Why traditional forecasts fail

Traditional methods of generating forecasts are based on the idea that you need expert knowledge and intuition of different products and services to model their future behavior. However, this approach has fundamental limitations, as follows:

• It's impossible to know everything about all products and services.
• Knowing how products perform today is not a good guide for predicting how they will perform tomorrow.
• The behaviors of many products are highly correlated and can be difficult to disentangle.
• Traditional models get overwhelmed by today's big data.
• The data itself keeps changing.

To address these challenges, we need new forecasting methods that can handle large amounts of heterogeneous data while producing forecasts that are more reliable, more accurate, easier to interpret and explain, and more useful for decision makers.

Let's explore these limitations in detail to lay the groundwork for why new forecasting...