justification for it. The documentation should contain the signatures of representatives from all areas affected by the new or modified software. Likewise, the requirements should be documented as well and contain a similar set of signatures. If you deal with other departments without sufficient correspondence or documentation, you're likely to find that those points you took as gospel are in fact open to interpretation. A documented set of requirements with appropriate approvals should not be regarded as more red tape. It is a concrete method of securing understanding among all parties.
The development of tangible design documents is a must. Depending on the scope of the development effort, you may have two levels of design: one that provides a birds-eye view of the architecture and basic elements for the users to see and a second that contains detailed designs for each object to be developed. Smaller projects may not necessitate this two-tiered documentation step but will employ a single-level design or set of designs corresponding to the individual modules. Designs come directly from user requirements but provide specific information, such as inputs, outputs, processing requirements, and perhaps program logic.
Always remember that the software being designed has a customer in mind--usually the internal user group making the request. They have a right to expect consistent correspondence regarding project status and schedule, and they should be provided sufficient user documentation on the back end of the lifecycle--if not before--for them to understand and operate the software.
Continuing Education
Understanding how a methodology works and how to apply it is one thing. Selling your company on developing one and then implementing it successfully is quite another. Because of the tedium frequently associated with the related red tape, many users balk at having to take the extra administrative steps generally required by a structured lifecycle model. Even more unfortunate are those individuals within IT itself who would rather return to the days of on-the-fly development in order to lessen the friction with the user community caused by a lengthy development process.
One way of getting users to work effectively within a lifecycle model is to involve them in the creation and implementation of one. This will help them see that the methodology is being developed as much for their benefit as for the software developer's. Since it is true that IT and the user community frequently endure rocky working relationships, helping to educate the users is also an excellent way of fostering better understanding.
Embracing a software development model makes good sense for an organization from all angles. Chances are that the elements are already in place in your company, and the concept just needs a sturdier formation. Whether you start from scratch or create a development strategy from existing practices, the benefits to both the users and IT are worth the effort involved to provide enhanced structure to your development environment.
Steve McGuire is an author and midrange applications developer in North Carolina. Steve can be reached at
LATEST COMMENTS
MC Press Online