Avoid drawing tools that aspire to be data modeling tools. A good data modeling tool supports defining tons of metadata with business relevance. Think of the diagram as just the tip of the iceberg – where you don’t see the 90 percent of the mass that is underwater. The same is true for data modeling. If you concentrate only on what the picture is, you’ll probably compromise the effectiveness of the resulting database.
Choose a tool that fits your needs. Often, people purchase a killer modeling tool that offers everything imaginable. But, if all you need or will use is the data modeling portion, why pay for more? The key concern here is that the more any tool does besides data modeling, the better the chance its data modeling capabilities may have been compromised to do everything else. Sometimes more is not better.
Data definition language. This is another case where more might not be better. It is better if your tool supports 100 percent accurate CREATE or ALTER scripts for a few databases important to you than all of them at a lesser level. But be very careful - the DDL generated by many tools, even those focusing on just a few databases, can often generate less than optimal DDL. You have to know what to look for; so, engage your database administrator in making the decision, just to be safe.
Verify that your data modeling tool provides robust model consistency and accuracy checking reports and/or utilities. As data models grow (and they will), it can be quite overwhelming to have to manually check everything. And you cannot expect the poor DBA to sanity check the thousands or tens of thousands of DDL lines a data modeling tool can quickly generate. Effectiveness is mostly on your shoulders, but efficiency can be aided by good data modeling checking utilities.
