Monte Carlo simulation is a mathematical technique that predicts possible outcomes of an uncertain event. We’ve previously written about the benefits it provides to R&D professionals and hosted webinars about the power of Monte Carlo simulation to help products meet intended specifications, predict process capability and determine optimal process settings. However, Monte Carlo can also be a powerful tool to estimate the duration, or even the cost, of a project.
Project timelines are often uncertain events, particularly when the projects are cross-functional. While project managers typically guesstimate what the timing of a project could be, they could use a simple simulation to take out the guess work.
Imagine a software development project that is generally built on 2-week sprints. To keep things simple, let’s assume that all the requirements have been specified and a group of developers are ready to work. Not surprisingly, the boss asks, “Can we have this new feature delivered by our conference in 80 days?”
When estimating a project timeline, the project manager considers a development time of 3-4 sprints (42-to-56 days), quality testing of 1-2 sprints (14-to-28 days), and 1 week for deployment (7 days). A more cautious approach involves adding the maximum timeframes: 56 days for development, 28 days for testing, and 7 days for deployment, resulting in 91 days. Alternatively, a manager may average the estimates, arriving at 77 days (49 days for development, 21 days for testing, and 7 days for deployment). However, with a delivery request of 80 days from the boss, the certainty of meeting the deadline becomes a critical consideration.
No one likes to tell the boss it can’t be done. Perhaps, even worse, is telling the boss you’re not sure of the outcome. This is where Monte Carlo simulation can help!
As you can see from the diagram below, I’ve entered this project manager’s dilemma in Minitab Engage to predict the duration of the project. I’ve entered the three phases of the project (in days) and an upper specification of 80 days, which is when my imaginary boss wants to present it at our user conference.
Once I run the simulation, I gain two insights: first, it shows me that my most probable outcome (highlighted by the mean, in the middle of the distribution) is 77 days. Second, I can see that there’s a 23.8% chance that I do not hit the deadline of 80 days.
Armed with some data analysis, I can make some educated decisions. Maybe I’m comfortable with only a ~76% chance of making my boss happy. Or more likely, I need to add additional resources to the project to ensure success. Being able to tell my boss my confidence level and why I need to reallocate or invest into meeting their 80-day timeframe will likely result in a much more productive conversation ahead of the project and much more understanding at the end of the project, should I miss the deadline.