Title: Lecture 6. Forecasting
1Introduction
2- Course content
- Demand Management
- Components of demand
- Qualitative technique in forecasting
- Time series analysis
- Causal relationship forecasting
3Forecasting
- Forecast plans It is not possible to make
decisions in production scheduling, purchasing,
and inventory levels until forecasts are
developed that give reasonably accurate views of
demand over the forecasting horizon. - Forecasting is a prediction of future events used
for planning process - Management needed accurate forecast to ensure
supply chain management
4Forecasting
- Accurate forecast allows schedulers to use
capacity efficiently, reduce customer response
times and cut inventories. - Managers need forecast to anticipate changes in
prices or costs or to prepare for new laws or
regulations, competitors, resource shortages or
technologies.
5Demand Management
- Demand Management
- - Demand management is to coordinate and control
all sources of demand so the productive system
can be used efficiently and the product delivered
on time. - Demand are of two types
- Dependent Demand
- If the demand can be tabulated from the end item
configuration. For developing a Car, number of
tires, wheel barrow etc can be known. Usually
sub-components of the final product. - 2. Independent Demand
- It cannot be derived directly from that of other
products. E.g sales prediction
6Handling Demand
- Handling Demand
- Holistic approach or active approach to influence
demand - (Developing aggressive marketing strategy to
influence demand) - 2. Passive role or simply respond to demand
- (Firm accept the demand and runs with it
passively accepting it) - Our primary interest is in forecasting for
independent demand
7Components of Demand
- Demand for product or service can be broken down
to six components - Average demand for the period
- Trend
- Seasonal elements
- Cyclical elements
- Random variation
- Autocorrelation
8Components of Demand
- Trend are the usual pattern of demand. Trends are
of 4 types. Trend is a systematic increase or
decrease over time.
Linear Trend It is a straight continuous
relationship
Asymptotic trends starts with the highest demand
growth at the beginning but then tappers off.
Objective of capturing the market and gradually
saturating it.
An exponential curve is common in products with
explosive growth. The explosive trend suggests
that sales will continue to increase.
S-Curve is typical of product growth and maturity
cycle. Important point in S-Curve is a point
where the trend change from slow to fast growth
sales
quarters
9Components of Demand
- 1. Seasonal Elements Seasonal fluctuation in
demand - 2. Cyclical Elements Cyclical influence on
demand comes from macroeconomic factors as
political, war, sociological pressure etc. Time
span is unknown and cause of cycle may not be
considered. - 3. Random variations are caused by chance events.
The cause for the this type of demand is unknown.
- 4. Autocorrelation It denotes the persistence of
occurrence. The value expected at any point is
highly correlated to its own past.
10Historical product demand consisting of growth
trend and seasonal demand
Number of units demanded
Trend
Average demand
seasonal
11Types of Forecasting
- Forecasting is classified into four basic types
- Qualitative Technique
- Time series analysis
- Causal Relationship
- Simulation
Quantitative Technique
12Qualitative Techniques of Forecasting
- Qualitative analysis are subjective, judgmental
and based on estimate and opinions - Grass Roots
- As per this, the person near to the customer
knows the market more better. His information is
taken as a base for further forecasting. - 2. Market Research
- Market survey, personal calling, data collection
etc are done to collect information from market
and then test the hypothesis to better understand
the management decision problem (MDP). - 3. Panel Consensus
- Group discussion to exchange the ideas. The
problem is the lower management staff may not
fully participate in idea sharing. - 4. Historical analogy
- Grouping of the customer based on product
category purchased. If an person has bought a
DVD, then it is likely that he is interested in
purchasing DVD movies. Also people who had
previously purchase some items before are
interested in new product of same category.
13Qualitative Techniques of Forecasting
- 5. Delphi method
- Delphi method is a form of panel consensus but
the identity of the individual participating is
not open. The view of everyone has same weight. - View of individual is known through open end and
closed end questionnaire. - The result from all the participant is
summarized. - Distribute the final result to participants.
