· software-development

Parkinson's Law and Iteration Zero

I’ve been thinking a bit about Parkinson’s Law recently and its' applicability in software development.

Parkinson’s law is defined as follows:

Parkinson’s Law is the adage first articulated by Cyril Northcote Parkinson as the first sentence of a humorous essay published in The Economist in 1955: “Work expands so as to fill the time available for its completion”

My colleagues quite frequently reference this law with respect to stories taking the amount of time that reflects the story point estimate assigned to them.

I haven’t noticed this so much but I think we’re more susceptible to the law when what we’re working on doesn’t have a clear goal or doesn’t have a fast feedback loop.

One of the times where I think we run into this problem is in the agile 'iteration zero'.

Iteration Zero is an iteration reserved at the start of a project for setting up project infrastructure, building a walking skeleton and other such tasks. It’s typically a week long.

While I think it’s useful to do some up front work like this what I’ve noticed having participated in several of these is that we probably don’t need a week for this type of work even though it will easily fill a week if it needs to.

Despite my belief that the work fills the time it’s interesting to notice that we still don’t get everything perfectly setup in iteration zero and probably wouldn’t even if it was 2 weeks long instead.

We need to actually start driving some user functionality end to end and getting it deployed across our environments so that we can get proper feedback on the work that we’ve done.

For example one of the tasks of an iteration zero might be to have all the developer work stations ready to go by the time we start.

It’s certainly possible to get the work stations to a stage where we think we’ve got everything setup correctly but it’s not until people are actually using them that we know for sure.

The idea of an iteration zero is still useful but we should try to keep it as short as possible and accept that we’re still going to be learning/spending time in the first few iterations on iteration zero type stuff and expect a 'slower velocity' accordingly.

  • LinkedIn
  • Tumblr
  • Reddit
  • Google+
  • Pinterest
  • Pocket