|The purpose and value of the data flow
diagram is primarily data discovery, not process
mapping - hence the name "data flow diagram"
(DFD). The pioneers of structured analysis and design
methods - Gane and Sarson, DeMarco - devised
their DFD approaches to formalize analysts' anecdotal and
idiosyncratic data safaris. When a search is launched for
raw data with which to populate a model, the DFD provides
the cage to bring it back alive. Data items thus
discovered are cataloged in a data dictionary, then
subjected to algorithms for normalization and aggregation
into entities of a data model or entity-relationship
In the picture below we see the funnel of methods channeling discovered data into more formal and abstract representations.
While technically sound, the classic DFD-normalized E-R methodology was opaque rather than transparent in that it deferred any overview of the emerging data structure until all the tedious research and preparation had been done. Nonetheless, the data flow diagram remains one of the best techniques available for developing a comprehensive inventory of data items in a problem space.
Unfortunately, the use of DFDs in information analysis has fallen into disfavor due to lack of education in their use. Most analysts and information users mistakenly look for process sequence and event control in DFDs. In fact, sequence and branch logic have no place or meaning in a DFD and so their disappointed viewers abandon them. As Gane89 says (page 53):
One solution would be to evolve a method of data discovery which augments the proven value of DFDs with appropriate notation for sequence and control. Many newer techniques of data discovery have been offered ; none so far seems to have gained the broad acceptance and usage which DFDs once enjoyed.
Another direction would be to offer a modern CASE tool for the preparation of DFDs, a tool which would coach and guide in its appropriate use. In fact several vendors of data modeling products now offer DFD tools as well and we will try to provide coverage for these soon.