The sections that follow describe these four broadly different types of projects:
SWATAccomplish something relatively small but particularly difficult.
InvestigativeReport on the technology.
ProductionProduce a new or replace a legacy production system.
Full-CommitMove the entire organization over to objects.
Each has a different set of ground rules, and each requires a different set of staff skills. One of the survival recommendations is: Be clear and consistent about the purpose of the project.
SWAT, a term used by the police force, stands for Special Weapons and Training. It is an appropriate term for some software projects, because the purpose of a SWAT project is to get as much done as quickly as possible. Typically, this method is used on a small, difficult task. To approach such a project, take the following steps:
Relax other restrictions on the project.
Create the project with a clear and simple set of requirements.
Use a small, high-powered team, and give them whatever tools and training they need ("special weapons and training").
Free the team from phase deliverables and most of the intermediate documentation deliverables that accompany larger projects.
Let the team get their primary job done as quickly as possible.
The people you choose for a SWAT team should be able to "think outside the box." They need this kind of ability, or you would not need them for this project. Use two, three, or at the most, four people, sitting in close proximity to each other to avoid additional communication hazards.
Review Project Reginald (see Chapter 2) for a potential pitfall, that of adding more people as soon as the original team's work looks promising. A larger team has different needs and should be set up appropriately. If you wish to make the shift, declare the first project at an end, and set up for a Production project.
Suppose you wish to find out how well suited object technology is for your organization. The purpose of the investigative project is to detect the problems you may encounter, and to learn a way of working around them. Be sure to design and instrument the project to give you that information or you will not be any further ahead at the end of the project.
First, unless you are a small company, assign at least four people to the project. If you assign just two designers to learn object technology, you run the risk that they get trained, update their résumés, have fun, and produce a graphical prototype. What should you decide on the basis of this? Your development manager needs answers to the following questions, which your project setup team should take care to address:
How long will it take to train the average person in the department? How many people need to be trained? To what extent does the field staff need to be retrained to use the newer products?
How do we manage development of the project? Same as before or different? How do we control versions and share development information between groups?
What new tools do we have to buy? How many of the old tools can we use?
How do we test the new system?
What are the maintenance characteristics? When a change request comes in, how much improvement is there in turnaround time?
The project team should consist of four to seven people so that the communication issues show up. In contrast to a SWAT project, which should have only a few people to minimize communication needs, the investigative project needs enough people to encounter these communication issues. Experience indicates that OO development requires more person-to-person communication than structured programming (see Chapter 2). You want to discover how to synchronize the teams' versions of the software and documents, and to let them experience their interdependencies.
Staff the project with a business expert, one or two of your better programmers, two or three of your standard development staff, and an object technology mentor.
The business expert will turn your exercise into an investment. The team will design business objects right away, and the business expert will be the only person on the team who can properly describe the interdependence of various rules, attributes, and other elements of the business. Without this person, the rest of the team will design something that possibly will be quite good, but will not accommodate needed variations. The business expert will become accomplished at object modeling and will act as the team's ultimate information source.
The good programmer will ensure that the program gets written. A good systems or C programmer with a lot of programming experience will do well. She or he will be able to adapt quickly to the needs of object technology. This person may also write small tools, interfaces, or algorithms, or establish the team's programming style.
The two or three standard developers act as a test of the technology on the people who are going to have to adopt it wholesale. They will show where the technology is hard, easy, different, or the same. You will probably want to choose some of your better developers, both to increase the chance of success and because they will become team leaders on the next project. If their expertise and capability are too different from the rest of the company, the project will produce unreliable information.
The OO mentor's role is to provide a safe journey through this first project. The mentor must be comfortable with one or more ways of developing a system and be able to communicate with the team, to put them relatively at ease. She or he is responsible for safe design habits, and because her or his design habits and programming style are likely to be copied throughout the organization, they are important. The mentor need not be full-time, although half-time or more might be needed.
The charter of the project must include tests to identify the issues that will affect success in deploying the technology to the organization. These are some questions that may arise:
Do we know how to estimate a project? plan the project? test the system?
How much training and education is needed?
What is the transaction throughput?
What is maintenance like?
You might even decide that the team's first assignment should be to articulate the proper questions. A project topic that is a reasonable precursor of what will be asked of the development department needs to be selected.
Allow 6 to 10 months for the project. Six months is the minimum amount of time to give people to learn about the technology. Ten months is the most time that should be allowed for such an experiment.
Have the executive running the experiment prepare two project statements, the first to give to the group at the beginning to help them to learn the technology and to come to terms with the development process. When that time is up, the second project statement should be revealed. It should be a typical change request that affects the user interface, program logic, and database. The stopwatch runs from the time the second project statement is shown to the group until they deliver complete results. That is the way maintenance improvement claims can be judged.
