Agile Business Change Blog Thoughts on Agile Strategic Business Change and Agile Delivery

Tweet

Mail

Until very recently, I had only one approach to estimating a backlog on Agile projects. I had a number of variations worked out to cover complexities and uncertainties, but planning poker was what I did, and I was sure this was the one true way to estimate. Except...

I couldn’t escape the nagging feeling that I was missing something, just by having settled on one technique to the exclusion of all others. In particular, several of my colleagues were quite persistent in their support of something called “T shirt sizing”.  

Like a good Agilist should, I tried many times to get my head round this approach, but despite all the explanations and beer, I just couldn’t get it. My argument was; “Given a situation where a medium story is twice the size of small, and large is three times that, then why don’t I just use the numbers, 2, 4, & 12?”  

But recently I realised that my need to understand this small:medium:large ratio early in a project was causing me to drive teams towards premature design. Worse, the teams were becoming reluctant to engage with estimating, as the job of reviewing each story in turn, comparing it to others, understanding each new ‘wrinkle’ as it emerged and finally agreeing a size, just overwhelmed their energy.

In search of an answer to these issues I have discovered affinity estimating. In this model, the team sorts all the stories to simply put like with like.  Physically, this could be done as putting cards into piles, placing stickies on a whiteboard, or sorting rows in Excel. Whatever the physical mechanism, all the team has to do in considering each story is ask ‘Is this story like this pile or that pile?’.

In this exercise I’ve found it very helpful to label these piles (as they emerge) as small, medium, XXL and so on. The end result - very rapidly achieved - is some groupings of similar size stories.  Now the team can judge the relative size of stories by inspecting and comparing the groups. Since this means judging the relative size of maybe 6 piles of stories, not tens or hundreds of individual stories, its very much faster than planning poker and energises teams. It also seems to be more accurate, and as a result I may never play planning poker again.

I've been guided in this by Chris Sterling's great step by step description, and I agree with David Campey that this is magic.