Requirements traceability

[Graham_Labdon]

Hi All

Is it necessary/desirable to map all functional requirements to use cases?

Graham

[Oliver F.]

As always- it depends on the product/project you are modelling, the stakeholders and their motivation, etc.

If you are modelling a small throw-away project for a customer with overseeable requirements, then you will propably not need it.

However for long product lifecycles with many stakeholders and maintenance processes behind it there will one day come the question: "In which feature did we introduce this ?" (mostly immediately followed by a "who was that jerk" if that "jerk" has left the company a while ago so he can get blamed without risk).

Or "which use cases do we have to touch to extend feature X".

EA gives you a lot of tools to handle this but there is no silver bullets and generic golden rule for it- you will have to tailor it for yourself.

Oliver

[Transition Technologies]

Hi,
I agree with Oliver it really depends on you whether to map every functional requirement to use case.

From my experience I try to do so. The better the model is described the more it can give you for the later phases (design, implementation...)

With Relationship Matrix that is very easy and not so much time consuming, especially with new enhanced options.

[«Midnight»]

Another good thing about the matrix is that it works well with a simple (i.e. not overpopulated) set of traces.

It is very easy to get into 'forcing' traces where they really don't add (net positive) value to the model. Just because there is a mention or hint of something does not mean there should be a trace (or other link). Falling into this trap is easy. Once you do, traces don't add much to the model beyond maintenance overhead. At best this will result in a brittle system, at worst the brittle system will be late and will not match the model (nor the specification, from which the model will likely diverge).

Just my 0.02 CAD on the matter...