Data model patterns: A metadata map

Read this excerpt from Data Model Patterns: A Metadata Map for a guide to all things metadata. That is, data about data.

The following is an excerpt from Data model patterns: A metadata map, by David Hay.

About Metadata Models

There once was a fellow named Corey
Whose career was not covered in glory
He had a bad day
When he just couldn't say
Me-ta-da-ta Re-pos-i-TOR-y.

What are metadata?*

During the 1990s, the concept of data warehouse swept the information technology industry. After many years of trying, it appears finally to be possible for a company to store all of its data in one place for purposes of reporting and analysis. The technology for doing this is still new, and the first attempts have had mixed results, but the effort has been quite serious.

One of the problems that arose from this effort was the realization that if a senior executive is going to ask a giant database a question it is necessary to know just what is in the database and what types of questions to ask. In addition to the data themselves, therefore, it is necessary to keep data about the data. The term coined for "data about data" during the 1990s was metadata.

Since then, numerous books and magazine articles have been published on this subject, but most have focused on why metadata are important and on technologies and techniques for managing them. What these publications have left out is a clear description of exactly what the stuff is. After a decade, there is still no simple, clear description of metadata in a form that is both comprehensive enough to cover our industry and comprehensible enough that it can be used by people. This book is an attempt to produce such a description.

As with all buzzwords, once invented the term metadata has taken on a life of its own. It is variously described as:

  • Any data about the organization's data resource [Brackett 2000, p. 149].
  • All physical data and knowledge from inside and outside an organization, including information about the physical data, technical and business processes, rules and constraints of the data, and structures of the data used by a corporation [Marco 2000, p. 5].
  • The detailed description of instance data. The format and characteristics of populated instance data: instances and values, dependent on the role of the metadata recipient [Tannenbaum 2002, p. 93].

Several significant points come out of these definitions. First, as Mr. Marco pointed out there is a difference between business metadata and technical metadata. The business user of metadata is interested in definitions and structures of the language as terms for the types of information to be retrieved. The technician is concerned with the physical technologies used to store and manage data. Both of these points of view are important, and both must be addressed.

Second, the subject is concerned with more than just data. It is, as Mr. Brackett said, "any data about an organization's data resource." Once you have started looking at the structure of an organization's data, you have to also account for its activities, people and organizations, locations, timing and events, and motivation.

Third, as Ms. Tannenbaum pointed out, the "meta" aspect of the question is a matter of point of view. There is metadata relative to the data collected by the business. There is also meta-metadata, which is used to understand and manage the metadata.**

This last point is illustrated in Figure 1–1. Here, the bottom row shows examples of things in the world that are often described in information systems. "Julia Roberts" is a real human being. The "Wall Street branch" of a bank is a physical place were business is performed. Checking account "09743569" is a particular account held in that bank by a particular customer (Julia Roberts, for example). The customer of that account may then perform an actual "ATM Withdrawal" at a specific time.


This Book (Metametadata) Elements of metadata (metadata model) Objects: "Entity Class" "Attribute" Objects: "Entity Class" "Attribute" "Role" Objects: "Table" "Column" Objects: "Program module" "Language"
Data Management (Metadata) Data about a database (a data model) Entity class: "Customer" Attributes: "Name" "Birthdate" Entity class: "Branch" "Employee" Attributes: "Employee.Address" "Employee.Name" Role: "Each branch must be managed by exactly one Employee" Table: "CHECKING_ ACCOUNT" Columns: "Account_number" "Monthly_charge" Program module: ATM Controller Language: Java
IT Operations (Instance Data) Data about real-world things (a database) Customer Name: "Julia Roberts" Customer Birthdate: "10/28/67" Branch Address: "111 Wall Street" Branch Manager: "Sam Sneed" CHECKING_ ACCOUNT. Account_number: = "09743569" CHECKING_ ACCOUNT. Montly_charge: "$4.50" ATM Controller: Java code
  Realworld things Julia Roberts Wall Street branch Checking account #09743569 ATM Withdrawal

Fig. 1–1: Data and metadata.

