# A Simple Look at Forecasting Methodologies By:: [[Brian Heath]] 2023-09-03 When forecasting the future, at least three methods have very different outcomes. The first method is to say what happened yesterday will happen today. This is a very reliable approach for near-term predictions. In all likelihood, tomorrow will be like today, as things take a long time to change. Furthermore, the incremental differences between today and tomorrow are small enough that most people are satisfied with the small amounts of error. Consider if one is forecasting the temperature. If one says today will be the same as yesterday, one will likely never be that far off. This is true despite the changing seasons. The seasonal changes are just too subtle daily for one ever to be extremely wrong consistently. However, what if one wants to predict what the temperature will be six months from now? Saying the temperature six months from now will be the same as today would be drastically wrong. So, the second forecasting method is to build a model about how things work today and see what the model says will happen in the future. While this method can be just as effective as the first method in predicting tomorrow, it is much more costly to develop. So, people often revert to the first method in this case. But this second method is much better at predicting the temperature six months from now. Here, one builds a model of the oscillating temperatures, calibrates it based on past temperatures, and calculates a future temperature at any time. In all likelihood, this forecasting method is ideal for predicting outcomes 3 to 36 months out. Why not further out? Because systems change, and this method does not account for this. Instead, this second method only models how things are operating today. If the system is very stable today, this method could be reliable for quite some time. But even stable systems change over time. Eventually, the "current state" model will fail. One needs to employ the third forecasting methodology to provide forecasts beyond five years. These methods model systemic changes and are the newest forecasting methodologies. Here, these methods' accuracy depends entirely on how much the system is understood. Often, we know little about these systemic changes because no one was measuring them or the changes are being seen for the first time. If one wants to forecast the temperature in 20 years, one would employ the third method that models how climates change over time. It is worth noting that this third method is regularly accurate at predicting future outcomes but often not great at predicting exactly when that outcome will occur. This is due to the many compounding factors that impact the rate of change when looking at distant, future possibilities. Suffice it to say, when dealing with systemic changes, it is much easier to say what is likely to happen than when. As much as we like to think about time and change as smooth and continuous, life and the universe are chaotic and sudden. The next time one forecasts the future, consider what method is the most useful and least wrong for the situation. #### Related Items [[Forecasting]] [[Predictive Analytics]] [[Systems Thinking]] [[Models]] [[Analytics]] [[Intention]] [[Accuracy]]