Entity Relationship Modeling - Complete Model?
(Page 15 of 21 )
But, is this based on a complete model of the reality? Where do we get our BALANCE from? It does not come from out of the blue. A full and true picture of the reality would be something like:
Quarterly balances are computed from of transactions. The entities for BALANCES and PERIODS are denormalizations; derived (or computed values) that has to come from transactions. It MAY be correct to use the former model (in a data warehouse), but it violates 3NF if you expand your view to look at the complete business model.
I call this The Analysis Trap 1. This may happen if one tries to limit the scope of the task at hand, and it would lead us into a situation where, whenever a transaction is inserted, updated or deleted, we MUST have a business rule that says ‘Update BALANCES whenever something happens in TRANSACTIONS. There may even be a worse cause for this analysis trap 1 to happen: Someone might say: ‘Performance will be lacking if we do not sum up here and there’. But those quotes are a certain sign that one is thinking physical implementation of the model; tables, SELECT statements, and so on. However, a computer’s performance ability has NOTHING to do with the analysis of the business!
(We will return to that in an upcoming Design Phase topic).
Remember, it is the business; how we run it (or would like to run it), that determines what information (entities) we need, and how the different types of information interact with each other (relationships). As you work your way through the analysis phase, the business model (and implicitly the E-R model) will become more and more accurate. Applying normalization techniques will ensure low (or no) level of redundancy (repeating the same information), and as a consequence also deliver a high level of referential integrity (correct relationship values).
Analysis Trap 2 - Incomplete Business Understanding
The Analysis Trap 2 is about incomplete business understanding or misinterpretation of the business processes. This is a common source for inadequate data models. There are (at least) two factors that may lead to this; The system analyst is lacking experience either from the specific business area, or in general, and/or the business fails to bring forward enough detailed information about the business needs.
I have come to the conclusion, after more than 15 years of ER modeling, and a total of 28 years in the software industry, that more often than not, the "professional" part in the business analysis task, the system analyst, must take a fair share of the burden here.
Being a system analyst is not the same as being a programmer. These are two distinct different professions, and different skills are necessary. It is not a disadvantage for a system analyst to have a programming background; you may quickly see areas where the model will influence the development phase in a negative way.
Next: Building Queries >>
More Development Cycles Articles
More By Alf A. Pedersen
/ 53 
