Business Intelligence: Counting on Business Metadata

In recent years, interest in the management and use of metadata within our business intelligence environment has grown dramatically. But, 2000 will always be remembered as the year when a lack of business metadata dominated the news. The lack of a clear definition of the term "legal vote" resulted in a situation where, over a month after the election, the world still didn't know who would be the next U.S. president. Why? To put it succinctly, a lack of managed business metadata in the legislative metadata repository.

Metadata, in an information architecture, is all the physical data that describes and makes useful, the data, processes and knowledge within an enterprise. Good business metadata helps enterprises simplify, or avoid, unnecessary problems. There are many examples of business metadata, including entity/table names, column/attribute definitions, data structures and data dependencies. Looking at the election, the problem is definitional.

Definitional business metadata should be developed and assessed based on five criteria: 1)Does it increase vocabulary?; 2) Does it increase specificity?; 3) Does it decrease vagueness?; 4)Does it allow computation or selection?; 5)Does it sway attitudes?

These five items do not apply to all definitional metadata, but it is usually clear which criteria are applicable and which are not, as the term is applied to the criteria.

Many years ago, while working on generating lists of customers that fit certain criteria for marketing campaigns, a team generated multiple lists of customers based on different selection criteria and models. The lists were loaded into a set of tables for further manipulation and to trigger the marketing campaigns. These lists were soon used for many purposes that required a list of parties, including customers, prospects and employees. However, there was no common term to communicate the meaning of these lists. There was no shared vocabulary. The answer was to call each list of parties a population. The definition, any list of parties and associated attributes, added a new word to the enterprise vocabulary and prevented misunderstandings. The term and the definition represent business metadata that increased vocabulary.

In business, manufacturing may believe that a customer is any person that receives their outputs (e.g., an assembly department is the customer of a molding department, shipping believes that the customer is the person that acknowledges receipt of shipped goods). By separately defining terms, such as internal customer and ship-to customer, the enterprise increases the specificity of each type of customer.

By increasing specificity of business metadata, we, also, avoid disagreements that are merely definitional in nature. By specifying internal customer, parties engaged in a discussion, reading a report or exploring a data mart can absorb and convey thoughts and knowledge effectively. Consider how many combinations of the term "revenue" are needed in a business environment to achieve an appropriate level of specificity.

Vagueness is the primary metadata failing in the recent election debacle. In a business environment, consider the term "sale." Is a sale made when it is agreed to verbally? When credit is approved? A promotion manager would say that a sale occurs when the business discounts products and services as part of a promotion. "Sale" is a vague term that can lead to misunderstanding, even though we all "know" the meaning. When developing or reviewing definitional business metadata, evaluate the definitions from a wide perspective.

Many business metadata definitions must provide information for computation or selection. Computation is the result of applying a formula to arrive at a result. For example, the definition of total invoiced amount may include the phrase "the sum of invoice line item amounts," providing a method of computation. However, which line items are included from a table of invoice line items? By adding the phrase "for a given invoice," selection criteria are also defined.

Traditionally, computation and selection criteria are more carefully defined than other aspects of definitional business metadata because it is necessary to write programs that perform these functions to meet the most rudimentary of business requirements. The potential pitfalls are that the computation and selection criteria are not effectively communicated or not uniformly applied across different information systems - or across independent data marts. Proliferation of independent data marts without a shared source of managed and controlled metadata is the business intelligence equivalent to marketers holding ballots up to the light to see if the light shines through, while manufacturing simply uses a machine count.

Business intelligence projects involve data from throughout the enterprise, thus magnifying the importance of business metadata. Definitions of objects and entities are at the core of metadata - so, make sure you make your business metadata count.