Estimation in Kanban?

Before I start, I should note that there is a lot going on in the Agile community related to Lean and Kanban. In case you've been in a cave, you might want to check out this great slide deck by Henrik Kniberg comparing and contrasting Kanban vs. Scrum, or check out Karl Scotland's blog on the topic.

That being said, Anna Forss blogged today about how estimates do or don't fit into Kanban. An excerpt:

The main reason for estimating tasks is to be able to decide on an appropriate work load during the sprint. During the sprint planning meeting, the team estimate how big tasks are and when the work load (estimated task sizes) matches the available resources, the team commits to the sprint backlog.

So, how about kanban?

Jump to her full post to see how she answers this question.

Here's my comment on the topic, even if it is a little open-ended:

I was under the impression that estimates aren’t needed as much in Kanban because items are supposed to be closer to being “of similar size”? Looking back to the original Toyota manufacturing process application, this would have been true.

Otherwise, a large item might hang in your work queue for weeks (because it is too large) while small items fly through. If several large items “get stuck” then it might clog the whole kanban system.

So… do you need to estimate to insure that items are small and closely equivalent? (maybe)

This also make the concept of MMF (minimal marketable feature) an important goal when writing stories!