CHAPTER FOUR

(DEMAND MANAGEMENT)

OVERVIEW

Demand management involves identifying all potential sources of demand as well as

influencing the level and timing of demand. Attempts at gauging initial and the effects of
demand management are called forecasts. Managers of operating processes must have
forecasts on which to base all their activities… if the forecasts are wrong, it is unlikely that the
operation will be effective, let alone efficient.

This chapter deals with demand management and particularly with forecasting. We will
deal with both qualitative and quantitative forecasting models. We will also look at forecast
error, for it is error on which the selection of the most appropriate forecasting technique is
based.

Managing Demand over the Planning Horizon

Demand management involves recognizing the sources of demand for a firm’s goods and
services, forecasting demand, and determining how the firm will satisfy that demand. Together,
marketing and operations provides the goods and services needed.

Demand sources include ultimate customers and downstream members of the value
chain. Demand will be for products as well as for service parts, spares, prototypes and mockups
of products, transfer parts among plants, and additional, materials and products to cover
breakages and unplanned shortages.

Typically, forecasts are prepared in the following sequence:

Economic Forecasts: economic forecasts are predictions of what the general business
conditions will be several months or years in the future.

Technological Forecasts: technological forecasts predict the probability and significance of

various possible future developments. They indicate the expected direction of technology
change and the expanded rate of change.

Demand Forecasts: demand forecasts predict the quantity and timing of demand for a firm’s
goods and services.

Resource Forecasts: resource forecasts predict the timing and quantity of demand for a
firm’s facilities equipment, and workforce as well as for purchased parts and materials.

Long-range forecasts are used for planning major expenditures on equipment and
people. Intermediate-range forecasts are more detailed than long-range forecasts and are used
for aggregate production plans and staffing plans. Finally, short-range forecasts are very
detailed- by individual product and by week or day-and drive the master production schedule.

The shorter the forecast horizon, the more accurate the forecast; the more aggregate
the product group, the more accurate the forecast and the larger the population, the more
accurate the forecast. As there is a need to “freeze” the planning horizon if work is to be easily
managed through the plant, reducing the frozen portion of the plan will increase forecast
accuracy.

Managing Demand in a Fast-Response Organization

Customer criteria demand management involves identifying what customers want and
deciding how that demand is to be met. To accurately identify true demand, we must
understand customer needs and expectations.

Customer must be well known; this allows predictions to be based on good knowledge.
The best way to get good knowledge is to ask actual customers, not samples of the general
population. Changing needs are used to drive product developments; these developments can
be tested on loyal customers.

This feedback is quicker than is possible with extensive market research. The window
will therefore have shifted only a little between the time of asking and the time of new product
response. More accurate and quicker data reduce product risk.

In FRO’s the forecasting focus is on the short and medium term, not the long term. It is
assumed that the operating systems are responsive enough to be able to use short-range
forecasts. There will be little need to build up a safety stock of finished goods inventory “just in
case”. Customers are more likely to pay a premium if their needs and expectations can be
quickly met. And repeat customers are more profitable than is prospecting for new customers.

Qualitative Approaches to Forecasting Demand

These approaches are all subjective, relying on intuition. When historical data cannot be
used or when quantitative data modelling would be too expensive, qualitative approaches make
sense. The techniques include:

Executive Committee Consensus: This involves getting a group together and having the
members discuss their opinions regarding future values of items being forecasted.

Delphi Method: Anonymous opinions from a group are circulated through the group several
times to encourage the group to reach a consensus.
Sales Force Composite: opinions of regional experts are aggregated to provide a firm-wide
forecast of immediate sales.

Customer Surveys: this method forecasts on the basis of the stated intention of current and
potential customers.

Quantitative Forecasting Models

Time Series Modeling: This involves plotting demand data on a scale and studying
the plots (past) to predict the future. Time series are frequently analyzed into four parts.

1. Trend – gradual upward or downward movement of the data over time.

2. Seasonal – seasonality is variations that repeats itself at fixed intervals of a year or
less.
3. Cyclical – cyclical variations has a duration of at least one year, and its duration varies
from cycle to cycle.
 4. Random Variation – variations that cannot be explained by the first three
components.

Smoothing Models – are often useful as they provide reasonably good forecasts quickly and
inexpensively. Ex. Simple moving average involves assigning equal weight to all past
observations.

Exponential Smoothing – assigns the highest weight to the most recent observation and
successively lower weights to older demand observations.

Exponential smoothing with a trend – there is a trend estimate, and it use trend –adjusted
forecast.

Focus forecasting – uses a series of different rules to generate independent forecasts for the
last three-month period, using historical data. The best forecast is the one with the smallest
forecast error for the past three-month period. The rule is used to forecast the next three
months’ demand for that particular product. The rules are:

- What was sold in the past three months is what will probably be sold in the next three
months.
- What was sold in the same three-month period last year will probably be sold in the
same three-month period this year.
- Ten percent more will probably be sold in the next three months than was sold in the last
three months.
- Fifty percent will probably be sold in the next three months than was sold in last three
months
- The percentage change experienced over the past three months over the same three
months this year will probably be the same for the next three months.
This technique cannot be used for new products or for items that are rarely purchased-
especially when the occasional purchase may be very large.

Causal Modeling: These models directly identify and measure the effects of specific
forces that influence demand. If the causal relationships are known, these models will be more
appropriate than Time Series Models.

Models include:

- Linear regression models

- Multiple regression models
- Non-linear models

Selecting and Monitoring Forecasting Systems

Before a demand forecast can be prepared, the following questions should be answered.

- For which goods and services is the firm attempting to forecast demand? What level of
aggregation will be used? What is the time horizon of the forecast, and how long is each
time period in the time horizon?

- Which sources of demand will be considered? How will the necessary data be obtained?

- Are there any cause-and-effect relationships in the data? Are there any trend, seasonal,
or cyclical components?
- What is the purpose of the forecast? Is the firm trying to predict the effects of its actions?
Project an existing demand pattern? Predict a turning point?

- Who will be using the model? With what level of forecast sophistication are the ultimate
users of the forecast comfortable? ( If the users do not understand the model, the
chances are slim that the forecasts generated by the model will be used.)

- How accurate must each forecasts be? What is the cost of forecasting errors? What is
more detrimental: underestimating or overestimating demand?

- How quickly must each forecast be prepared? How much time and money can be spent
developing, implementing, and monitoring the forecasting model?

Once these questions have been answered, one or more forecasting techniques can be
selected. Often, a mix of qualitative and quantitative techniques are used. Here are some
hints:

 Time series techniques are suitable for detailed short-term forecasts.

 Exponential smoothing models have relatively minimal data storage and
computational requirements compared with other time series techniques; these
costs are important if forecasts are frequently updated.
 Time series models assume that the past explains the future; this often not a realistic
assumption.
 Causal models provide insight into cause-and-effect relationships. These models are
expensive to develop and operate; they are more suited to long-term forecasting.
 Qualitative models are also used for long=term forecasting. They provide good
predictions of what will happen as well as insights into why it will happen. Where
there are few past data, these models should be used.

Often a forecast’s accuracy can be improved; this will cost the firm in terms of data, time
and money.

Measuring and Using Forecast Errors: Forecast errors is the difference between actual
demand and forecast demand for the same period. Measures of forecast errors are used
to help judge forecasting model accuracy.

Example: RSFE (running sum of forecast error) – is a model that indicates the bias of a
model. A positive RSFE indicates a model that tends to underestimate demand; it is
positively biased.

Prepared by: Mr. Romano D. Flores- Instructor (CBA)-BPSU