Joel -
This is something we're currently wrestling with as well. We came to
the decision to *almost* completely re-write our requirements in the
requirements management system we use (HP Quality Center) after we've
progressed to an integration testing/stabilization iteration. We
arrived at this decision after considering the following:
1. We (and those outside our teams) frequently refer back to our
biz+functional requirements for: support, training or "how does that
work again?" scenarios.
2. We don't want to turn stories into requirements documents. Our
stories do contain detailed acceptance criteria, but we want to treat
our stories as we think the methodology meant them to be treated (as
delta-type documents for a given iteration). This means that a story in
iteration 2 may supersede a story in iteration 1. Therefore, if our
stories were exported into "requirements" someone doing research would
find a lot of contradiction - not good.
4. We love change and know that our process must enable it. Per the
above we thought that if we started to treat stories as requirements
documents we would lose all the good things we had hoped to gain for the
project team and slowly migrate back to a waterfall way of doing things.
5. We really have it out for our analysts ;). On our teams our analysts
play the product owner (proxy) and unfortunately will have to perform
all the duties of that role as well as define the "final" requirements
in our requirements management system too. We know that we've reduced
our velocity by deciding to do things this way, but until our system(s)
documentation plays less of a role in support, training, etc we're
saddled with this process.
We are starting to pilot this process on a couple of teams to see how it
goes. Let me know if you'd like to be kept in the loop. I'm sure we'll
learn some lessons along the way.
Thank you,
Julie