July 22, 2010

If It's Painful, Do It More Often

I've worked with a number of teams who couldn't deliver. They could write code all day long for months on end, and they'd do just that. But then they'd spend a month or three shaking bugs out and still not be able to get the product into production. In many of these cases, the programmers just want to program. They don't want to or know how to do the messy task of deployment. Install and upgrade is not their core competency.

It's instinctive to cram more features into the release when it is just so hard to deliver. If you get few chances to deliver improvement, you want to cram as much in the release as possible. But that just makes it harder. More beautiful garbage to push over the goal line. This raises the inventory of unfinished work, the amount of work in process, which further delays deployment and hides the real problem.

I'm reminded of the Israelites when Moses instructed them to not save the manna that God provided from one day to the next. Do not keep an inventory of the stuff. But some of them didn’t listen and kept some of it until morning. But by then it was full of maggots and had a terrible smell. Software development with too much WIP is exactly like that.

Anyway, Lean software development reduces work in process in order to surface problems and reduce costs.

I like this illustration from page 7 in Mary Poppendieck's book, Implementing Lean Software Development. When you first look at it, your reaction might be "No! RAISE the water because there are rocks!" But the rocks are still there. Some day you are going to hit those rocks. Better to hit the rocks and deal with it in a dingy than in a yacht. Better to deal with the inability to deploy using a small release than a big one. It's better to lower the inventory, surface the issues and work to resolve them. Don't leave them there, lingering.

In eXtreme Programming we say that if is something is painful, do it more often. What we mean is that we should not avoiding dealing with problems. Solve them right away.

1 comment:

  1. I like the idea of being derived mainly by exposing issues rather than cramming features into the release. Example, addressing technical debt rather than adding features.