On the surface, forecasts are binary: they are either right or wrong. Even a forecast that is "mostly accurate" can result in an outcome that is far from where you were expecting to land, which poses a significant risk in many industries, especially low margin ones.
To avoid this, forecasters have a tendency to value forecasts that are presented with extreme precision — often down to the decimal point. Often these forecasts are well intentioned and thoughtful, with conclusions drawn from countless data sources striving for the most complete picture.
Precision Isn't Accuracy
Precision gives everyone involved the feeling of safety and security: "I am prepared for every bump and cobblestone in the road ahead, because my roadmap is very precise!" Well, maybe not so much. In forecasting, the roadmap is for a road you've yet to go down. As we've discussed in past entries, looking to past events as a predictor of the future can only tell you so much. The future is still unwritten, so we look for ways to mitigate risk and prepare ourselves for what might happen.
A detailed, precise forecast is simply that: Precise. What it isn't (necessarily) is accurate. And in the development of precise forecasts, we may lose track of the pursuit of accuracy by introducing unnecessary noise into the mix.
Leveling Out - An Example
By creating countless mini goals to measure the success of a forecast, the FP&A function creates noise and variances that demand analysis & explanation. This can be the kiss of death for forecast accuracy.
"The more granular one makes the prediction, the more likely it is that there will be error."
We have been working with a global manufacturer to help measure and improve forecast accuracy. When measuring the accuracy of a sales representative's past forecasts by part and customer for a particular month, the accuracy rate is often below 50 percent. But when we look at forecasts vs. actuals over a 2-month period, the forecast accuracy rate approaches 90 percent.
The ebbs and flows of daily activity, such as the exact timing of orders, order processing, production, shipping, etc., often do not coincide neatly with accounting's month-end cutoff and include many elements beyond the control of the sales rep. Since we focus on the objective of providing good information to support manufacturing requirements aligned with lead time, the 2-month window is more than adequate. Of course, we need to work on financial targets but that can be achieved in other ways, such as through improving operational processes.
A Winning Season
Seeking to create precise forecasts should be a red light for overly-ambitious forecasters thinking that they can predict every single event. The point should never be to create a forecast for every detail of enterprise operations, but rather to predict the end result. FP&A professionals need not define a successful forecast as a precise one; forecasters should strive for forecasts that are approximately right rather than precisely wrong.
