Marcus wrote a blog post about predicting progress in ‘well-defined’ projects. He used a simple example where uncertainty was very limited. You can read the details in the blog post. However, even when uncertainty about the work was low, the predictions he was able to do (and he collects quite a few metrics in that story), was limited.
In this episode, we explore the limits to prediction, and how that affects how we should look at prediction in our own software projects.
Embracing uncertainty and what that means in practice for software projects
We start this episode by discussing what uncertainty really means for us product developers. We also talk about the uncertainty that exists even when the whole project is already predefined (the example that Marcus uses in his blog post).
We hear often the word “uncertainty”, but what does that mean in practice. We explore that, and what Marcus learned as he understood uncertainty: “This has led me to ditch the idea of being able to estimate how long a task or project will take before we start.”
Ways that help us make predictions
Marcus is a self-described metrics nerd. He has started looking into metrics many years ago and is constantly experimenting with ways to measure and predict progress in his projects. All of that work has helped him understand some practices that work in practice, and what does not work.
We discuss how predicting is a dynamic process, and not a “number” or a “point in time”.
He shares the story of a project, where there was progress, but the prediction (done using estimates) had an “infinite” completion time for most of the projects. Something that Vasco argues in his book on #NoEstimates: we don’t know when we are going to be done if we base our predictions on estimates.
Here’s the graph Marcus shared:
The experiments that Marcus ran over time, helped him find a faster method to assess and predict progress in his project.
In this segment, we refer to the book “A seat at the table” by Mark Schwartz.
Predicting on a ‘well-defined’ project
The blog post that prompted this episode is about a well-defined project. Or so Marcus thought. In this segment, we explore how he started to predict in that project, collecting some data and projecting that into the future. We also discuss some of the failures, and problems with probabilistic forecasting. Finally, we discuss the idea that making decisions is the goal of using data, not making predictions.
In this segment, we refer to the book Actionable Agile Metrics by Dan Vacanti and the book How to Measure Anything by Douglas Hubbard.
About Marcus Hammerberg
Marcus is the author of Salvation: The Bungsu Story (available on Amazon), an inspiring and actionable story about how simple tools can help transform the productivity and impact of an organization. The real-life stories in The Bungsu can help you transform the productivity of your team. Marcus is also a renowned author and consultant in the Kanban community, he authored the book Kanban in Action with Joakim Sundén.