I'm trying to drive the process of creating traceability matrices for future development projects at my company. The challenge is that currently, the requirements documents (MRD and Functional Specification) created at this company for past projects have never contained uniquely numbered requirements.
They need to be numbered to create a traceability matrix. Or at least it makes creating the matrix a whole lot easier.
Now the big question. Who should number them? One approach would be to have the QA Lead assigned to the project "tag" the requirements as part of his/her review of the document. It's definitely important that the QA Lead highlight each requirement as part of his/her review of the documents anyway. QA's numbering could then be delivered to the author of the document for inclusion in the final revision.
Another part of me says that it's the author's responsibility. The problem here is most of them don't have the "detailed eye"
of a QA Lead. This approach also assumes that every author has a solid understanding of what a requirement is. Many Product Managers don't.
I have seen it done both ways and, as you state, it depends on how good the author is at understanding that a sentence contains a requirement.
I encourage the author, whose responsibility I believe it is, to number them and then spot the ones they have missed in the review.
The best requriements spec for this that I have seen numbered them also according to priority. The priority was set as MUST (the product cannot go live unless requirement met), SHALL (the product would be seriously degraded unless this requirement met, SHOULD (this requirement is desirable) and MIGHT (it would be nice), then within these priorities they were numbered 1 to n. These priority levels were debated during the reviews and then agreed.
This enabled us to create a matrix for each priority level which made life considerably easier (especially as we had over 3000 requirements).
As a QA person reviewing the requirement, it is more difficult to find the time to do this exercise, as reviews are almost always on the critical path.