4.1 Good decisions begin with good forecasts
Consumer demand varies greatly from month to month and varieties go in and out of fashion quickly. There are techniques that can help us to predict what will be needed in the future such as product demand, transportation logics, facilities, capacities, materials, personnel hiring and personnel schedules etc.
4.2 Forcasting methods
It is very difficult to find a universal method that can forcast everything. Before choosing a general forecasting method, it is important to answer the following questions:
1, What is the purpose of the forcasts? For what decisions will they be used?
2, What lead time is necessary for the forecast to be of value?
3, What specific entity (variable) is to be forecast?
4, What do we know about the entity being forecast? what factors affect or are related to it?
5, What data or information about the entity are valuable?
The role of time
1. Short-term forecast usually look no more than 3 months ahead.
2. Intermediate-term forecast has a time frame of 3 months to 2 years.
3. Long-term forecast usually has a time frame of 3 to 5 years.
Quantitative versus qualitative methods
Quantitative forecasting methods use mathematical models to represent relationships among relevant variables based on historical data. Because of this method's preciseness, it sometimes referred to as objective forcasting methods.
In contract, qualitative methods rely on one or more individuals to generate forecasts without using mathematical models alone; for example, a sales manager may predict future sales for the division based on informal discussions with some customers. Qualitative forecasting incorporates the forecaster's experiences, intuition, values, and persoanl biases into the forecast. These are considered subjective forecasting methods because there is no way to determine exactly what information is being used by the forecaster and how. Such forcasts are specific to the forecaster and cannot be duplicated by others.
Although it may appear that quantitative one should be more consistent and accurate, in practice the results depend on the circumstances or relatively stable environment fulfilling short-term and intermediate-term forcasting. However, over longer time horizons or in unstable circumstances, such as when the possibility of a war, trade embargo, change of government, or major technological innovation exists, fundamental environment conditions and relationships may change. In these circumstances, qualitative one can often be superior to quantitative methods.
4.3, Qualitative forecasting methods
Quantitative methods require a substantial amount of reliable, consistent historical data or which affect variables same. However, in many circumstances these conditions are not satisfied, so qualitative methods should be used.In general, qualitative one is considered when one or more of the following conditions exist:
1, Little or no historical data. For example, forcasting the demand for a new product presents a problem.
2, The relevant environment is likely to be unstable during the forecast horizon. For example, the oil embargo of 1973, the integration of the European Union in 1992.
3, The forecast has a long time horizon, such as more than 3 to 5 years. Because technological innovation, competition, or government regulations may occur.
The primary drawback of using qualitative methods is that they rely primarily on the information gathering and processing capabilities of ther person making the forecast and cannot be independently duplicated. But this method can be improved over time by using the following methodological enhancements.
1. Standardize the process. For examples, what data should be used, how the data can be combined etc. Survey same customers, ask same questions.
2. Monitor forecasts. Most people are not very good forecasters unless they have received some training. One of the best ways to improve qualitative forecasting is to monitor the performance of the forecaster(s). Each time a forecast is made, it should be compared later with the actual outcome to determine its accuracy helping the forecaster to correct his bias.
3. Remove Incentives for accuracy. If sales manager's performance is based on his future sales quota, he may under forecast the sales in the future to assure he can sell more than the forecast to get the bonus.
4. Use group or panel such as group averaging, group consensus or Delphi Method.
4.4 Quantitative forecasting method
Analyze the relationship between independent (or predictor) variables and dependent variables.
1. Steps in modeling
1.1 Graph the relevant data
1.2 Select a general form of the function
1.3 Estimate the parameters of the function
1.4 Evaluate the model quality
1.5 Select and implement the best model.
2. Time series and causal models
Time series forecast the variables affected by time such as the products need in next year. Causal or associative models correlate dependent and independent variables such as the friction and force.
Time series usually include permanent component (P), trend component (T), seasonal component (S), Cyclic component (C) and random components (Sigmai)
4.5 constant process and the cumulative average
If there are different constant periods, moving average is often used to smoothe the fluctuation, weighted moving average, simple exponentional smmothing.
4.6 Quazi constant processes
4.7 Comparing alternative models
4.8 Linear trend processes
4.9 Seasonal processes
4.10 Causal models
4.11 Advanced models