Available demand forecasting models fall into two broad categories:
- qualitative – Qualitative methods include expert opinion polls, the Delphi technique, and consumer surveys.
- quantitative – Quantitative methods include Time Series techniques and causal models.
To use an expert opinion poll, first identify individuals with relevant experience – perhaps specifically with relevant products. Then solicit their opinions about what the demand for the product will be. Because these experts deal in the same or similar products, they can predict the likely demand for a given product in future periods under different conditions based on their experience. If there’s a large number of experts and their experience-based reactions vary, an average can be taken. This can produce unique forecasts. However, the method can be highly subjective.
The Delphi technique is a variant of the expert opinion poll method. It seeks to arrive at a consensus in an uncertain area by questioning a panel of experts repeatedly until their responses appear to converge along a single line. The participants are given the opinions of everyone on the panel of experts. This may cause them to revise their earlier opinions and narrow the gap between the divergent views. The Delphi technique is intended to generate a reasoned opinion in place of unstructured opinions. But this is still a poor substitute for studying the market behavior of economic variables.
Time Series methods use historical data, assuming that the causes of demand variation in the future will be the same as in the past. Time Series methods are the most common forms of forecasting. These models are called naïve because they’re based on the assumption that historical patterns of demand are good indicators of future demand. The advantage of this method is that it doesn’t require formal knowledge of economic theory or knowledge of the market. All you need is time series data. Its limitation is its assumption that the future will repeat the past. Also, while the Time Series method is appropriate for long-range forecasts, it’s inappropriate for short-range forecasts.
Causal model forecasts use Time Series data, but with a mathematical technique known as regression analysis. By using that statistical technique, causal forecasts can mathematically relate dependent variables like demand to independent variables like time, price, or promotion. Unlike naïve forecasts, causal models take account of real-world economic factors. But there are some disadvantages to causal modeling as a forecast technique. First, there are the time and cost implications of collecting and analyzing data. And second, relationships between variables can change over time.
One can classify the various models available for forecasting into three categories:
- Extrapolative models: They make use of past data and essentially prepare future estimates by some methods of extrapolating the past data. For example, the demand for soft drinks in a city or a locality could be estimated as 110 percent of the average sales during the last three months. Similarly, the sale of new garments during the festive season could be estimated to be a percentage of the festive season sales during the previous year.
- Casual models: It analyses data from the point view of cause-effect relationship. For instance, to the process of estimating the demand for the new houses, the model will identify the factors that could influence the demand for the new houses and establish the relationship between these factors. The factors, for example, may include real estate prices, housing finance options, disposable income of families, and cost of construction and befits derived from tax laws. Once tea relationship between these variables and the demand is established, it is possible to use it for estimating the demand for new houses.
- Subjective judgments: Another set of models consist of subjective judgment using qualitative data. In some cases, it could be based on quantitative and qualitative data. In several of these methods special mechanisms incorporated to draw substantially from the expertise of group of senior managers using some collective decision making framework.
Selection of a forecasting technique: The selection of a forecasting technique depends on the following three factors:
- The characteristics of the decision making situation, which include: (i) The time horizon (ii) Level of detail (iii) Number of items (iv) Control versus planning
- The characteristics of the forecasting methods: (i) the time horizon (number of periods for which forecasting required) (ii) The pattern of data (horizontal, seasonal trend etc.) (iii) Type of model( casual, time series or sta6tistical) (iv) Cost (v) Accuracy (vi) Ease of application
- Present situation which includes: (i) The item that is being forecast (ii) Amount of historical data available (iii) Time allowed for preparing forecast.
Although there are the below mentioned forecasting models we shall be concentrating on Weighted moving averages model.
- Weighted moving averages
- Casual forecasting model
- Linear regression analysis
- Multiple regression analysis