Monte Carlo forecasting is simpler than most people believe. To prove this we teach the basics using paper tutorials and dice. We make these tutorials publicly available for you to use to teach others within your company (or at home with your kids). We welcome all feedback on what works and what doesn’t.
Basic Monte Carlo Forecasting Tutorial
Get the Basic Monte Carlo Forecasting Tutorial here.
“Discover what Monte Carlo forecasting is by performing it by hand. This exercise simulates completing a project many times and plots the outcomes. Perform 9 more trials. Each trial involves filling all rows in a column until the remaining work count reaches zero.”
This tutorial is the simplest burn-down style Monte Carlo forecast. It simulates 10 trials starting at 20 stories being completed at a rate determined by a roll of a six-sided dice. The burndowns are plotted on a chart to demonstrate that each trial takes a different route to completion.
Throughput Monte Carlo Forecasting Tutorial
Get the Throughput Monte Carlo Simulation Tutorial here.
This is a more advanced tutorial. It has extensive instructions on its opening page, and a completed example as the last page. We use this in our training courses, and often play this with all staff during our forecasting engagements. It get students to perform 3 of 10 simulation trials (7 are already performed) using historical throughput samples (given to you). It walks the students through the mathematics of computing percentiles on result trials to help people understand how to interpret 85th percentile for example.
There are many teaching moments during this tutorial. The most common is why is all this “complexity” necessary. We answer that in the document –
“Historical throughput data for teams measures delivery rate for a wide portion of the development system (the wider the better). Team throughput per week accounts for delays; for example waiting time, impediments, staff availability, interruptions and un-recorded work. The impact of these delays is more significant to a forecast than the hands-on time alone. This is a reason developer estimates are unreliable when forecasting projects, they don’t account for delays and system dynamics. In a stable system (e.g. the team isn’t blown-up), throughput will be a good predictor of future delivery rate even with large item size variability.”
We will add more of these over time as we learn ourselves how to teach the methods we use and why they give superior results than dividing average rate of work delivery by an average amount of work remaining.