In the most general terms, a prototype is a small test program designed to tell you something about another larger project that you have not yet built. You build the prototype and then use the information that you learn by building and using it to predict things about the bigger project. A prototype can also let you try different approaches and perfect development techniques. The following are some of the most important uses of a prototype:
Define application functionality
Test application architecture
Test untried technology
Predict project duration
Predict needed staffing
Determine project feasibility
Refine development techniques
Build the development team
The following sections describe these prototyping benefits and tell how you can maximize their value. Some of the information that you can gather during a prototype has subtle uses, so it's not always obvious how to use what you learn. If you don't keep some of these in mind while you develop the prototype, you may miss out on important clues.
This information is at least as important for team leaders, project managers, and customers as it is for developers. For example, if the developers build a prototype containing too much user interface code, they will probably give you low estimates of the time and staffing needed to finish the application. You need to be aware of these issues so that you can keep developers on track and so that you know how much to trust their results.
Because some of these issues are subtle, you need to pay close attention to them during prototype development. Make a list of the things you hope to learn from the prototype, and look over this list frequently. Be sure to share the prototype's goals with the other developers and customers. If everyone stays focused on the goals, you can greatly increase your chances of success in relatively little time.
Probably the most important use of a prototype is to define the application's functionality. The goal is to ensure that the customers and the developers have the same view of what the final application will look like.
The project begins with a statement of what the application will do. The preliminary design suggests a possible method for performing the application's key tasks. A prototype solidifies that preliminary design so that everyone can see what the developers have in mind.
The prototype shows what the application will look like and, to some extent, what it will do. At a minimum, the prototype should let the developers perform a carefully scripted walkthrough to explain the program's most important features to the customers. Ideally it also demonstrates those features so that the customer can experiment with them and get a feel for how they will work in the final product.
Nonprogrammers often have trouble understanding what programmers mean when they describe a program. A programmer may say, "This form will display inventory in a grid control," and the users may think they understand what that means, but they often have trouble really understanding how that applies to their work until they see the form in action. A working prototype is the best way to give the customers a really good appreciation of what they are buying.
After they have seen and possibly experimented with the prototype, the customers can give feedback. The developers can adjust the development plan to make the product better fit the customers' needs. This greatly increases the chances that you will produce a product that the customer can actually use. This is one of those rules that is completely obvious but often ignored by developers. It's amazing how often developers build a product that no one wants.
Meeting with users gives developers more motivation because they see that they are building something that will actually be used by someone. They often get caught up in the customer's enthusiasm and want to do the best job possible. Working in isolation may be useful at times, but developers should stay focused on the end result.
If user feedback leads to major design changes, the developers can build a new prototype to let the customers see what the revised application will look like. You should repeat this cycle as often as necessary to ensure that the customers and the developers all agree on what the final system's functionality.
Incorporating feedback into new prototypes lets the customers know that you are listening to them. It shows that you take their concerns seriously, and is a great way to win customer buy-in. A customer who makes a personal contribution to the project is more likely to do so in the future and is more likely to struggle for the project's success.
Tale from the Trenches
One system we built was designed to assist dispatchers in routing several hundred repair trucks to their daily jobs. The work center where this was going to be installed was full of fear and resentment. Dispatchers were afraid they were being replaced and might lose their jobs. Operators and drivers didn't want to participate in something that might get their coworkers fired.
We picked out the dispatcher who was most traditional, was most antiautomation, and was totally convinced that our system would never work. It couldn't possibly do a better job than he did! We sat him down in front of a prototype and showed him how to use it. We listened to his feedback and adjusted the prototype as quickly as possible, sometimes the same day. After a few sessions, he realized that this was not a replacement for him but a tool he could use to do his job better. He saw that we valued his input, and he became one of our strongest supporters. That convinced everyone else. After all, if Skeptical Steve believed in the project, then it must work!
In some applications, defining the functionality the customer needs can be difficult. Customers may not know exactly what they need, or they may think they need one thing when they really need something else. Building a prototype can help clarify these needs. Sometimes the project definition emerges incrementally as the customers see a series of prototypes and make evolutionary adjustments to each. Other times after seeing an initial prototype, the customers realize what they really need, and the design may take a completely new approach. In any case, prototypes let you get feedback early and often.
When developers and customers have agreed that a prototype represents the system's functionality, the prototype can serve as a living specification. When customers have a question about what the final system will do, they can run the prototype and see.
Equally important, developers can look at the prototype to see what the final program should do. This lets them concentrate on writing code rather than trying to figure out what the program should do. It also helps prevent mismatches in which one developer thinks a feature should work one way while another developer thinks it should work another way. If everyone understands what the prototype does, they will not need to argue about how things should work.
By defining the application's features, a good prototype lets several different tasks move forward in parallel. That lets you apply more people to the development effort at the same time and thus reduce total development time. The following sections describe some of the more important of these tasks.
With an approved prototype in hand, you can begin preparing a detailed specification. Even the best prototype leaves some questions unanswered, and the specification answers them. Writing the specification is a lot easier once the prototype is complete because it gives everyone a clear idea of how things will work. After all, it's a lot easier to write about how an existing prototype works than it is to describe how a program that doesn't exist might work.
You can streamline the specification by referring to the prototype. Instead of describing every detail about how a scrolling list of inventory items works, the specification can state that it works as demonstrated by the prototype. A programmer or customer who has a question about some detail during development can simply fire up the prototype and see.
