What is a Business Rule?

[gleman]

The term "business rule" gets bandied about quite a bit in data quality circles. In most of the data cleansing world, a business rule is a pattern that a particular data item must match. In the data profiling world, a business rule entails a pattern, a set of relationships, or a range of acceptable values.

My thought is that all of these are correct, but the term "business rule" really means something quite broader as it applies to data quality. A business rule is a statement about what constitutes the set of valid values for a set of data elements.

Notice that it's not just a statement about a particular field. A business rule can have a single "parameter" that defines the behavior of a field, but it can also extend to many fields across a record, a join of several records, or even an aggregate of tables or an entire database. For example, a business rule could say that the average of a column formed by the join of several tables will never exceed 150.

Business rules also include algorithmic statements. I think most DQ vendors would claim that they support business rules because they can define a pattern that a data element must match, inclusion in a set of lookup tables, or a simple expression. But real world data is never that simple. To really support business rules, you have to be able to handle the IF-THEN-ELSE and FOR-NEXT cases. For example, a business rule could be as complex as "If the value of Column A = 1, then the value of Column B must be in the range of A-D, otherwise Column B must be blank and Column C must be computed by taking half of the average of Column D from another table using Column E on the first table as the key to the lookup table." That's a fairly tough pattern to write even in the most sophisticated pattern matching environments, but it's a very easy snippet of code to write in almost any programming language.

Your thoughts?