14Time series Analysis
- Predict the future based on past data. The
analogy of past weeks data can be used to predict
the future data. - Short term forecast under 3 mts
- Medium term forecast 3 mts 2 years
- Long term forecast greater than 2 years.
- Short term compensate for tackling, problem in
hand, - Medium term compensate for seasonal effects
- Long term compensate for identifying change in
trends and consumer habits - Forecasting models available for firm are
- Time Horizon to forecast
- Data Availability
- Accuracy required
- Size of forecasting budget
- Availability of qualified personals
15Time Series Analysis
- Calculation of the forecasting based on time
series analysis can be done on following basis - Simple Moving Average
- Weighted Moving Average
- Exponential Smoothing
- Simple Moving Average
- The demand for the product or service is
relatively constant, neither growing nor
declining, with no seasonal slump, then in such
scenario, a moving average is preferred. - The simple moving average method is used to
estimate the average of a demand time series and
thereby remove the effects of random fluctuation.
16Simple Moving Average
- Applying a moving average model simply involves
calculating the average demand for the n most
recent periods and using it as the forecast for
the next time period.
Where D actual demand in period t n total
number of periods in the average
Forecast for period t1
With the moving average method, the forecast of
next periods demand equals the average
calculated at the end of this period
17Simple Moving Average
- a. Compare a three-week moving average forecast
for the arrival of medical clinic patients in
week 4. The number of arrivals for the past
three weeks were
Weeks Patient arrival
1 400
2 380
3 411
b. If the actual number of patient arrivals in
week 4 is 415, what is the forecast for week 5
18Simple Moving Average
- Solution
- The moving average forecast at the end of week 3
is
b. The forecast for week 5 requires the actual
arrivals from weeks 2-4, the three most Recent
weeks of data
The forecast at the end of week 3 would have been
397 patients for week 4. The forecast For week 5,
made at at the end of week 4, would have been 402
patients. In addition, at the end of week 4, the
forecast for week 6 and beyond is also 402
patient.
19Simple Moving Average
- Determining the value of n
- The stability of the demand series generally
determines how many periods to include (i.e the
value of n). Stable demand series are those for
which the average (to be estimated by the
forecasting method) only infrequently experiences
changes. - Large values of n should be used for demand
series that are stable and small values of n for
those that are susceptible to change in
underlying average. - If the underlying average in the series is
changing, however, the forecasts will tend to lag
behind the changes for a longer time interval
because of the additional time required to remove
the old data from the forecast.
20Simple Moving Average
Example Exhibit 13.5 Pg. 546 Eleventh Edition
Chase/ jacob
21Weighted Moving Average
- In simple moving average, each demand has the
same weight in the average. But in the weighted
moving average method, each historical demand in
the average can have its own weight. - The sum of the weights equal to 1.
- The Formula for a weighted moving average is
Where W1weight to be given to the actual
occurrence for the period t-1 W2weight to be
given to the actual occurrence for the period
t-2 Wn weight to be given to the actual
occurrence for the period t-n ntotal number of
periods in the forecast
22Weighted Moving Average
- A department store may find that in a four month
period, the best forecast is derived by using 40
of the actual sales for the most recent month,
30 for two months ago, 20 for three month ago
and 10 of four months ago. If actual sales
experience was -
January February March April May
100 90 105 95 ?
0.1 0.2 0.3 0.4
Find the forecast for month 6 if the sales for 5
mts turned out to be 110 (Ans 102.5)
23Weighted Moving Average
- Using the weighted moving average method to
estimate average Demand - a. The analyst for the medical clinic has
assigned weights of 0.70 to the most recent
demand, 0.2 to the demand one week ago, and 0.10
to the demand two weeks ago. Use the data for the
first three weeks from the table below to
calculate the weighted average for week 4. (Ans
403) - b. If the actual demand for 4th week is 415
Patients, what would be the forecast for week 5.
