Demand Forecasting in a Supply Chain - Review Notes

Chopra and Meindl's book, Supply Chain Management: Strategy, Planning, and Operation, is a comprehensive introduction on supply chain management.

Importance of Forecasts

Forecasts of future demand are essential for supply chain management decisions.

Demand forecasts are used in supply chain design, planning as well as in operations.

Demand forecasts are used in various subcomponents of supply chain.

Production: for aggregate planning, inventory control and scheduling,
Marketing: for new product introductions, promotions, and sales-force allocation
Finance: Plant and equipment investment decisions, operating budgeting
Personnel: Workforce planning and resulting hiring and layoff.

Characteristics of Forecasts

1. Forecasts may always go wrong. Therefore a rigorous presentation of forecast should include both the expected value and a measure of forecast error.

2. Long-term forecasts are usually less accurate in comparison to short-term forecasts.

3. Aggregate forecasts are usually more accurate in comparison to disaggregate forecasts. For example, forecast of the food consumed by a group of students in a college canteen can be forecasted more accurately than the food consumed by each and every student.

Forecasting Methods

Forecasting methods fall into four categories

1. Qualitative: The forecasts are based on the human judgement and opinion. Market research falls in this category.

2. Time Series: These methods use historical demand data of an item.

3. Causal: Causal forecasting uses data of multiple variable to forecast demand of an item.

4. Simulation: Simulation methods use what if questions and come out with forecasts. The underlying models for whatif analysis are time series or causal models. Even a hybrid model can be used for simulation.

When quantitative methods are used for forecast, the effort is to isolate systematic component and random component using the available data. The systematic component gives the expected value and the variation around the expected value happens in the future periods due to the random component.

Static and Adaptive Methods of Forecasting

In a static method, a single forecasting model is applied to the currently available data to derive forecasts for all the future periods for which forecasts are to be generated. In adaptive methods, as new data arrives, the new data is incorporated into the forecasting model to derive forecasts for future periods from then on.

Basic Approach to Demand Forecasting

1. Understand the objective of forecasting: Determine the decisions which are taken based on the forecast.

2. Integrate planning and forecasting in the entire supply chain: Different units in the supply chain should not forecast separately. All the required forecasts have to be generated from uniform premises and tools.

3. Identify major factors that influence the demand: This identification helps in choosing the forecasting technique.

4. Understand and identify customer segments for which you want forecast of demand.

5. Determine the appropriate forecasting technique

6. Establish performance and error measures for forecast.

Time series methods

In static methods, estimates of level, trend, and seasonal factor are derived using the past data. These three factors give the forecast of the systematic component for future periods.

Adaptive methods:

Moving average is an adaptive method. Exponential smoothing is also an adaptive method. Holt model is trend-corrected exponential smoothing model. Winter's model is a trend- and seasonality corrected exponential smoothing model.

Measures of Forecast Errors

An estimate of the forecast error is to be given along with the forecast of an expected value. As actual values are realized, a forecast error can be calculated and managers perform error analysis to satisfy themselves that the current forecasting method is accurately predicting the systematic component of demand. Contingency plans have to be put in place to account for the predicted forecast error.


Some popular measures for forecast error are:

Mean square error
Mean absolute deviation
Mean absolute percentage error
Tracking signal

References


Sunil Chopra and Peter Meindl, Supply Chain Management: Strategy, Planning and Operations, Prentice Hall, 2001.

Chamber, YJ.C., K.M. Satinder, and D.D. Smith, "How to Choose the Right Forecasting Technique," Harvard Business Review, July-August 1971, pp. 45-74.
Georgoff, David M., and Robert G. Murdick, "Manager's Guide to Forecasting," Harvard Business Review, January-February 1986, pp. 2-9.

Article originally posted in

http://knol.google.com/k/narayana-rao/demand-forecasting-for-supply-chain/ 2utb2lsm2k7a/ 1356

Updated 20.1.2012

Kindly Bookmark and Share it: