Alexandria's Services

Sales Forecasting

Defined by the American Marketing Association as an estimate of sales, in dollars or physical units, for a specified future period under a proposed marketing plan or program and under an assumed set of economic conditions and other forces outside the unit for which the forecast is made.

Box-Jenkins Univariate Modeling

The Box - Jenkins Univariate Modeling procedure is presently the most accurate statistical routine available to forecast sales. This procedure is very good for forecast projections for up to 2 years. A Time Series is fitted with a mathematical model that identifies cyclical, seasonal, trend and random patterns of data. Box - Jenkins Univariate models require at least 3 years of historical data to be effective.

Accurate prediction of sales is vital in determining adequate inventory levels, warehouse space requirements, logistical requirements, sales promotion efforts, sales support activities and production planning.

Box-Jenkins Multivariate Modeling

The Box - Jenkins Multivariate modeling procedure is currently the most accurate statistical routine for projecting sales using leading indicators. This tool is excellent for long-term forecasting (2 - 5 years) because of the inclusion of leading indicators for prediction of turning points. Time Series data is fitted with a mathematical model using a series of regression equations incorporating multiple input variables. These leading indicators are used to add information to the model.

The Box - Jenkins Multivariate models are useful in strategic planning for determining changes in plant capacity for capital expenditures, geographical location planning, and product and promotional mix planning. The tool's power lies in the ability to incorporate economic and demographic factors that affect future forecasting decisions into the model.

Decomposition

The Decomposition Model is a Univariate modeling procedure that forecasts sales using the trend and seasonal components of the time series. This model is used when less than 3 years of historical data exist. The model is less reliable that the Box - Jenkins modeling procedure for time series data because of the constraint that it is limited to a single model. The Decomposition model is simpler and less time consuming to apply than the Box - Jenkins modeling procedure, however, it is impossible to determine the validity of the Decomposition conclusion.

Figure 1: