• »

# Demand Forecasting Methods in Supply Chain Management: Techniques for Predicting Future Business Inventory Needs

Updated on September 18, 2012

In supply chain management, reducing inventory is a critical component of cost control. While there are many methods for reducing inventory, one of the primary ones is demand forecasting – trying to obtain an accurate picture of what demand will look like in the future.

Approaches to demand forecasting vary, but generally fall into one of two categories:

• Quantitative Forecasting
• Qualitative Forecasting

Quantitative forecasting utilizes statistical models to predict future values. These models may take into account current and historical trends.

Qualitative forecasting is less mathematical and more intuition-based. In many instances, especially those that feature rollouts of “unique” products (like the original iPhone), statistical models can provide inadequate results because they don’t have enough past data to make an accurate prediction of future demand. This is where human expertise enters the process: smart demand forecasters can often take into account factors that statistical models cannot turn into mathematical equations.

## The Time Series Method

The time series method is a quantitative forecasting technique that bases calculations of future demand on historical patterns. It assumes that the past is a good indicator of the present and future – the assumption of continuity.

The time series method statistically analyzes three main components: basic value, trend, and seasonality. Basic value represents the “baseline” – the historical rate of sales. The trend analyzes which way (if any) demand is trending – is demand for this item increasing or decreasing? Seasonality takes seasonal variations into account: obviously, some goods (like ice cream) are in higher demand at certain times of the year (in this case, summer).

Under a time series analysis, a projection of future demand is constructed from historical data of these three components. While the results can be skewed if other factors need to be taken into account (say, price changes), it is usually reasonably accurate.

## Forecast Accuracy and Key Performance Indicators (KPIs)

In inventory control operations, demand forecasting is usually performed for each stock item SKU – stock keeping unit. Different metrics, called KPIs (key performance indicators), can be used to ensure that customer service remains at or above acceptable levels for each SKU. Common KPIs include case fill and order fill.

To fulfill these objectives, it’s necessary to keep a certain level of safety stock (for a discussion of safety stock vs cycle stock, see this article). However, holding excess safety stock can lead to higher inventory costs, and furthermore, increases risk that some inventory will have to be written off during the later stages of the product lifecycle.

If supply chain managers are interested in reducing safety stock, there are several methods they can use, foremost of which is demand planning. Increasing forecasting accuracy, then, is a primary goal in the supply chain. Forecast accuracy can be calculated using a formula called MAPE – Mean Absolute Percent Error. The equation is as follows:

Error % = |Actual – Forecast|/Actual
(note that |x| means the absolute value of x)

The reason that absolute value bars are used in mean absolute percent error is because for statistical reasons, magnitude of error is often more important than the sign.

Forecast accuracy is a metric that is inversely related to MAPE (as MAPE increases, FA decreases). The equation for forecast accuracy is:

FA = 100 – MAPE
(expressed as a percentage)

## Order Cycle Management and the EOQ (Economic Order Quantity): Applications of Demand Forecasting

Since the point of demand forecasting is to control inventory, applications of demand forecasting are obviously found in daily inventory management. Businesses fall into two classes when it comes to their order policies for ordering cycle inventory: there are those that engage in continuous review and those that engage in period review.

Continuous review systems review inventory on a daily basis and reorder a fixed quantity of stock whenever it drops below a certain threshold known as the reorder point. The classic example is a trucker: whenever his fuel drops below a certain point, he fills up his semi’s fuel tank with a (relatively) fixed quantity of diesel.

Period review takes an alternative approach. At regular periods of time (which depend on the industry as well as the specific business), inventory is reviewed and stock is ordered in sufficient quantities to keep inventory at a predetermined level. Thus, the order quantity is variable, and depends on the amount of stock used since the last review date. (Period review is also known as min-max policy because inventory varies between a set minimum and maximum.) These review dates are known as review points.

In general, continuous review results in more frequent orders and lower inventory levels. Period review results in higher inventory levels and less frequent orders. While the specific method used varies across industries and organizations, ABC analysis is often useful in determining which method to use (in the absence of other necessitating factors). Fast moving A items will more likely utilize continuous review, while slower moving B and C items are better suited for period review.

Economic order quantity is the final point that needs mentioning in the context of demand forecasting. As we just established, there is always a tradeoff between order frequency and inventory levels. Each have separate benefits. Keeping inventory levels low and having frequent orders reduces the risk of obsolescence, or in the case of food items, waste due to spoilage. On the other hand, frequent orders will increase transportation costs – it’s cheaper to have one 100 pound bag of salt shipped than it is to have 100 1-pound bags shipped. So while inventory levels are higher with period review, the higher inventory holding costs are sometimes offset or outweighed by reduced transportation costs.

A commonly used model for calculating EOQ utilizes the following equation:

EOQ = square root of [(2CR)/(PF)]
with C = order cost (per order), R = annual demand (per unit), P = purchase cost (per unit), F = annual holding cost (as a fraction of unit cost), and PF = annual holding cost per unit

The EOQ can be obtained by plugging those variables into the formula. It is not always an accurate representation, however, because it makes certain assumptions:

• linear order costs
• constant/continuous demand
• enough warehouse capacity for storage of EOQ

## Popular

29

0

• ### Achieve the Goals You Set: Use the Law of Attraction

8

0 of 8192 characters used

• Author

skylergreene 5 years ago

That statement is entirely correct. Models are never the same as the real world, but in many instances they can be useful for approximation and planning purposes.

• Michael VO 5 years ago from Warrenton Missouri

Forecasting is kind of like trying to see the future and no one can see the future. As an engineer I am familiar with models. I model projects in my Critical Chain Project Management efforts. I heard some one say that all models are wrong, but some models are useful, what is your take on that statement?

working