Judging the Quality of Models
Here are some general criteria for judging the quality of models.
Simplicity. A model should not be needlessly complex. Many applications can be modeled with no more than 50 entity types. Few applications require more than 250.
Names. The names in a model should be crisp and meaningful. If a name confuses business experts, change the name. It is difficult to devise good names, so don't be surprised if you encounter lengthy discussions.
Scope. A model must fully address the requirements, but not greatly exceed them. At least 80 percent of a model should pertain to immediate needs. As much as 20 percent can anticipate future needs.
Extensibility. A model should be readily extended. New requirements should cause additions, but little alteration. If you add 20 percent new entity types to a model, you should not need to revise more than 20 percent of the original model.
Visual display. The visual layout should make a model easier to understand.
Documentation. A model should have a thorough written explanation. The explanation should define important application terms, explain the rationale for subtle decisions, and walk the reader through the model. Typically, every five entity types should require one page of explanation.
Review. A model is more likely to be of high quality if it has been subjected to vigorous review.
Bear in mind that it may take 510 revisions before a model converges and satisfies these criteria and essential business needs.