Level 2 Data Flow Diagrams

Granularity - what is the "right" level of detail for a DFD?

The purpose of a Level 1 DFD (physical or logical) is to try and describe the workings of the whole system at a glance. One way of doing this is to put all the processes you can identify on the Level 1 DFD, and this is fine when the system has relatively few processes, (various sources will differ in what is considered acceptable For A2 level I think you should have six to eight processes no more than nine (nine because of research on human short-term memory limitations says that's all a reader can cope with). Do not worry if there are only a few processes it doesn't mean a low scoring project.

Reducing Granularity

If you have more than about 9 identified processes, there are several things you can do.

• Combine process boxes until you reach an acceptable number. It is a good idea to combine process boxes that don't do much work anyway, as it is helpful to get a similar level of detail in each of your process boxes. So if you have two processes which just receive inputs and store them, or two processes which are linked by Process 1 receiving inputs, doing something to it and then passing it on to Process 2, it is a good idea to combine them.

• If the processes look reasonably detailed, and there are maybe ten or eleven of them, one way to reduce the complexity of the DFD is to combine Data Stores if their contents look pretty similar.

• If the processes look reasonably detailed, and there are maybe ten or eleven of them, one way to reduce the complexity of the DFD is to combine externals if they look pretty similar, (e.g. customers and exhibition holders on WordWise).

Remember, this "lost" detail may well be recaptured at Level 2. Basically, we're producing a multi-level model, the context diagram is at Level 0, then we get a high level view of the relationships between processes at Level 1 DFD, each of the Level 1 processes can then be decomposed into more detailed Level 2, 3 etc. DFD's.

Decomposition of Level 1 Processes at Level 2

After you have sorted out your Level 1 DFD to your satisfaction, making sure that it gives an overview of the system, but is not too complex to understand, you need to work out which processes need further description at Level 2.

If a Level 1 process box has a brief, specific name, (i.e. it is obvious and unambiguous what work the process is doing), and it has no more than about six information flows entering or leaving it, it is probably OK to leave it.

Processes should be decomposed at Level 2 if:

• There are more than about six data flows into or out of the process

• The process name is complex or very general, (e.g. "Record Sales & Deal with Customer Orders", "Liaise with Clients")

© LEV