Monitoring and controlling forecasts

Comprehension check.

USING COMPUTERS TO FORECAST

Forecasting often involves manipulation of great amounts of data. Computers can greatly reduce the work and time involved in developing and evaluating forecast models and in making routine forecasts. A great deal of work may be involved in evaluating possible forecast models and selecting the one that seems to best fit a demand pattern. Many combinations of independent variables may be examined in the search. The ability to perform repetitive calculations rapidly also makes it feasible to try values of the independent variables one period prior to demand, then two periods prior to demand, and so forth. Thus one can investigate several amounts for ‘lag’ to find the one that best predicts demand.

Computers can be applied to use a forecast model after the model has been selected. Companies sometimes have thousands of items for which they routinely develop forecasts. Automation of much of the forecasting calculations improves the speed and reduces the cost of developing these forecasts. A forecasting program can access a company’s data base to process data in a demand history file and routinely make forecasts for products or product families.

1. Why are computers involved in developing forecasts?

2. What are the advantages of applying computers to forecast?

Sales patterns are seldom static, and sales forecasts should strive to move with them or even anticipate their movements. New accounts are gained and old ones lost. The level of activity of existing accounts may rise or fall. Sales forecasts should be reviewed and revised periodically. Sometimes the forecasting method may be changed.

Many companies revise or update their forecasts for every month or every quarter. As new data become available they are incorporated into the forecast because usually these recent data are most relevant, at least in short-range planning. New data can also be compared with the forecast values to evaluate the forecasting method’s performance. One way of determining whether a forecasting model is performing adequately is by visually comparing new data and forecast values. Another method is to use a tracking signal.

The tracking signal is recalculated each time the actual demand data for a sales period become available and the forecast is updated. The tracking signal should remain fairly small (near zero) if the forecasting method is performing adequately. Deviations will occur but should balance each other – some deviations will be positive and others negative.

A method of checking forecasts is particularly useful if forecasting is done by a computer, as it often is when forecasts must be maintained for a great variety of products.

Дата добавления: 2014-11-29