Let me say something obvious. Estimating is hard. So how come that we at Lunar Logic barely remember times when it was an issue? The standard approach to estimation is based on expert guesses. This means that subconsciously we make ourselves vulnerable to cognitive biases which render our estimates too optimistic and unreliable. That neither is fun for development teams nor provides much value to clients. We can employ a different strategy. Instead of relying on vulnerable human memory we use the actual historical data and run statistical simulations to produce a forecast. Not only is the outcome significantly better than what expert guess provides but also it requires less work from development teams. I will share a story how our approach to estimation at Lunar Logic evolved from a common scenario to what we call statistical forecasting. We’ll share the theory behind, show how we do it and present some real data to back up the story. After all, once we understand why estimation is hard we can easily improve how we do it.