Demand Management SIBM

You might also like

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 38

Demand Management

What is Demand Management ?


• Defined as “focused efforts to estimate and manage
customers’ demand, with the intention of using this
information to shape operating decisions.”
• Set of activities and decisions, tools and techniques,
that firms adopt to assess, and predict the purchase of
company’s products and services.
• Seeks to forecast and even regulate, the quantity, mix,
price and timing of such purchases.

1
Demand Management Objectives
Gathering and analyzing knowledge about
consumers, their problems, and their unmet
needs.
Identifying supply chain partners to perform the
functions needed in the demand chain.
Moving the functions that need to be done to
the channel member that can perform them
most effectively and efficiently.

2
Sharing with other supply chain members,
knowledge about consumers and customers,
available technology, logistics challenges and
opportunities.
Developing products and services that solve
customers’ problems.
Developing and executing the best logistics,
transportation, and distribution methods to
deliver products and services to consumers.

3
Demand Forecasting
A major component of demand management is
forecasting the amount of product that will be
purchased by consumers or end users.
In the integrated supply chain all other demand
will be derived from the primary demand.
A key objective is to anticipate and respond to
primary demand as it occurs in the market
place.

4
Forecasting process comprises of two elements
(a)Nature of demand, and
(b)Forecast components

Nature of Demand

Dependent Demand Independent Demand

5
Dependent versus Independent
Demand
• Vertical dependent is characterized by sequence of
purchasing and manufacturing, such as number of
tyres used for assembly of automobiles.
• Horizontal dependent occurs in a situation where an
attachment, promotion item or operator’s manual is
included with each item shipped.
(a)The demanded item may not be required to complete
the manufacturing process but may be needed to
complete the marketing process.

6
(b) Once manufacturing plan for base item is
determined , requirements of components/
attachments can be calculated directly and no
separate forecasting is done.
• Independent demands are ones that are not
related to the demand for another item.
• For instance, demand for refrigerator is not
related to the demand for milk.
• Independent demand items are forecasted
individually.

7
Forecast Components
1. Base demand
2. Seasonal factors
3. Trends
4. Cyclic factors
5. Promotions
6. Irregular quantities.
Mathematically forecast is expressed as
Ft+1= (Bt x St x Tt x Ct x Pt) + I, where
- Ft+1= forecast quantity for period t+1
8
- Bt= base level sales demand (average sales level) for
period t+1
- St= seasonal factor for period t
- T= trend component (quantity increase or decrease
per time period)
- Ct= cyclic factor for period t
- Pt= promotional factor for period t
- I= irregular or random quantity.
 All forecasts may not include all components.
A. Base demand is based on average demand over an
extended period of time.
(a)There is no seasonality, trend, cyclic or promotional
component.

9
B. Seasonal component is characterized by upward and
downward movement in demand pattern, usually on
annual basis e.g. emand for woollen blankets is at
peak during winter months and lowest during
summer.
(a) Seasonality at wholesale level precedes consumer
demand by approximately one quarter.
(b) An individual seasonality factor of 1.2 indicates that
sales are projected at 20% higher than an average
period.
C. Trend Component exhibits long range movement in
sales over an extended period of time.
(a) Trend may change number of times over the entire
product life cycle.

10
(b) For instance, a reduction in birth rate implies
reduction in demand of disposal diapers.
(c) Trend component influences base demand as
Bt+1 = Bt x T, where
- Bt+1 = base demand in period t+1
- Bt = base demand in period t, and
- T= periodic trend index.
D. Cyclic component are known as business
cycles.
(a)Economies swing from recession to expansion
every three to five years.
11
E. Promotions are initiated by the firm’s marketing activities
such as advertising, and various other schemes.
(a) Sales increase during promotion as the consumers take
advantage of promotional schemes thus leding to
liquidation of inventories.
(b) Promotion can either be the deals offered to the
consumers or deals offered to the trade (wholesalers/
retailers).
(c) Promotions if offered on regular basis at the same time
every year will resemble a seasonal component.
F. Irregular components include random or unpredictable
quantities that do not fit into any other category hence are
impossible to predict.
(a) By tracking and predicting other components the
magnitude of random component can be minimized.

12
Forecasting Methods
Forecasting techniques can be divided into three
categories
(a) Qualitative methods
(b) Time-series methods, and
(c) Causal methods
 Qualitative Methods
- Subjective and judgmental and based on
opinions and estimates.
 Time series and causal methods
- Employ numerical data collected over a period
of time to predict future trends.
13
Demand Forecasting

Qualitative analysis Quantitative analysis

Customer Sales force Time series Causal


composite analysis
survey analysis

Simple Trend analysis


Executive Delphi Simple
moving
opinion method exponential
average
smoothing

Past analogy Holt’s double Winter’s triple


Forecast by linear
Exponential Exponential
regression
smoothing smoothing
 Qualitative Methods
- Based on judgments regarding factors
influencing demand and opinions about the
probability of factors affecting the demand.
- Methods range from scientifically conducted
opinion surveys to intuitive predictions about
future events.
• DELPHI Method
- Olaf Helmer and N.C.Dalkey of RAND
Corporation developed Delphi method in the
1950s while working on a project sponsored by
the US Air Force.
15
- The Delphi Method generally consists of the following
steps.
1. Selection of groups of experts, depending on the type
of expertise required.
2. Ideas and forecasts are obtained from all participants,
usually through a questionnaire.
3. The results are summarized and redistributed among
participants, along with appropriate new questions.
4. Any member whose response deviates from the
opinions of majority is requested to reconsider and
provide justification for the deviation.

16
5. The responses are again summarized, and new questions
are developed on the basis of the responses.
6. This cycle is repeated till the results are in a range
narrow enough to be used as a forecast.
- Success of this technique depends on the talent of the
coordinator and absence of bias on the part of experts.
- The coordinator should be competent enough to analyze
diverse and wide ranging statements and arrive at a
structured questionnaire as well as a coherent forecast.

17
Problems in this method are:
1. Members’ opinions might be influenced by a
socially dominant individual.
2. Members may fear the loss of credibility if
they back away from a publicly stated
opinions.
 How to overcome these problems?
- Membership is generally not revealed to the
panel and panel members are kept separate.
- The panel does not meet to discuss or debate
the issue.

18
• Nominal Group Technique
- Developed by Andrew Van de Ven, a Wharton Professor.
- The steps involved are:
1. Generation of ideas
- Group members write down their ideas regarding the
question/problem posed by a mediator.
2. Collection of ideas
- Group members’ ideas are collected and recorded on a
flip chart or blackboard that is visible to all members.
- No discussion is permitted during this stage.

19
3. Discussion
- Each idea is discussed.
- To avoid any wastage of time, similar or
duplicate ideas are clubbed together and
discussed.
- The ideas are discussed in terms of their
perceived importance, clarity and logic.
- Members are allowed to make brief impersonal
comments, on a voluntary basis on each idea.
4. Preliminary voting.

20
- Members are asked to cast their preliminary
vote to select the best idea.
- If there is no consensus regarding the best
idea, the ideas concerned are discussed further
to clarify their meaning and logic.
5. Final Voting
- Members are asked to cast their final vote.
- The result of the final vote is counted and the
most preferred idea, solution or forecast is
identified.

21
Time-Series Method
- Assumes that the past data is a good indicator of the
future.
- Instances when this assumption is not true are rare and not
significant enough.
- Hence, many operations managers use a time series model
to forecast the demand for their goods or services.
 Simple Moving Average
- SMA technique forecasts demand on the basis of the
average demand calculated from actual demand in the
past.

-
22
• SMA method is effective when a product does not
experience fluctuation in demand over a period of time
and the past demand for the product is not seasonal.
• Useful for removing any random fluctuation in demand
to get accurate forecasts.
• Mathematically SMA is calculated as
 Ft = (Dt-1 + Dt-2 + Dt-3 + Dt-4 + ………+ Dt-n)/ n
- Ft = forecast for the period t
- n= number of preceding periods taken for averaging
- Dt-1, Dt-2, Dt-3 and so on =Actual demand in the
preceding time periods.

23
• One of the key decisions to be taken when using
SMA method is the length of the time period to
be considered.
• The greater the moving average period, the less
vulnerable the forecast to random variations.
• A larger moving average period is taken when
fluctuations in demand are minimal.
• A small time period is taken when fluctuations
in the demand are high or when there is no need
to identify short-term fluctuations.

24
 Weighted Moving Average
- Sometimes the forecaster wants to use a moving
average but does not want all the ‘n’ periods
equally weighted owing to some trend and
seasonality in demand.
- Experience and trial and error methods are used
to assign weights to a particular data.
- Each element is weighted by a factor and sum of
the weights should be equal to one.
- Mathematically,
- WMAt+1 =∑Ct At,
-

25
- WMAt+1= W.M.A at the time period ‘t+1’,
- At = Actual demand in time period ‘t’
- Ct = Percentage weight given to time period ‘t’;
0<=Ct <= 1 and C1 + C2 + C3 + +Cn = 1
Exponential Smoothing Method
- Based on the assumption that the most recent
data is a better indicator of future trends than
the past data.
- Useful when used on data characterized by
seasonal tendencies.

