Demand Forecasting

Significance
Tools used for Demand Forecasting/Planning
Demand

Forecasting is the estimation of future demand which

helps managers
- to be proactive so that demand supply can be
managed and profits maximized
- Demand is essential for all planning and strategic
decisions
 Push vs Pull View –an Example
All push processes are performed in anticipation of demand
All pull processes are in response to customer’s demand

For push processes managers must plan level of production

whereas for pull processes managers must plan level of
capacity to be made available

In both cases forecasting demand is essential

e.g. direct marketing in DELL
 Dell has both Push and Pull processes:-
Push processes are Procurement cycle by Dell and
Components order Cycle by Intel
Customer order cycle and Assembly Cycle at Dell

On the side of Dell, its Production Manager must

ensure that that adequate number of components are
ordered in anticipation of demand ; and ; that the
requisite level of capacity is available for meeting actual
demand
On the side of Intel , it provides processors to DELL and since
processors take weeks and DELL can not wait for too long after a
customer order, therefore INTEL too produces processors in
anticipation of demand

Both DELL and INTEL forecast future demand , DELL uses this
forecast for replenishing components and processors needed in
future whereas INTEL uses the forecast for production of
adequate no of processers for DELL
Similarly production at INTEL drives demand for Intel
Suppliers, who must too forecast future demand and produce so
that Intel can meet its future production needs
Other decisions which use demand forecast
Production: Scheduling, Inventory control, Aggregate
Planning

Marketing: Sales force Allocation, promotion,, new

production Introduction

Finance: Plant/Equipment Investment, Budget Plan

Personnel: Workforce Planning, hiring, layoff

Characteristics of Forecasts
1. Forecasts are relative/ subjective and are prone to error
of judgment so these must include expected value and a
measure of forecast error(demand uncertainty) e.g.
100-1900 and 900-1100
2. Long term forecasts are less accurate than short term
forecasts i.e. long term forecasts have more standard
deviation of error relative to mean . For example
forecasting what will sell in less than 12 hours before
actual sale is more predictable
3. Aggregate forecasts are more accurate than disaggregate
forecasts
For example GDP of a country in a year as compared to
yearly earnings of a company with less than 20% error
or even worse the earning from each product line
within a company
In nutshell the higher is the level of aggregation less
is the error in forecasting demand
Components of Forecasting and Methods
Predictions are difficult especially about future, however it
must be understood that, the demand does not arise in vacuum.
Customer’s past behaviour can shed light on future behaviour as
well . Customers are influenced by a variety of factors, and,
demand can be predicted if a firm can determine relationship
between theses factors and future demand

Companies must balance subjective and objective factors

while forecasting the demand. There are several quantitative
methods but must include human input whenever final forecast
is made (e.g. in Ice cream demand )
Factors related to Demand Forecast
Past Demand
Planned Advertising and Marketing efforts
Display position in a catalogue
State of Economy
Planned price discounts
Actions of competitors

A company must take into account the above factors while applying
appropriate forecasting methodology( e.g. demand for soups in October
when the past data shows demand upsurges in November and December)
 Methods of Forecasting Demand
There are four methods:-
1. Qualitative Methods (used in new Industry)

2. Time Series Methods e.g. Regression, Simple

Averages, Moving average, exponential smoothening
etc.
3. Casual Forecasting Methods
4. Simulation Methods
Qualitative Methods: These are subjective, rely on
human judgment and opinion when there is little past
data available and when experts have market
intelligence about some happening which is critical
Time Series: These use historical data for forecasting
of the demand and are based on assumptions that past
history is a predictable indicator of future demand .
Theses methods are good when environment is stable
and basic demand pattern do not vary significantly
Casual Methods: These methods are used when demand forecast is
highly co-related with certain factors in the environment( e.g.
economic, interest rates etc.). These methods find the co-relation of
demand and environmental factors and thus use these associations to
estimate demand forecasts (e.g. price)

Simulation: Simulation uses technology and can use combination

methods e.g. both time series and causal methods together . For
example by using simulation method a firm can get answers to
questions such as ; what will be the impact of price promotion? and /or
competitor opening store nearby ?

We will focus on Time Series methods

 Time series Methods
When demand is forecasted on the basis of past data the future
demand will be effected by current demand, any historical
growth patterns, and any historical seasonal patterns

Moreover, the anticipated demand will have a random element

which can not be explained by current demand , historical
growth pattern and the historical seasonal variation

therefore,
Observed Demand(O)= Systematic Component(S)+ Random
Component(R)
 Systematic Component measures expected Value of
demand and consists of i) level of the current de
seasonlized demand; ii) trend i.e. the rate of growth or
decline in demand for next period and iii)seasonality
i.e. the predictive seasonal fluctuations in demand

Random Component is based on forecast error i.e.

the noise and measures its estimated size and
variability
Time series methods fall in two categories:-
1. Static
2. Adaptive

Static: Companies can estimate level, trend and seasonality of the systematic
component once and do not update it as it observes new demands( by
using averages, regression estimates)