(Ans. 410)
Weeks Patient Arrival
1 400
2 380
3 411
24Exponential Smoothing
- Exponential Smoothing is method is actually a
weighted moving average method that calculates
the average of the time series by giving recent
demands more weights than earlier demands. - It is most frequently used for forecasting due to
its simplicity and the amount of data needed to
support it. - Weighted moving average requires n periods of
past demand and n weights, whereas exponential
smoothing requires only three items to calculate
demand - The last periods forecast
- The demand for this period
- Smoothing parameter alpha (a) (Value of a is
between 0 and 1)
25Exponential Smoothing
- The equation for forecast is
Smoothing constant a is the level of smoothing
and the speed of reaction between forecasts and
actual occurrences. Value for smoothing constant
can be taken from organization requirement as per
their volume of demand. Or mathematically it can
be taken as 2/(n1)
The equation for exponential smoothing
highlights, the old forecast error portion
between previous forecast and what actually
occurred.
26Exponential Smoothing
- E.g In the given table below, consider the
arrival of patients, at the end of 3 weeks, using
a0.10, calculate the exponential smoothing for
week 4. Assume initial forecast as 390
Weeks Patient Arrival
1 400
2 380
3 411
27Exponential Smoothing
- In the above example, if the demand for 4th week
becomes 415, the new forecast for week 5 would be
as follow
Conclusion Using the exponential smoothing
model, the analysts forecasts would have been
392 patients for week 4 and then 394 patients for
week 5 and beyond. As Soon as the actual demand
for week 5 is known, then the forecast for week6
will be Updated.
28Trend Effect in Exponential Smoothing
- Exponential smoothing has an advantages of
simplicity, minimal data requirement, inexpensive
and attractive to firm. - But its simplicity is a disadvantage if the
underlying average is changing, as in case of
demand series with a trend. - Higher values of Smoothing constant (a) may help
to reduce forecast error to some extent, when
there is a change in the average of the time
series however, the lag will still be there if
the average is changing systematically.
29Trend Effect in Exponential Smoothing
- Assume that actual demand is steadily increasing
at 10units per period. Forecast using exponential
smoothing with a0.3
As we see, forecast using exponential smoothing
with a0.3 will lag severely behind the actual
demand even if the first forecast is perfect. To
improve the forecast, we need to calculate an
estimate of the trend, we start by calculating
the current estimate of the trend which is the
difference between the average of the series
computed in the current period and the average
computed last period. Another smoothing constant
delta (?) is added to reduce impact of error
30Trend Effects in Exponential Smoothing Model
FITt Ft Tt Ft FITt-1 a(At-1 - FITt-1) Tt
Tt-1 ? (Ft - FITt-1 )
Ft the exponentially smoothened forecast for
period t Tt the exponentially smoothened trend
for period t FITt the forecast including trend
for period t FITt-1 the forecast including
trend made in prior period or period t-1 At-1
actual demand for prior period or period t-1 a
,? smoothing constants
31Assume a initial starting Ft of 100 units, a
trend of 10 units, an alpha of 0.20 and a delta
of 0.30. If actual demand turned out to be 115
rather then the forecast 100, calculate the
forecast for the next period.
Hence, the forecast for next period turned out to
be 121.3 with a trend of initial 100 units.
32Mean Absolute Deviation (MAD)
- MAD is the average error in forecasts, using
absolute values. - MAD is computed using the differences between the
actual demand and the forecast demand without
regard to sign. - It equals the sum of the absolute deviation
divided by the number of data points or stated in
equation as follow - Where
- tperiod number
- Aactual demand for the period
- F forecast demand for the period
- Ntotal number of period
33Mean Absolute Deviation (MAD)
Month Motorcycle sales
Jan 9
Feb 7
March 10
April 8
May 7
June 12
July 10
August 11
Sept 12
Oct 10
Nov 14
Dec 16
- Compute a 3 month moving average forecast of
demand for April through January (of the next
year) - Compute a 5 months average for June through
January - Compare the two forecasts computed in parts a and
b using MAD. Which one should the dealer use of
January of the next year.
34Mean Absolute Deviation (MAD)
35Mean Absolute Deviation (MAD)
- MAD is often use to forecast errors.
