Production Scheduling Design

What Production Planning Does with Forecast Error

Executive Summary

  • Production planning must absorb the forecast error created by the forecasting system. This has important implications for stock and safety stock management.
  • This brings up the topic of both make to stock versus assemble to order, as well as estimating the costs of forecast error.

An Introduction to Forecast Error Usage

In a make to stock environment, when a company has a low forecast error, it enables production to make to the forecast. Most companies have a problem mastering statistical forecasting systems. And because they make their demand history as un-forecastable as possible by doing things like not performing historical substitution Not accounting for promotional sales – the majority of make to stock companies put their plants in a position of having to decide whether to essentially re-forecast.

See our references for this article and related articles at this link.

How Supply Planning and Production Planning and Scheduling Absorbs Forecast Error

Of course, safety stock is supposed to be the place where the forecast error is managed. This is taken into account through dynamic safety stock, which accounts for both forecast and lead time variability. However, the higher the forecast error, the less likely the company will be willing to carry the safety stock as calculated by the dynamic safety stock formula. A poor quality forecast increases the safety stock — but does not decrease the necessity to carry the calculated safety stock. This reality is covered in this article.

The dynamic safety stock calculator is available in this article.

What This Means For Companies in Safety Stock Management

What this means is that for many companies, they do not account for variability in safety stock entirely, or even mostly — and when this happens, it is left to the plant. In this situation, the plant receives mixed messages. They are told to build to the forecast. However, they are also measured on machine utilization and other efficiency factors — which means that a plant that makes a poor quality forecast will not be able to meet the demand as they will be making the wrong things. As many companies can often switch between producing different finished goods from the available input items, plants often do have leeway in what they can produce.

High Forecast Error at the Finished Goods Level?

Businesses that have a high forecast error at the finished good, but relatively few input materials that go into a relatively large number of finished goods — should test the forecastability of the input materials against the forecastability of the finished products.

If the forecastability of the input materials is higher, the company should consider switching to an assemble to order environment. This is explained in the following article.

Make to Stock or Assemble to Order?

Many companies that think they are made to the stock environment — are, in fact, following an assemble to order model for at least some of their products.

This is because production is actively changing the finished good produced based upon shorter-term demand signals, such as the open orders.

For whatever reason, the terms make to stock (which should probably have been called make to forecast from the beginning) and make to order are commonly understood in the industry, but assemble to order is far less well understood.

Therefore many companies that are actively engaging in assemble to order planning do not realize they are doing this. (1)

Production and Error

Do companies want their plants to make a bad forecast? If companies with a poor quality forecast required this of their plants, they would end up with large quantities of finished goods inventories combined with a large percentage of lost sales.


In companies with a poor quality forecast accuracy, the supply chain forecast is not merely performed in one place — that is in demand planning. In the companies that I have worked in that have had poor prediction accuracy, the same tendencies appear. The supply planning/inventory group makes adjustments to the supply planning parameters, which are done because they think the forecast is not useful. They need a way to account for it in supply planning. In these environments, procurement does not check the purchase requisitions for quality and then order what the software recommends. Instead, they check orders and consumption and adjust the orders based upon this as well as their intuition. And of course, production does not produce the production plan. But they change production based upon its “forecast.”

Having seen these environments enough times now, it is evident the inefficiency imposed upon the company by high forecast error. Furthermore, it sets up businesses to fail in supply chain planning implementations because, for decades, the company has been training its employees to react rather than to take the long view and plan. Companies will often take these same re-actively trained employees, provide some software training, and then post go-live find that they have user acceptance issues.

Estimating the Costs of Forecast Error

In recent projects, I have estimated the total costs of the inefficiency imposed by forecast inaccuracy. And while it is not a natural value to quantify, and it will never include all of the costs, it is something companies should attempt to quantify. One of the problematic areas to quantify is the fees absorbed in production.

We know that the costs incurred by inefficiencies in production are higher than those imposed by safety stock. Since most companies with low forecast accuracy do not carry the calculated safety stock, they are incurring those costs regularly. Unfortunately, “planning” as a concept is still not strong in companies, and the focus is very much more on “getting orders out the door,” then on planning how to do this efficiently. Many companies have purchased planning software, but still, don’t have the planning culture required to take advantage of this software.


There are many hidden costs imposed upon production by poor forecast quality. And poor forecast quality is the norm within companies, not the exception.

That is a lot of costs that companies are not estimating. They are implicit costs, but my analysis at several companies now shows they are quite high. And the costs I have calculated have been a gross underestimation of the total costs. So far, it has always been more than enough money, even at a partial estimation, to invest in improving forecast accuracy.