At this point, review the successful, if aggressive, program adopted by Project Alfred (see Chapter 2). A small company may need to adjust the setup of its Investigative project. It may not have four to seven people to put on an experiment. Such a small company should use the SWAT approach to evaluate and learn the technology for the company. The SWAT team must explore various avenues of development rapidly and come up with recommendations for the rest of the group. Not only should this SWAT team consist of good designers, they also need to be articulate in passing on their recommendations, and accurate in their estimation of the company's capabilities. The team will set the company's programming standards and be the chief educators.
A Production project puts a real system into use. It is not an experiment; it must succeed, but it should not break the company if it is late. Whether it is a new system or a replacement for an existing legacy application, there probably will be a considerable number of interprogram interfaces to track. I assume here that the project involves both object and nonobject technology, and describe the legacy application version of the Production project. The only difference is the presence of additional experts on the legacy system.
Your project will probably select itself. It is usually clear which application needs replacing or creating, provides noticeable benefit to users, or does not endanger the company. One common hazard is that the users get so excited by the first glimpse of the new system that they overwhelm the development department with new demands.
The Production project will use a mix of people from the department because it is too large to staff with "superstars." The staff should be made up of experts on the existing system, standard developers who will get trained, and some object technology experts. The experts on the existing system will speed up knowledge transfer about how that system works or how the new system needs to fit in. They know the interfaces and can provide insurance against forgetting necessary functions.
You may decide to hire an external team of specialists to do the bulk of the work. This can be very effective: They are more expensive per day, but they work faster. I assume, however, that you will use your present developers to make use of and train your existing staff.
As explained in detail in the Day Care strategy in Appendix A, create two categories of teams. Put the expert developers into Progress teams to do most of the development work. Keep them as free as possible from distractionsteaching, in particular. Assign one or two of the experts to training the novices in Training teams. You may let the Progress teams do 90 to 95 percent of the system development, and assign the Training teams only 5 to 10 percent of the development work. The idea is to let the novices learn by doing, give them a dedicated teacher, give them time to deliver useful function, and keep them out of the way of the expert developers. You may find that the Progress team works 10 or 20 times faster than the Training team. Base the project schedule on the Progress team, not the total number of bodies. As the novices become adept with the technology, let them migrate to the Progress teams.
This strategy is in stark contrast to the more common one, which is to build mixed teams with three or four novices and one expert. The idea of mixed teams is to keep the student-to-teacher ratio low. The result that shows up repeatedly is that the expert does not have time either to design or to teach, and does not have a peer discussion partner. Thus, progress and quality both suffer. In the meantime, the novices do not have a full-time teacher, so they do not develop good design habits. Forming Progress and Training teams makes it possible to tackle the two issues separately.
A Full-Commit project needs your best manager and technical leader, a group of really good programmers, a team of standard developers, and one or two full-time object experts. Because the project will have a large number of developers pulling in different directions due to differing philosophies, your technical leader must be respected and persuasive to be able to align them in whatever technical direction gets selected.
Your really good programmers will learn the nuances of OO design quickly and become development leaders; the standard developers are simply your regular staff. The one or two full-time object experts will provide the training, the standards, and expertise to the rest of the organization. They will work with the technical leader to establish the organization's techniques, standards for design, coding, testing, training, and so on. The object experts will help establish and disseminate your organization's new methodology. Plan on using them for one to two years.
The Full-Commit effort should start with one or more projects that are important to your organization, but that, if delayed, will not bankrupt it. Not every company has this luxury, of course. Use incremental development to give yourself space for course corrections, and trust your best manager to pay close attention to the health of the projects.
Other Project Categories
Although I classified projects by size and purpose in this chapter, other classifications are possible. Information Systems (IS) groups and their projects differ from scientific/engineering groups and their projects. Real-time projects differ from end-user application projects, and so on. Their differences show up in technology selection and methodology rather than in setup. Whether you are in an IS or an engineering department, creating an application or a real-time program, you may choose a SWAT project, investigate, or you may move the whole organization.
How bold should the project be? Should you choose a safe topic, or one on the edge of possibility? Your organization will probably do on this project much as it did on its previous project. If your group is used to large projects or new technology, then a major step into object technology may be within the team's comfort zone. If, on the other hand, your organization is used to making small changes, then you will probably choose a safe topic.
Brooklyn Union Gas launched a 150-person, mission-critical project as its first venture into object technology, partly because the organization was used to large-scale projects (see Chapter 2). If you select an unusually aggressive project, be sure to include an extra risk factor just for the shift in the boldness. To counter this risk, add someone to the project who has experience with bolder projects. This theme is picked up again later in the Technology section of this chapter.
Ultimately, the size and boldness of the project will be determined by the expertise of the people you can bring in at various times and costs. The next section discusses this.
Let the charter and staff fit the purpose.