Scrum meets Monte Carlo – How to make Scrum more predictable
Customers and stakeholders like ‘predictable’ Scrum teams, which reliably forecast their progress – particularly in contexts with strict deadlines or where cross-team dependencies must be managed. However, this is where many traditional estimation methods prove to be insufficient as they often include variation due to estimation flaws and unanticipated difficulties during the Sprint, making estimates less reliable. In our presentation we will present Monte Carlo Simulations as a statistical model to create reliable forecasts. We will show how they increase transparency within the Scrum Team and are a powerful communication vehicle towards stakeholders. Thus, they enhance empirical process control in Scrum and help us inspect and adapt. Monte Carlo is an established mathematical model, which is used in many fields such as science, engineering, supply chain management, meteorology, physics, finance, and traditional project management. Simply put, it uses historical data to calculate the likelihood of future events. In our talk we will see how this works in Scrum. We will show how to use historic Sprint data in order to create probability distributions for a forecast. This way, Monte Carlo Simulations make visible more than one possible outcome, allowing us to select different levels of confidence according to the specific needs of our project. Our goal is to show that Monte Carlo Simulations are not rocket science but a tool that anyone can use easily. We will look at straightforward real-life examples that show the superiority of the Monte Carlo-based predictions over human intuition and traditional estimating approaches. Also, we will discuss how they help you to have the right conversations in retrospectives, reviews, and other events and how they are crucial when it comes to understanding your process.