Bring Back The Old
The figure below shows the phases of Scrum as defined in 1995 (“Scrum Development Process“) and 2011 (The 2011 Scrum Guide). By truncating the waterfall-like endpoints, especially the PSA segment, the development framework became less prescriptive with regard to the front and back ends of a new product development effort. Taken literally, there are no front or back ends in Scrum.
The well known Scrum Sprint-Sprint-Sprint-Sprint… work sequence is terrific for maintenance projects where the software architecture and development/build/test infrastructure is already established and “in-place“. However, for brand new developments where costly-to-change upfront architecture and infrastructure decisions must be made before starting to Sprint toward “done” product increments, Scrum no longer provides guidance. Scrum is mum!
The figure below shows a generic DRE system where raw “samples” continuously stream in from the left and value-added info streams out from the right. In order to minimize the downstream disaster of “we built the system but we discovered that the freakin’ architecture fails the latency and/or thruput tests!“, a bunch of critical “non-functional” decisions and must be made and prototyped/tested before developing and integrating App tier components into a potentially releasable product increment.
I think that the PSA segment in the original definition of Scrum may have been intended to mitigate the downstream “we’re vucked” surprise. Maybe it’s just me, but I think it’s too bad that it was jettisoned from the framework.
The time’s gone, the money’s gone, and the damn thing don’t work!
Leave a Reply
Why am I here?
WTF?
Meh!
D'oh!
My BTC Address
Bulldozer00's Blog 
hmmm…having start and end sprints favors new development while not having them favors maintenance work. Perhaps having the option might provide a more flexible toolkit?
Hi Gene,
Sounds good to me. Like you said yesterday, the removal of the bookend phases doesn’t mean they don’t have to be performed – especially the PSA phase. Explicitly including them as first-class, but optional, citizens (like the 1995 definition did) at least clearly points out their existence.
Indeed. Taking a toolbox approach and encouraging configuration of the process to individual contexts (perhaps with some generic samples as guidance) seems most likely to succeed as far as I can see.