Could They Build the Sears Tower Without a Blueprint?
	Why Data Modeling Is Imperative
	By Larry English President, Information Impact International, Inc.

This article is copyrighted by PLATINUM Technology, Inc., and was originally published in the PLATINUM Edge: Magazine. PLATINUM Technology, Inc., Volume 2, Summer, 1995. All rights reserved.

Introduction

I started my career in "data processing" at Sears in 1973 when the construction on the Sears Tower was nearing completion. My signature is on the top beam of the Tower, some 1454 feet above ground level, in what is still the world's tallest office building. The Tower has a fascinating architectural plan that calls for nine inter-linked sections or tubes to form a strong, stable, yet flexible structure that allows its 110 floors to sway safely in the Windy City. The Tower has reminded me of the requirement for data modeling during the past 15 years of helping organizations apply management principles to information as a strategic business resource. What if, I thought, the Sears Tower had been architected like some organizations design their databases? The thought is truly alarming.

Just as a complex building requires a guiding architectural plan, building a complex data infrastructure requires a guiding data model. When performed properly, data modeling is valuable and absolutely essential for effective management of the enterprise's information resources. It is required for both effective shared databases with integrated information systems, and for effective data warehouses with comprehensive decision support systems.

The case for data modeling is a case for information management. Information management is not possible without a well-defined data model that accurately portrays an enterprise's knowledge requirements and controls database design.

What data modeling is

Data modeling is not simply defining data for an application. It is not developing a logical database design for an application.

Data modeling is the process of analyzing and representing the "things" (entities) about which an enterprise must know.¹ A data model represents the business facts (attributes) that the enterprise must know about those business entities, along with their relationships and business rules. Data modeling requires business personnel to wrestle with and achieve consensus on the definition of data -- the business vocabulary -- with the facilitation of a data architect.

The data model as an architecture

A data model is an architectural blueprint that enables stable and flexible database development. Stability means new applications can reuse the database(s) directly or by simply adding (but not modifying existing) new entities, attributes and relationships. Flexibility means the database(s) can support changes to how the business accomplishes its business processes without major modification. A stable and flexible data model:

• Reduces the costs of application development and maintenance through reuse of commonly defined data and reuse of the resulting databases. A survey shows that of organizations most effective in implementing information resource management principles, 75 percent have reduced the costs of application development and 83 percent have reduced the costs of maintenance.²
• Reduces the costs of interface programs that transform data and move it from one place to another by sharing data directly from commonly defined databases.
• Reduces the costs of managing the consistency of unnecessary redundant data across different applications by eliminating multiple sources storing the same (or similar) data.
• Reduces the costs of correcting inconsistent data by eliminating the need for multiple sources of data and interface programs that can be the source of error .

The data model as an abstraction of business resources

A well-defined data model enables cross-functional business communication and represents the business knowledge required to achieve the enterprise's objectives. John Naisbitt states that in the new economy of the Information Age, successful companies must acknowledge that their primary resource is information and their assets are their employees.³ As a resource, data shares business resources. For example, the organization chart is a picture, or "model" of how it has organized its human resources. The chart of accounts is a "model" of an organization's revenue and expense categories. Organizations use these models to understand, communicate, and manage these resources. In the same way, a data model is a tool to understand, communicate and manage the organization's knowledge resource. Data is not simply a resource -- it is a master resource. Data is an abstraction that represents the business resources and events, and the data model represents the data. For example, customer John Smith places order 12345 for item "widget." The data model in figure 1 represents this business exchange.

Figure 1

A characteristic of all resources is that it has a common life cycle. As a resource itself, information has a resource life cycle similar to that of financial and human resources. Each resource has five major processes: planning, acquisition, maintenance, application, and disposition. Figure 2 represents the resource life cycle for human resource management.

Figure 2

Figure 3 represents the resource life cycle for information resource management.

Figure 3

Four of these processes are cost-adding. Only the process of applying the resource adds value. Resource management means maximizing resource value in the processes that apply that resource while reducing costs in the cost-adding processes. The data model supports a data-driven approach to application development. This update approach minimizes unnecessary create and processes (cost-adding) and maximizes the retrieve processes (value-adding) by sharing the data from commonly defined and accessed databases.

