Automated manufacturing processes are fascinating: Each machine must function close to perfection independently yet work seamlessly with the other machines. The manufacturing process works because the engineers understand the big picture of how everything fits together.

This need to know the intricacies of a system is not limited to the production process. The operational applications that run the production machinery and the applications that automate other organizational processes such as accounting, supply chain and logistics must also seamlessly work together and share information as needed.

An organization's complexities must be well understood, captured and communicated in the form of an enterprise data model (EDM). A data model uses symbols and text to help developers and analysts better understand a set of data elements and their business rules. An EDM is a subject-oriented and integrated data model describing all of the data produced and consumed across an entire organization.

"Subject-oriented" means that the concepts on a data model fit together as the CEO sees the company, as opposed to how individual functional or department heads see the company. There is one "Customer" term and definition, one "Order" term and definition, etc.

Integration goes hand in hand with subject orientation and implies a single view of the data along with a mapping back to the chaotic real world. For example, if Customer Last Name lives in 10 applications within an organization, the integrated EDM would show Customer Last Name only once and would capture the mapping back to these 10 applications.

EDMs help organizations integrate their data with the goals of improving enterprise analytics, strategic planning and knowledge sharing. Because resource and skill challenges can hinder the creation and maintenance of an EDM, many organizations choose to purchase starter industry data models instead of trying to reinvent the wheel.

An industry data model is a pre-built data model that captures how an organization in a particular industry works or should work. Teradata offers eight industry data models called industry logical data models, one of which is the Manufacturing Logical Data Model (MLDM).

The details are in the data
The Teradata MLDM provides the big picture for a manufacturer and is, in essence, a living and breathing view of the manufacturing business. It contains 83 broad subject areas, such as Inventory, Invoice and Item. The current version of the MLDM is extremely robust, containing 1,649 entities, 6,092 attributes and 2,165 relationships. These numbers—and model features—are continuously updated through new releases.

As Teradata Professional Services members work directly with clients in the manufacturing field, they gather feedback for model changes and enhancements. The Teradata Product Manager captures these new requirements so they can be considered for addition in the next MLDM release. Also, market/industry trends and innovations, such as the use of radio frequency identification (RFID), are evaluated to ensure that additions to a new release provide increased value. Each release of the MLDM, therefore, provides manufacturing clients with new features.

The MLDM has a number of important characteristics:

	Operational. How a manufacturing company works instead of how a manufacturing company typically does reporting is detailed in the MLDM. In other words, the vast majority of the structures in the MLDM capture the data elements and business rules that govern the day-to-day operations of the business.
	Logical. The MLDM is completely independent of technology, so its business concepts are not tainted by a particular type of software, hardware or network.
	Extensible. Most manufacturers use the MLDM as a foundation, adding and removing structures and enriching the provided definitions to make the data model more meaningful and distinct to the particular organization.
	Abstract. A fair amount of abstraction is contained in the MLDM. Abstract refers to combining like things together under generic terms such as Event and Party to facilitate integration and gracefully handle future requirements.
	Global. The structures and terms on the MLDM are designed for international use.
	Standard. Best practice naming standards, including the use of class words, are followed in the identification of the data elements. A class word is the last part of a data element name that represents the high-level category to which the data element belongs, such as "code" and "amount."
	Digestible. The MLDM is sectioned into subject areas. Subjects are neatly captured in separate views, and the use of color distinguishing each subject area makes it easier to digest the larger models.

Simplify complexities
To illustrate one use of the MLDM, let's examine the concept of Plan. Plan is a critical concept to any manufacturing company, as it is a key indicator to product demand and affects every part of the production process, from the purchase of raw materials to the transportation of finished goods. Assume one manufacturing company has three different definitions of Plan:

	Night shift plant manager: A Plan is a good guess as to how much inventory will sell within a time frame, such as a month or quarter. It helps determine, among other things, the quantity of raw materials needed to produce the finished products.
	Carrier liaison in the transportation office: A Plan is a set of point-to-point mappings that capture the route each carrier must travel to deliver the finished goods to the warehouses, distribution centers or customers.
	Human resources department manager: A Plan is a career path for each employee in the organization. In addition to identifying roles, this Plan includes training and other forms of professional development.

enlarge

A conceptual schema or conceptual data model (CDM) is a map of concepts and their relationships that describes the semantics of an organization and represents a series of assertions about its nature. The MLDM comes with a CDM for the manufacturing industry that contains about 50 key concepts and their relationships. It was built by including at least one major entity from each subject area and then generalizing the rules among these remaining entities. For example, more than 15 entities, including Invoice, Invoice Adjustment and Invoice Type, are represented by just the single Invoice entity on the Manufacturing CDM. The figure at right contains a subset of the Manufacturing CDM.

One of the entities on this Manufacturing CDM is Forecast Plan. The first sentence in the three-paragraph definition of Forecast Plan provided in the MLDM is: "This entity defines a collection of predictions and estimations by particular timeframe, of product demand/sales/purchases/
production/etc. at some designated level of aggregation." This definition will most likely satisfy the night shift plant manager's definition of Plan.

Other subject areas in the Manufacturing CDM will accommodate the other views. The Forecasts and Model Scores Subject Area carries the following sentence in a two-paragraph definition of Plan: "This entity identifies an intended course of action."

This entity can then be copied into the CDM, and the modeling technique of subtyping can be used to fit each definition into this concept. For example, the human resources department manager's definition of Plan was called Employee Development Plan, a new entity on the Manufacturing CDM with a detailed definition provided by this manager. The carrier liaison's definition of Plan was called Transportation Plan, with a detailed definition provided by the carrier liaison.

enlarge

This is an example of obtaining the big picture, the system's intricacies. Eventually, these definitions must also be captured at a data element level. To provide a taste of what is involved in obtaining this detailed information, one of the Plan data elements from the source systems was mapped into the corresponding MLDM data element. (See table.)

Note that this mapping is overly simplified, as usually complex transformation rules as well as other types of metadata need to be reconciled, such as format, granularity and nullability. Also note that the model needed to be expanded to include the concept of Employee Development Plan, which was not on the original MLDM. This illustrates the ease with which the model can be customized for a specific organization.

Common understanding
Many integration battles are quickly defused using the MLDM, because instead of win/lose definition debates among business areas, it becomes a mapping exercise where both parties agree on a single, external, unbiased view.

Companies can save substantial amounts of time and money with a detailed and well-proven MLDM. The generic data model serves as a foundation for companies' EDMs and can be easily extended as the businesses grow. With an MLDM, manufacturers are given a common understanding of business terms as well as a big picture view of how each department, though independent, fits seamlessly with all of the others. T

Steve Hoberman has worked as a business intelligence (BI) and data management practitioner and trainer since 1990. He is the inventor of the Data Model Scorecard and author of two books, including "Data Modeling Made Simple."

Photo illustration by Jeff Grunewald

Teradata Magazine-September 2008