Estimating software projects

My Twitter stream flooded with estimation related topics yesterday. We’re currently starting up a new project, so the timing was perfect and the topic highly relevant. Here are some recommended blog posts from yesterday on the topic:

Estimation is at the root of most software project failures
In this blog post Rob Bowley explains why we’ll never be very good at estimation, especially the long range high level estimates – mostly for reasons outside of our control. Still many people in the Agile community talks about estimation as if it’s a problem we can solve. The discussion after the post is also highly recommended:

You Mean I Can’t Even Estimate? The Planning Fallacy in Action
A post by Jim Benson, down the same alley as Rob Bowley. Explains how we might deal with the fact that there will be a variation in every activity, and that we tend to underestimate a given task even if we’ve done that task many times before:

Estimate the total cost of Agile projects
Kane Mar first describes why estimating the cost of software is, at best, an educated guess. Then he explains how to deal with this when estimating the total cost of Agile projects:

Story estimation on Agile teams
A comprehensive Q&A by Praful Todkar. Who should be involved in estimation? Should we time-box estimation? Is pre-analysis necessary before estimating a task? What should we include in our estimates?

On the previous project we started off estimating stories before taking them into the sprint. When we gradually moved into the maintenance phase, we discovered that we only selected stories by priority – not size. Also, the stories were quite small, so estimating them seemed like a major overhead.

Not giving any real value anymore, we decided to stop estimating stories. Later, when we applied Kanban, we removed the iterations as well. We’ve managed well without estimation for about eight months now, instead focusing on the flow and maximizing throughput.

So here we are, about to start a new project. I think the time is right for us to reflect on why we estimate stories, and why we (try to) estimate the total size of the project. After all, for this particular customer, Scrum is the “company standard” – they’re used to estimating stories before prioritizing and starting them. Therefore, I think this is where we will start, this is what the company knows. But who knows where we will end up?

Do we still need estimates, or do we eventually get into trouble anyway when using them? Do you know other relevant blog posts on this topic?

Kanban and Scrum after 500 tweets

I joined Twitter June 22, 2010, so, needless to say, I was not an early adaptor. Nevertheless, I just passed 500 tweets, and most of my tweets are somehow related to agile development, so did I come across something useful? The answer is yes, and I believe it’s an appropriate time to reflect – focusing on Kanban and Scrum this time.

During this period I’ve learned to enjoy Lean thinking, and Kanban in particular. The difference, you may ask? David Anderson puts it well:
“People ask me what is the diff between Lean & Kanban? Answer: Lean is a destination, Kanban is a means to get there.”

If you don’t know Kanban, here is a clean and to the point overview:

And here is a first hand experience on what you can expect if you introduce it:

It’s not always easy to map a process to the Kanban board, so you might need some tips when you are stuck:

After a while, you will know that keeping an inventory is not free of charge, and there might be an upside if you manage to keep it down:
”In 1987, while GM held 2 weeks of parts in inventory, Toyota held 2 hours.” (Jeremy Lightsmith)

By then you should know that you need to pull work items, not push them. This strategy, if you see the analogy, might even affect your leadership:
“Management is push, leadership is pull.” (Jim Benson)
…and what’s really the difference between a project leader and a project manager?

Kanban takes an evolutionary approach, so big changes won’t happen overnight, but you should still be prepared for resistance in the organization. Here are 5 arguments against Kanban – and how to respond to them:

Even though I find Kanban very appealing, I still like Scrum. Recently, Scrum even got a major update to the Scrum guide! I found it really interesting that they now had removed release planning, cause it might compensate for poor development practices:
This doesn’t mean you have to stop all release planning, it’s just not mandatory anymore, and you should use it with care. This is an example of one small step you can take to improve the quality.

And quality, that’s what you need to focus on, no matter what agile process you are using. If you shorten down the feedback cycles, that would help you on the way. To get fast feedback, there’s a number of things you might consider:
…shorten the iterations:
…keep the releases small and deploy to production often.
…but maybe not 50 times per day?

If you want to go fast, you need to move with a sustainable pace. You need to stay clean so you can keep going fast:

So, to be able to shorten the feedback cycles, you need to focus on getting working software at the end of each sprint. And if you don’t, you’re definitely not alone. Over 50% of Scrum teams can’t get shippable code at the end of each sprint. Research shows that this will slow you down between 2 and 24 times:

There are many topics I didn’t cover this time, so maybe there is room for another post?