When asked the question, “What is an Entity Relationship Diagram (ERD)?” the typical response I hear is “It’s a graphical representation of a database.” Methodologists and data folk have generalized the most recognizable deliverables of my line of work into this simple sentence.
As the quote “A picture is worth a thousand words.” implies; there is real power in the visual representation of data. It is the thousand words that give more substantial meaning and value to the ERD. Here are some of my thoughts on the importance of and usage of text on ERDs.
Use the principles of desktop publishing.
Find a readable font and use it on all of your objects. Multiple fonts are confusing and make reading the model much more difficult. Watch your case and use upper, lower and mixed case consistently. Your users will quickly key in on your usage pattern and what it means. Bold is good for highlighting important things. Use it wisely and don’t overuse. Underlining is just not a good idea. It makes words hard to read, and I avoid using it.
Relationships do need verbiage
It is not acceptable to build a data model with no text on the relationship lines. Relationships define the business rules in your ERD. Omitting them gives the reader no clue what the relationship means. They should read like a sentence. I prefer to include both a parent-child and child-parent phrase. i.e. “A person pays the invoice.” and “The invoice is paid by the person.” Be consistent on the position on the relationship line. I like to put constraint names on the physical relationship line.
Legendary text leads the way
An often overlooked object on the data model is a legend. Almost all models for a corporate application are pretty complex with many entities. Color coding by business area, application functionality, or the source/target helps model users understand the ERD. This value is minimized or lost when the data modeler fails to include a legend explaining the color coding scheme. Make it a practice to include a legend in a consistent place to lead the model viewer through your ERD.
You may need to say something more
There are times in data modeling when text on relationship lines and in a legend does not tell the whole story of your data design. Complex business rules, domain of values, and other business and technical items of interest can be documented in the metadata under the data objects. Read your model with the eyes of one of your model users. If your data story does not come through clearly, you may need to include text on the diagram. I use value added text sparingly. I prefer floating text rather than text in a box or circle. The ability to tell an accurate story is your main goal.
Learn more at the Data Modeling Zone
The above guidelines come from my Data Modeling Zone 2014 presentation, Is your data model a work of art?. The session is geared to a beginner’s audience. The focus is on tips to make your ERD more readable and understood by your users.
I am presenting at Data Modeling Zone 2014 in Portland, Oregon. I hope that you will join me in my sessions: Is your data model a work of art? and Relationship versatility and the data modeler #DMzone