Forecast Accuracy has been a topic of discussion for decades. Just from my own time in Supply chain (20 plus years), it’s always been front of mind for everyone I know.
So here are my two questions. If we’ve been working on improving it for so long, and we’re still working on improving it, 1) Have we made any progress at all? And 2) Is there something we can do instead?
I grew up with a friend who was always late. Only occasionally would he be on time. And frankly, never early. If I continued to think I could make things better (my words), I would be very frustrated after 40 years. So, instead, I realized the situation and I work around the reality I face.
The funny thing about forecasting is that even if we make it 100% accurate, it will not fix the problems we are trying to fix. Bold statement, I know. So, let’s explore.
First, Forecast Accuracy only addresses one of the sources of variability. We still have Supply Variability, Operational Variability and Management Variability. Fixing the Forecast does nothing to address these other sources?
Second, the scope of the forecast matters. What I mean is, your forecast is a future prediction and it’s usually a monthly total. And often, it’s a family aggregate number which is a group of parts. Understanding what it means to turn a monthly, aggregate number into your daily, specific number is a key point. Predicting 2,000 cars sold is not the same as predicting 25 blue Fords, 25 green Hondas, 25 white Buicks, and 25 black Subarus per day.
Third, what is your expected outcome from the forecast? The two big ones are:
1) Finance – we use the forecast to set financial expectations. Whether we are reporting up to management or to wall street, the forecast gives us our income (and some extrapolated expense) predictions. For this, we don’t need to be too specific.
2) Supply Chain – we use the forecast to attempt to be prepared with component materials and some other various resources. But since our material orders must be placed to arrive on specific days, the forecast is necessary, but not sufficient. We need to further calculate the daily numbers. And for this, we need to be very specific.
Bottom line, forecasts are important. But if we are expecting to use a ballpark number (forecast) to set our very specific and precise supply orders… well, it’s just never going to perform as you hope.
There is a better way to use the forecast. But AI will not help until and unless you change the foundational approach to the real situation.