26
- Demand for the most recent data is given the
maximum weightage.
- Weights assigned to the preceding periods
decrease exponentially.
- The data required for making forecast are the
most recent forecasts, the actual demand for that
period and a smoothing constant ()
- The value for  lies between 0 and 1.
• First-order exponential smoothing
• Ft = Dt-1 + (1- ) Ft-1

27
• Ft-1 = Forecast for the period ‘t-1’
• Dt-1 = Actual demand for period ‘t’
  =Smoothing constant
 Selecting a smoothing coefficient ()
• The smoothing coefficient  takes any value between 0 and
1.
• High  results in more weightage for the most recent
months and low  results in a relatively lower weightage for
it.
• A high  is more appropriate for new products for which
demand is more dynamic and unstable.
• If demand is stable and believed to represent the future, a
low  can be selected to smooth out the effect of any noise.
28
Causal Quantitative Models
• The demand for product or service is
dependent on different factors or variables like
price, quality, availability of substitute and/or
complementary products/ services, income
levels of customers, number of competitors,
etc.
• Organizations must identify the variables that
affect the demand for a product and service.
• A causal method evaluates the relationship
between different variables and their influence
on each other.
29
• Causal methods include linear regression and
multiple regression analysis.
Linear Regression
- Refers to the functional relationship between
two or more correlated variables.
- Linear regression refers to the functional
relationship between a dependent variable, for
which the forecast is needed, and a group of
other variables, known as independent variables,
which influence the dependent variable.

30
- For instance, let us assume that the sale of
televisions is dependent on the advertising
budget and the number of retailers.
- In this case, television sales is the dependent
variable and the advertising budget and the
number of retailers are independent variables.
- In linear regression, the relationship between
the dependent variable and one independent
variable is defined by a straight line.

31
• Y= a+bX,
• Where Y= Value of dependent variable
• X= Value of independent variable.
• a=Y intercept (constant value)
• b=Slope of the line
- Widely used by operations managers because it predicts
demand with high level of accuracy.
- Useful in long term forecasting of major occurrences and
aggregate planning.
- Useful for forecasting for product families, where demand
for individual products within the family may vary widely
during a time period though the demand for total product
family remains smooth.
32
Least Square Method
• Used to generate a regression model by
assigning data to a single line.
• Past demand data is used to form a linear model
by regressing data points to a single line.
• Once a linear equation is formed, future demand
(Y) can be predicted by substituting value of X.

33
Selecting a Forecasting System-Time Span
Time Decision Areas Techniques used
Horizon
Short- Purchasing, job scheduling, Time series –
term project assignment, and machine SMA,WMA and
scheduling Exponential
Smoothing.
Medium Capital and cash budgeting, sales Regression
term planning, production planning, analysis
and inventory budgeting

Long- Product planning, facility Regression


term location and expansion, capital analysis, Delphi
planning method and
market research.
34
Measures of Forecasting Accuracy
• Since the demand for a product is dependent on
various factors, all of which cannot be represented in a
forecasting model, obtaining accurate results from
forecasting methods is highly improbable.
• A forecasting error is the difference between the
forecasted demand for a particular period and the
actual demand in that period.
• To determine how well the forecasts from a
forecasting model fit with the actual demand pattern,
the average error of the model is calculated.

35
• The average error of a forecast model provides
a measure for examining how well the forecast
value of demand matches the pattern of past
data.
 The measures of forecasting errors are:
(a) Mean Absolute Deviation
• Mean of the errors made by the forecast over a
period of time without considering the
direction of error.
• Does not determine whether the forecast was an
overestimate or underestimate.
• MAD= 1/n ∑│At – Ft │
• │At – Ft │indicates the absolute value of
deviation.
36
(b) Mean Square Error
• Mean of squares of deviations of forecast
values from the actual result is calculated.
• MSE= 1/n ∑ (At – Ft )2
• Large errors are penalized more than the small
ones because of squaring .
(c) Mean Forecast Error
• Calculated in the same way as MAD,only
difference is that in MAD, the absolute values
are taken in consideration whereas in the MFE
method, the real values are taken.
37
• MFE= 1/n  (At – Ft)
• The closer the value of MFE to zero, the more accurate
the forecast is.
(d) Mean Absolute Percentage Error
• MAPE indicates relative error.
• MAPE= 100/n  ⃒At – Ft ⃒ At
Tracking Signal
• Measure of accuracy that assesses the accuracy with
which forecasting methods are able to predict the demand.
• TS=Actual demand – Forecast Demand  MAD or
RSFE MAD

38

You might also like