How to Deal with Inaccurate Forecasting?

From the cases of Nike, Xilinx and P&G, we can see that forecasting tools are not always accurate. There are several reasons which will lead to stock-outs. For instance, when the demand increases sharply due to sales promotion while the production did not aware of this; when both the retailer and wholesaler are misled by distorted information signal and maintain certain amount of security floating rate of stock in order to avoid Bullwhip effect; or when intermediate distributors mismatch order, the supply chain disaster would happen. 

Facing the disadvantages of current forecasting tools, what should we do to decrease the forecasting uncertainties? The two articles from Todd Taylor and Lee, Hau L give us some tips on it.

Since Bullwhip effect is one key cause of inaccurate forecast, methods for eliminating the Bullwhip effect become important to SCM. The article of "Taming The Bullwhip" demonstrated some strategies to tame the bullwhip effect, it pointed out that "understanding true causes of demand, gaining visibility, and investing in collaboration with partners can help to counter the distortion of bullwhip effects. [2]

Todd Taylor provided some more tactics to decrease uncertainties during forecast. there are some overlap opinions from Todd with Lee, Hau L, for instance, they both emphasis on understanding true customer demand and information sharing with partners. however, this article discuss solutions further and give some more different demonstrations.

Here are the main strategies provided by the article:

1. Focus on the customer:  Maintain an optimal network design centered on your customer and how they want to receive and consume your product.  Segment your supply chain based upon this understanding and a clear picture of your unique value proposition. [1]

2. Define the right push-pull boundary and strategy.  Optimize your inventory allocation process based upon an understanding of your demand certainty.  If you have stable demand for some products, a push strategy can be employed.  But where demand is uncertain, a pull strategy will need to dominate your policy.[1]

3. Share information to increase the visibility to demand shared along the value chain.  [1]

4. Manage your product portfolio.  This entails honestly and accurately evaluating the costs associated with the products in your portfolio.  reduce the complexity of the portfolio. Complexity management is much more than a one-time, revenue and activity-based costing exercise.  It is not only a thorough evaluation of the products in your portfolio, but also a joint agreement with product development, product management, marketing, ops and finance on the criteria and guidelines for new product introduction (NPI) and the refinement of the product management process and metrics to ensure the company adheres to these decisions.  [1]

Both articles have provided some demonstrations on optimizing demand forecast, there is no universal tools to benefit every supply chain system since every production process has its own unique character, but these concepts can help us to consider more comprehensively when design and use our forecasting tools. 

My question after reading are, how to segment supply chain according to customer demand?
if  20% products takes 80% of the total variable cost in the supply chain. what shall we do to improve it? and what is the ideal proportion of this cost?

References:

[1]Todd Taylor. Dealing with inaccurate forecast. http://www.supplychaindigital.com/outsourcing/dealing-with-inaccurate-forecasts
[2]Lee, Hau L.TAMING THE BULLWHIP. Journal of Supply Chain Management 46. 1 (Jan 2010): 7.