The next row up shows, in the first three columns, the data that might describe those three things: (1) A Customer has the name "Julia Roberts" and the "Birthdate" of "10/28/67". (2)A Branch has the address "111Wall Street" and a manager, "Sam Sneed". (3) The checking account has an account number "09743569" and a monthly charge, "$4.50". In the fourth column, the first row from the bottom shows that a particular program, called here "Java code", is responsible for a "Withdrawal Transaction". These are the things that would concern a person managing data for a banking business. Note that each of the terms was described as to what it was: customer name, branch manager, account number, and so forth.

The third row from the bottom collects those descriptors and labels them in turn. This is to create what we in the data administration world call the metadata. There are two components to these labels. First are the names of the things of significance being described by the business data, such as the entity classes "Customer" and "Branch". Second, each of these is in turn described by attributes, such as "Name", "Address", and "Birthdate". We also discover, in the case of the bank branch, that there is really an additional entity class, "Manager", and that it is related to "Branch". ("Each Branch must be managed by exactly one Employee.")

In the checking account column, we see that a checking account is actually the subject of a table in a database. The table is called "CHECKING_ACCOUNT" and has columns "Account_number" and "Monthly_charge". The ATM program described in the second row simply as "Java code" is actually a program module with the name "ATM Controller" written in the language "Java". As we can see, the metadata row itself encompasses several different types of objects ("Entity class", "Attribute", "Table", "Column", "Program module", and "Language"). The assignment of this book, represented by the top row, is to show how these objects relate to one another.

Metadata don't just describe data. They describe how the organization understands not only its data, but also its activities, people and organizations, geography, timing, and motivation. Yes, metadata describe the entity classes and attributes of an entity-relationship model, and the tables and columns by which these are implemented in a computer system. They also provide, however, structure for describing the activities of the organization and the computerized processes that implement these activities. They describe who has access to data, and why. They describe the types of events and responses that are the nature of an organization's activities. They describe where the data and processes are, and they describe the motivation and business rules that drive the entire thing. So, from all of this comes the following definition of metadata.

Metadata are the data that describe the structure and workings of an organization's use of information, and which describe the systems it uses to manage that information.

One anomaly has revealed itself in the line between business data and metadata. The information about what constitutes a legal value for a product category or an account type in the business model is often captured in separate reference tables. To reflect these validation structures, a typical data model often has many "type" entity classes (account type, status, day of the week, and so on) describing legal values for attributes. These are part of the business data model.

But because they are in fact constraints on the values of other attributes in the same data model, they are also included in the category of metadata. Where a table designer would be required to specify the domain of a column, the data modeler (who is instructing the designer) must now provide the values that constitute that domain. Here you have business data acting as metadata.

Be aware, of course, that even this line between business data and metadata is not as clear-cut as it seems. product type, for example, is about reference data that constrain many attributes in a business model. Even so, specification of the list of product types is very much the domain of the business, not the data administrator. This plays both the roles of business data and metadata. Probably more in the metadata manager's domain would be product category. There should be relatively fewer of these, and the list should be relatively stable.

*Ok, it's true. I studied Latin in high school and have always held that data is the plural form of the word datum. I realize that I may be swimming against the current, but, hey! It's my book!

**While delivering a lecture on cosmology one day, Sir Arthur Eddington gave a brief overview of the early theories of the universe. Among others, he mentioned the American Indian belief that the world rested on the back of a giant turtle, adding that it was not a particularly useful model as it failed to explain what the turtle itself was resting on. Following the lecture, Eddington was approached by an elderly lady. "You are very clever, young man, very clever," she forcefully declared, "but there is something you do not understand about Indian cosmology: it's turtles all the way down!"

Copyright info

Printed with permission from Morgan Kaufmann, a division of Elsevier
Data Model Patterns: A Metadata Map
By David Hay

Publication Date: 23 June 2006
ISBN: 0-12-088798-3
For more information about this title and similar books, please visit


This was last published in July 2006

Dig Deeper on MDM best practices



Find more PRO+ content and other member only offers, here.

Start the conversation

Send me notifications when other members comment.

By submitting you agree to receive email from TechTarget and its partners. If you reside outside of the United States, you consent to having your personal data transferred to and processed in the United States. Privacy

Please create a username to comment.