As a "picture" of the business resources, a data model:

• Increases communication among knowledge workers in different business areas by have commonly defined data.
• Increases business opportunity by relating data across functional areas and supporting cross-functional queries.
• Increases data quality by eliminating unnecessary redundant data and sources of error.
• Enables business process re-engineering by eliminating non-value-adding processes (redundant data entry and unnecessary interface processes).
• Allows effective management of the data resource by having a model that represents the enterprise data rather than going out to all the business areas to know and manage the data. (Consider managing inventory by physical inspection of the stock versus information representing the stock.)

Modeling operational data versus data warehouse data

Both operational databases and data warehouses require data models for sound design. The data warehouse model has distinct differences from its operational counterpart:

If an organization has not taken advantage of the benefits of data modeling for its operational data, the data warehouse represents a new opportunity to "get it right the first time." Failure to do so will spell disaster for the data warehouse effort. The resulting data model with consensus definition of key data can drive new development and re-engineering of legacy applications and databases.

How to make the case for data modeling to management

The benefits of data modeling are not automatic. Data modeling requires cross-functional resources up front to develop effective models and data-centric applications that maximize data value. How does one acquire and sustain management commitment?

• Start by identifying key objectives of the management whose support is required
• Describe in business terms how the results of data modeling enable accomplishing those objectives
• Illustrate the costs of not having a consensus data model and shared databases:
  • Conduct an inventory of redundant files for key data such as customer, item or other critical information
  • Quantify (estimate if there are no accurate records) the cost of developing and maintaining those redundant files
  • Quantify (estimate if there are no accurate records) the cost of developing and maintaining the unnecessary redundant applications that create, update, and interface the data
• Illustrate the value of integrated databases to exploit new business opportunity:
  • Identify business opportunities of interest to management that are not now achievable, or achievable only with difficulty and with excessive costs and/or unacceptable time delays
  • Document the problems in the current data structure and application design that cause the excessive costs, unacceptable delays or makes the opportunity impossible to obtain
  • Quantify (with the help of business personnel) the value of seizing those business opportunities
  • Document how a well-defined data model and data management enables seizing those opportunities

Conclusion

So, just what might the Sears Tower look like if it had been constructed without an architectural blueprint? It might look like a modern-day version of the Winchester House, a Victorian mansion in San Jose, California. Sarah Winchester, the widow of the maker of the famed Winchester rifle, perceived the spirits of those killed by her husband's rifles were haunting her. Seeking relief, she sought the aid of a spiritualistic medium. The medium convinced her she could obtain relief by moving west and continually building on a house.

The Winchester House is the result of constant construction, 24 hours a day, for 38 years without a master plan or blueprint. Mrs. Winchester only made sketches of individual rooms one at a time (like designing databases for applications one at a time). Because there was no rhyme or reason to the construction, the 160-room house -- which started with only 8 rooms -- has many anomalies. These include staircases that lead to nowhere, windows that open up to brick walls, and a chimney that reaches within 6 inches of the ceiling, and a door opening up to a two-story drop.

Will this be the fate of some organizations' database infrastructures? "Large organizations will have little choice but to become information-based," Peter Drucker asserts.⁴ The only way an organization can become information-based is to have a sound foundation of integrated data to support the business processes and decisions of the enterprise. The only way to have a sound foundation of integrated data is with a stable and flexible data model.

ABOUT THE AUTHOR

Larry P. English is president and principal of Information Impact International, Inc. He is an internationally recognized speaker, teacher, consultant, and author in information management.

He specializes in analyzing trends for effective implementation of information management. He is actively involved in all aspects of information management, including planning, organization, modeling and methodology implementation. He has provided consulting and educational services in at least 15 countries on four continents.

1 - Larry English, Conceptual Data Modeling Student Guide. Oakbrook Terrace, IL: Platinum Technology, 1994, pp.1-10.

2 - Larry English, "The Information-Age Organization: A Profile," fmi Journal, vol 3. no. 4. Winter 1993, p.3. The IRM survey was co-conducted by the Database Research Group Inc. and Larry P. English. Respondents included organizations from North America and Europe that have an implemented data management function.

3 - John Naisbitt, and P. Aburdene, Re-inventing the Corporation, New York, NY: Warner Books, 1985, p.5.

4 - Peter Drucker, The New Realities., New York, NY: Harper & Row, 1989, p.207.

Larry English is featured at DCI's Data Warehouse World.