Forecasting Methods

Introduction

All businesses are confronted with the general problem of having to make decisions under conditions of uncertainty. Management must understand the nature of demand and competition in order to develop realistic business plans, determine a strategic vision for the organization, and determine technology and infrastructure needs. To address these challenges, forecasting is used. According to Makridakis (1989), forecasting future events can be characterized as the search for answers to one or more of the following questions:

„X What new economic, technical, or sociological forces is the organization likely to face in both the near and long term?

„X When might these forces impact the firmÐŽ¦s objective environment?

„X Who is likely to be first to adapt to each competitive challenge?

„X How much change should the firm anticipate both in the short run and the long run?

In this paper, I will provide an overview of forecasting methods and compare and contrast these various methods. The paper will then focus on how Mattel, one of the nations largest toy manufacturers, uses demand forecasting under conditions of uncertainty ÐŽV most specifically those relating to the pattern and rate at which customers demand products.

What is Forecasting?

In Operations Management, demand forecasting is defined as ÐŽ§the business process that attempts to estimate sales and the use of products so that they can be purchased, stocked, or manufactured in appropriate quantities in advance to support the firmÐŽ¦s value adding activities.ЎЁ(Ross, 1995). Forecasting is a process that transforms historical time-series data and/or qualitative assessments into statements about future events. This process can produce either qualitative or subjective projections. Note that no forecasting process can consistently provide perfect forecasts. Any forecast that perfectly estimates subsequent events should raise cause for alarm, as this is probably indicative of improprieties such as ÐŽ§cooking the booksÐŽ¦ or reporting performance data that shows conformance with plans versus actual events (Makridakis, 1989).

Forecasting Methods

There are four basic types of forecasting methods: qualitative, time series analysis, causal relationships, and simulation.

Qualitative Techniques

Qualitative techniques are subjective or judgmental and based on estimates and opinions (Chase, 2005). These forecasts reflect peopleÐŽ¦s judgments or opinions and suggest likely conditions, such as peopleÐŽ¦s opinion about whether it will rain today. These forecasts are preferred when there is a desire to engage individuals within the organization with a key business process. A potential pitfall of this technique is that some individuals base their judgments of future events on historical data, which may not provide relevant demand patterns that are stable enough to warrant their use to forecast future events. Additionally, emerging demand patterns may be too unstable for a numeric approach. Consequently, intimate knowledge of the market should be the data source of choice.

There are numerous qualitative approaches to demand forecasting, following are some of the more common approaches:

„X Grass-Roots Forecasting seeks input from people at the level of the organization that gives them the best contact with the event under study (Chase, 2005). This technique may consist of conducting a marketing study of sales representatives for their readings on current market conditions. The potential fault with this tool is that it is subject to the short-term perspectives of the sources. The source of the data may be unduly influenced by recent events. For example, a sales person who has had a good day may provide an overly-optimistic forecast for the future that does not accurately represent market conditions on the whole.

„X Historical Analogy: Forecasting based on historical analogy explores the possibility that past events can provide insights into the prediction of future related events. This method ties what is currently being forecasted to a similar item (Chase, 2005). For example, utilizing the sales pattern of black and white television sets to forecast color television sales. Economists relay on this type of forecasting model to forecast business cycles and related developments. This method could prove inaccurate if the forces that drove past events are no longer present.

„X Market Research Forecasting: This forecasting method collects data in a variety of ways such as surveys, interviews and focus groups to evaluate the purchase patterns and attitudes of current and potential buyers of a good or service. Designers of goods and services use this method to understand their current customers and the buyers they would like to serve.

„X Dlephi Method: The Delphi method compiles forecasts through sequential, independent responses by a group of experts to a series of questionnaires. The forecaster compiles and analyses the respondentsÐŽ¦ input and develops a new questionnaire for the same group of experts. This sequence works towards consensus that reflects input from all of the experts while preventing any one individual from dominating the process (Chase, 2005).

Quantitative Techniques

Quantitative forecasting techniques transform input in the form of numerical data into forecasts using methods in one of three categories. Each category of quantitative forecasting methods assumes that past events provide an excellent basis for enhancing the understanding of likely future outcomes.

„X Time Series Analysis: Time series analysis is based on the premise that data relating to past demand or performance can be used to predict future demand. Examples of this method include:

a. Simple moving average, where a time period containing a number of data points if averaged by dividing the sum of the point values by the number of points.

b. Regression analysis, where the average relationship between a dependent variable, sales for example, and one or more dependent variables, price or advertising for example, is estimated by fitting a straight line to past data to relate the data value to time.

c. Trend projections, a forecasting technique that relies primarily on historical time series data to predict the future. This method involves fitting a mathematical trend line to the data points and then projecting it into the future.

„X Causal Studies: Causal studies look for causal