Adaptive: Companies update their estimates of level, trend and seasonality

of systematic component after they make each demand observation . They
assume that portion of error can be due to incorrect evaluation of the
systematic component also apart from random component( by using
moving averages, exponential smoothening techniques)
In nutshell static methods use average values of
level, trend and seasonality whereas, Adaptive
methods use moving averages, exponential
smoothening to take into account each demand
observation
Basic Approach to Demand Forecasting
Besides the choice of Methodology, all parties in Supply
chain should reach a consensus regarding forecast
assumptions, techniques, and final forecasts so that all
plans within supply supply chain are consistent and
supportive of each other. This can be achieved by the
following:-
1. Understand the objective of forecasting

2. Integrate Demand Planning and Forecasting

3. Identify major factors that influence the demand
forecast/accuracy desired
4. Identify Customer segments

5. Appropriate technique/ Forecasting

6. Establish performance and error measure of forecast

Let us understand in a little detail all the above

Understand the objective of forecasting:
- support decisions that are based on the forecasts
- link between decisions and forecasts ( e.g. when Wal-
Mart plans a detergent promotion in July then the
information has to be shared with manufacturer and
transporter)
- common forecast for promotion and shared plan of
action to avoid too much /too little product , transport
scramble due to lack of knowledge about envisaged
promotion activity
 - including knowledge about forecast decision on the basis of
geography, product, customer segment or aggregate plan
- Forecast horizon is the time lag between the point at which
forecast is made and the event being forecasted

Integrate Demand Planning and Forecasting:

A company should link forecast to all planning activities
within supply chain
e.g. sales and marketing and production activities not
integrated in the forecast process any promotional activities
envisaged. A firm must specify the desired forecast horizon.
Identify major factors that influence the demand forecast /accuracy desired: proper understanding of these
factors is crucial and central to developing a forecast technique
The main factors influencing forecasts are :-
- Demand
- Supply
- Product related phenomenon
Demand:
- On the demand side a company must ascertain whether
demand is growing, declining, or has a seasonal pattern .

- estimates must be based on demand data and not sales data( e.g. a super market may have
promoted a certain brand of cereal in July 2000. As a result , the demand for this cereal may have been
high while the demand for other comparable brand was low in july. Supermarket , thus, should not
use this sales data of july 2000 to estimate demand for july 2001 as this will only be the case if the same
brand is promoted again in july 2001 and the other brands respond the way they did previous year
 -while making the demand forecast, it must be understood that what
the demand would have been in absence of promotion activity and
how demand is affected by promotions. A combination of these two
pieces of information will allow the supermarket to forecast demand
for july 2001 given the promotion activity planned that year

-A company should also find out whether there is any relationship

between its different products(complimentary or substitution). If
promotion of one product is likely to steal demand from another
product, the compamy must account for this in the forecast.

- The Company should be aware of any planned changa in lead times

and service policies and their influence on demand
Supply:
 A company must consider the available supply sources to decide
on the accuracy of forecasts desired .If alternative sources with
short lead times are available, a highly accurate forecast may not
be especially important. However if a single supplier with long
lead time is available then accuracy is valuable
Product :
-No of variants of a product being sold and whether these variants
substitute or complement each other
-If the demand for a product influences or is influenced by demand
for other product, the two forecasts are best made jointly
e.g. when firm wants to introduce an improved version
of an existing product it is likely that the demand for
existing product will decline because new customers
will buy the improved version. Although the decline in
demand of the original product will not be indicated
by the historical data, historical demand will still be
useful in that it will allow the firm to estimate combine
total demand of two versions. Demand for these two
products should be forecasted jointly.
Understand and Identify customer segments: To understand and identify
customer segments, customers can be grouped by similarities in service
requirements, demand volumes, order frequency, demand volatility, and
seasonality. In general , companies may use different forecasting methods for
different segments. A clear understanding of customer segments facilitates
accuracy and simplifies forecasting
Determine Appropriate Forecasting Technique: In selecting appropriate
technique , a company should first understand dimensions that will be relevant
to forecast i.e. i) geographical area, ii)product groups, and iii) customer groups

Difference in demand across each dimension should be understood. It will be

wise to have different forecast for each dimension . Then , the firm selects an
appropriate forecast technique from four methods i.e. qualitative, time series,
causal and simulation and as well as combination methods
Establish Performance and Error measures for
Forecast: Clear performance measures to evaluate the
accuracy and timeliness of the forecast should be
established
For example: A mail order Company that uses a forecast to
place orders with its suppliers. Suppliers send in the orders
with a two month lead time and the products are then sold
Objective of company is to have order with a quantity that
minimizes the amount of extra product left over at the end
of sales season and eliminates any lost sales in the absence
of available products
The forecast is to be created at least two months before the start of
sales season because of lead time with its suppliers

At the end of sales season , company must compare actual

demand with forecasted demand to estimate accuracy of forecast
Observed Accuracy can be compared with desired accuracy;
resulting gap to be used for corrective actions on part of mail
order company
The desired accuracy is seldom attainable due to current ordering
and fulfillment process so sometimes steps like sending early
versions of its catalogues to select set of customers will provide
sufficient advance information to improve forecast accuracy