Forecasting drives supply in your network. It sets the tone of supply chain planning. In many different contexts, a one percent improvement in forecast accuracy saves sometimes millions of dollars in inventory and other short-term working capital requirements.
Quite often forecasting is made synonymous with statistics and consultants/organizations spend most of their time developing forecasting algorithms that can detect that slight trend or cyclicality or changing seasonality pattern. In most cases algorithms work on numbers that may not be even "forecastable". Organizations sometimes not even draw a consensus on what numbers to forecast upon.
There are unanswered questions like – when should an organization realize sales, how should returns be treated, should billing cycles be accounted for, should an organization forecast on shipment, invoice date/quantity or better still point-of-sales. Sometimes organizations are also constrained by what information they have and can get or have control over. Given all these variables, it is very difficult for a forecasting organization to process actionable projected sales numbers from even best of algorithms.
Imagine another situation where planners have to do forecasting distinctly on multiple geography-specific hierarchical structures or complex dynamic product hierarchical structures. One cardinal and painful truth that most organizations discover only after-the-fact is that a very complex forecasting business process breeds a very unmanageable demand planning solution with numerous customizations. The management consequently spends its attention and investment-worthy capital on maintaining this complex landscape. They spend time and energy in procuring and installing more powerful machines.
Forecasting solutions fail or do not deliver expected results if – either the solution is not embraced by the stakeholders or is way too complex to catch anyone's imagination.