Using a statistical forecast as a starting point often proves to be a solid start to a collaborative Demand Planning process. First, why do companies forecast? There are several benefits of more accurate forecasts:
- Improve customer service levels.
- Increased sales.
- Reduced cost of carrying inventory.
- Improved cash flow projections.
- Production smoothing (level loading).
- Reduced employee costs.
- Increased return on investment.
In basic terms, all manufacturing/distribution companies want to do the same thing. They wish to have the right stuff in the right place, at the right time. So then is it better to use a statistical forecast or one made by a person? Well, a statistician would call this a choice between statistical and empirical estimate. The statistical forecast gets generated by software that calculates that forecast strictly by applying statitical techniques to the sales/booking history while an individual or a group of people uses judgment to formulate a collaborative forecast.
So then, how does the statistical forecast fit in a collaborative process? Some experts recommend using a statistical estimate as a starting point rather than asking someone such as sales person, to create a forecast from scratch. This statistical forecast, should not be the final answer; rather it is the starting point of that collaborative forecasting process. It’s a great input to start cross-functional demand planning.
Another question that people ask is; are there other cases where we should not use a statistical forecast? And the answer there is yes. Any product that its history is not indicative of the future cannot use statistics effectively. Items that can be considered fads are particularly hard to predict. Many companies that have forecast increasing sales based on past sales trends for items that were a fads. They get stuck with inventory that can possible bankrupt a company. Remember the Atkins diet craze from a few years ago, Atkins Nutritionals went bankrupt after being to force to throw away millions of dollars of inventory.
New products, for example, pose a challenge for statistical forecast in that there’s no history, in this case, we suggest using a similar or like product, to model the projections for that new product.
Short life cycle products can also be a challenge because there’s a little history, and little history does not yield strong statistical forecasts. In other cases, volatility of historical demand can result in unstable or unusual trends that do not have any statistical correlation. And finally significant changes in demand patterns caused by acquisitions or other either company or external events can drive unreliable forecasts.
In summary, a statistical forecast can be an excellent tool but should not be the only method used to create an accurate forecast.