The specification can also state that one part of the system will be like another. Some large business applications contain dozens or hundreds of nearly identical reports. They are all very similar in format although they display different data. In that case, the prototype can demonstrate one report, and the specification can state that the others are similar.
If you use a series of prototypes, the specification also should state the version referenced. For example, it might say the Work List Screen will look like the one in prototype version 0.09.
The specification should explicitly point out any features in the final application that differ from the prototype. Those are the main points where confusion might arise. Even with small features, if customers expect one thing and developers deliver another, the customers may feel betrayed. If the prototype cannot be trusted about this, how do they know that they can trust the prototype about other things?
In one project, our team built a prototype and then transferred it to another part of the company for implementation. That team went off on its own for several months and created what it considered the final application without any feedback from the customers or the prototyping team.
When the customers saw the final product, they noticed that the implementation group had made several "improvements." Our project leader sat through a series of meetings in which the customers hammered the implementation team:
CUSTOMER: With this version of the system, how would I do this?
DEVELOPERS: Uh, we didn't think of that. I suppose you could use this roundabout method to do the same thing.
CUSTOMER: What if you changed the application so that it worked this way? It would be a lot easier.
DEVELOPERS: I guess that would work.
CUSTOMER: That's the way the prototype did it. Change it back.
In the end, the developers spent a lot more time making the program work the way it did in the prototype. The customers were very unhappy. They didn't see why they had paid us to build a prototype, paid the development team to build the final application with improvements, and then paid the development team again to remove the improvements.
When the customers accept the prototype, the developers have a fairly good idea about what the final application should do. You can then start designing the application's architecture and building a development plan. Some of the fine points may change as the detailed specification is fleshed out, but the overall design should be clear.
At this point it is important to remember that the final program does not need to do things the same way the prototype did. The finished product should do what the prototype does, but it can use different methods. Prototypes are often built using quick-and-dirty methods that are effective in the short term but that may not be easily maintainable over time. If you blindly follow the prototype's methods, you may be stuck with a poor design.
Step back and think about the problem again from scratch so that you can build the best architecture possible.
After seeing the prototype, you may be able to tell what tools you will need to build the final application. For example, you may know that you will need an assortment of database development tools. Even before the main programming effort begins, the database team can start buying or building those tools.
Documentation and Training
User documentation and training depend only on what the program does, not how it works. This means that you can begin preparing user documentation and training material as soon as the customer accepts the final prototype. These tasks always take a lot longer than developers expect and are often left to the last minute. The result is a product with documentation that is weak, incorrect, or missing entirely.
When you have a final prototype, you can begin preparing documentation and training materials. However, not all prototyping approaches produce a single final prototype. In one method, developers build an initial prototype and get customer feedback. They build a new prototype and get more feedback. The process repeats until the prototype does everything the final application should do. In that situation, documentation has to wait until features are finalized.
I've worked on several projects that used evolutionary prototyping, and it can be very frustrating to the documentation and training people:
TRAINING: Can you give me a rundown of how the user assigns an employee to a department so I can write it up?
ME: Sorry, we don't know that yet.
TRAINING: Can you at least tell me where the program gets the list of departments?
ME: Nope. Perhaps in the next prototype.
Some companies require the developers to build a complete user interface prototype before starting the real programming. Documentation and development progress simultaneously. Although this makes it harder to accommodate last-minute changes, it makes it a lot easier to release a program with a good set of documentation.
A prototype does not necessarily implement features in the same way the final application will. The prototype may take huge shortcuts that make it less useable except in very specific instances. The goal here is to give the customer a feel for the application, not to build it yet.
Many programs begin with a sign-on form in which the user enters a user name and a password. The prototype may simply compare the values entered to a user name and password stored in the code. The final application will need to use a more flexible and robust system, such as trying to log on to a database using the entered values.
Prototypes are often short on formatting and error-handling code. For example, phone number fields may not verify that entered values look like valid phone numbers, so the user will need to use the right format.
Even so, the prototype should not be too stupid. It should save an invalid phone number into its test database, or at least present a somewhat meaningful error message, rather than crashing without warning. The prototype provides the first good chance for the customer to see what you are building. Making a good first impression is important, so don't mess it up with dozens of error messages and crashes.
If the prototype is too fragile to work safely, don't let the customer experiment with it. Instead, demonstrate it yourself following a thoroughly rehearsed script. Be sure you know what will work and what won't, and don't do anything that you haven't tried before.
First impressions can mean life or death to a new project. One application I built for the State of Minnesota began with a logon form shaped like the State of Minnesota—not just a rectangular form with a picture on it, but a form cut out to resemble the state.
When the users saw the first prototype, they loved it. When the demonstration was over, the first thing they asked was, "How did you make that logon screen?" They thought it was so cool. They also had lots of useful feedback, and we made a lot of changes to the design, but that one form had sold them. A little razzle-dazzle can make a powerful first impression.
One danger of a good prototype is that it may raise the customers' expectations. A good prototype looks like an almost-finished program. After seeing a demonstration, the customer may say something like, "The prototype looks great! Just fix up those few bugs and formatting problems in the next week or so, and we'll use that."
The customer doesn't understand that the prototype is like a movie studio set: It looks great in a carefully controlled demonstration, but behind the facade it's just a hollow shell.
Even if you convince the customers that you cannot just "fix up" the prototype, they may have trouble understanding why it will take you nine months to take the prototype from "almost done" to "done." You should warn customers of the prototype's "hollow" nature before you show it to them so that you don't raise their expectations prematurely.