- When errors that occurs in the forecast are
normally distributed, the mean absolute deviation
relates to the standard deviation as - Standard deviation
Conversely, 1 MAD 0.8 Standard Deviation
- The ideal MAD is zero which would mean there is
no forecasting error - The larger the MAD, the less the accurate the
resulting model
36Mean Absolute Deviation (MAD)
- The value of MAD to forecast in case of
exponentially smoothing is as follow
37Measurement of Error
- Tracking Signal
- It is a measurement that indicates whether the
forecast average is keeping pace with any genuine
upward or downward changes in demand. - Tracking signal is the number of mean absolute
deviations that the forecast value is above or
below the actual occurrence. - Tracking signal (TS) RSFE/ MAD
- RSFE running sum of forecast error, considering
the nature of the error
38Measurement of Error
- Computing MAD and Tracking signal
In a perfect forecasting model, the sum of actual
forecast errors would be zero the error that
results in overestimates should be offset by
errors that are underestimate. The tracking
signal would then be also zero, indicating an
unbiased model, neither leading nor lagging the
actual demand.
39Linear Regression Analysis
- Regression is a functional relationship between
two or more correlated variables. - It is used to predict one variable to other. Or
more precisely, relation of dependent and
independent variables. - Linear regression line is of the form Ymx C
- Where Y is the value of dependent variable that
we are solving for, C is the intercept and m is
slope, x is the independent variable. - ? In linear regression forecasting, the past data
and future projection are assumed to fall about a
straight line.
40Linear Regression Analysis
- Linear regression is used in for both time series
forecasting and for casual relationship
forecasting. - When the dependent variable changes as a result
of time, it is time series analysis. - If one variable changes because of the change in
another variable, this is called casual
relationship. E.g Death of lung cancer increasing
with the increase in number of people smoking. - Casual Method provides the most sophisticated
forecasting tools and are very good for
predicting turning points on demand and preparing
long range forecast.
41Linear Regression Analysis
- Least square method fits the line to the data
that minimizes the sum of the squares of the
vertical distance between each data point and its
corresponding point on the line. - Equation of st. line is Yabx
Standard Error of Estimate
42Linear Regression Analysis
- Example
- Following are the sales and advertising data for
past five months. The marketing manager says that
the next month, the company will spend 1750 on
advertising of product. Use linear regression to
develop an equation and forecast for this
product.
Month Sales (Y) Thousands of unit Advertising Thousand of
1 264 2.5
2 116 1.3
3 165 1.4
4 101 1.0
5 209 2.0
43Linear Regression Analysis
Month Sales (Y) Thousands of unit Advertising(X) Thousand of X.Y X2
1 264 2.5 660 6.75
2 116 1.3 150.8 1.69
3 165 1.4 231 1.96
4 101 1.0 101 1
5 209 2.0 418 4
44Correlation Coefficient for regression
- Correlation coefficients shows the strength
between the dependent and independent variable. - The value of correlation coefficient lies between
-1 to 1. - If r-1, it shows, negatively correlated
- If r0, there is not linear relationship
- If r1, highly correleted
45Casual Relationship Forecasting
- Casual relationship forecasting is the one in
which the causing element is known enough in
advance, it can be used as a basis for
forecasting. - E.g increase in rain will increase sales of
umbrella - Increase in car accidents, increase in number of
insurance - Identify the occurrence that are really the
cause. Often leading indicators are not the
casual relationship, but in some indirect way,
they may suggest that some other things might
happen. - Other non casual relationships just seem to exist
as a coincidence.
46Casual Relationship Forecasting
47Important Questions discussion
- PU 2003 Fall
- 5.a) From the choice of a simple moving average,
weighted moving average, exponential smoothing,
and linear regression analysis, which forecasting
technique would you consider the most accurate?
Why? (7) - 4.a)what is the difference between dependent
demand and independent demand. Why do firms keep
inventory? (5) - 6.c) Explain the features of a good forecasting
technique. (5) - 2.B What do you mean by demand management?
Differentiate between dependent demand and
independent demand. (5)
48End of Lecture