The greatest problem in application design is a lack of forethought. As it applies to database-driven applications, the design process must include a thorough evaluation of your database?what it should hold, how data relates to each other, and most importantly, whether it is scalable.
The general steps in the design process are
Define the objective.
Design the data structures (tables, fields).
Define and implement business rules.
Create the application.
Creating the application is the last step?not the first! Many developers take an idea for an application, build it, and then go back and try to make a set of database tables fit into it. This approach is completely backward, inefficient, and will cost a lot of time and money.
Before you start any application design process, sit down and talk it out. If you can't describe your application?including the objectives, audience, and target market?then you're not ready to build it, let alone model the database.
After you can describe the actions and nuances of your application to other people and have it make sense to them, you can start thinking about the tables you want to create. Start with big flat tables because, after you write them down, your newfound normalization skills will take over. You will be able to find your redundancies and visualize your relationships.
The next step is to do the normalization. Go from flat table to first normal form and so on up to the third normal form if possible. Use paper, pencils, sticky notes, or whatever helps you to visualize the tables and relationships. There's no shame in data modeling on sticky notes until you're ready to create the tables themselves. Plus, using them is a lot cheaper than buying software to do it for you; software ranges from one hundred to several thousands of dollars!
After you have a preliminary data model, look at it from the application's point of view. Or look at it from the point of view of the person using the application you're building. This is the point where you define business rules and see whether your data model will break. An example of a business rule for an online registration application is, "Each user must have one e-mail address, and it must not belong to any other user." If EmailAddress wasn't a unique field in your data model, your model would be broken based on the business rule.
After your business rules have been applied to your data model, only then can application programming begin. You can rest assured that your data model is solid and you will not be programming yourself into a brick wall. The latter event is all too common.