The term refers to methods that only take the previous (known) values as input. These are both such well-known models as moving average, exponential smoothing and some more complex models that determine the future value of demand at a particular point in time based on previous demand values. The methods can be applied to both short-term forecasting with planning horizons of one week to three months and medium-term forecasting with planning horizons of three months to several years. They should then be able to take account for seasonal, cyclical and trend factors.
Causal methods use static regression models to establish a relationship between the dependent variable (demand) and independent variables such as price, advertising budgets, etc. If such a model can be built, it enables scenario analysis, which is a significant advantage. However, it is necessary to remember that the independent variable has to be predicted well and that the relationship can change over time.
Unlike the quantitative methods described above, qualitative methods do not use a mathematical apparatus. They can be used to forecast product demand for both newly launched products, for which there are no statistics or data relevant to them, and existing product range. These methods include various expert judgment methods, including the Delphi method.
Predicting demand for new products cannot be linked to statistics describing sales because they are not available. In this case, supporting data are used, such as data on existing peers, data from research agencies, etc.