Purchasing Strategy & Forecasting. 2 Purchasing Strategy: Answering Questions of Future Question 1:...
-
Upload
augustus-mills -
Category
Documents
-
view
230 -
download
2
Transcript of Purchasing Strategy & Forecasting. 2 Purchasing Strategy: Answering Questions of Future Question 1:...
Purchasing Strategy & Forecasting
2
Purchasing Strategy: Answering Questions of Future
Question 1: What do you need?
Question 2: Will you buy it from someone else or make it
yourself?
3
Make or Buy Decision
It is not always economical for the companies to make all the materials used in manufacturing.
Some items are procured from others, and some are produced in the company.
4
Some Reasons For Making:
Lower production cost Unreliable or unsuitable suppliers Assure adequate supply (quantity) Utilize surplus labor capacity Obtain desired quality Protect special design or quality
5
Some Reasons For Buying:
Lower acquisition cost Inadequate capacity Reduce inventory costs Ensure alternative sources of supply Item is protected by a patent or trade license
6
Vertical Integration
Choosing between making or buying an item is largely dependent on the vertical integration strategy of A company.
By vertical integration, we mean developing the ability to produce goods or services that are previously purchased.
It can take the form of forward or backward integration
7
Forward vs. Backward Integration
CURRENT PRODUCTIONRAW MATERIAL BUYERS
BACKWARD INTEGRATION
FORWARD INTEGRATION
STEEL AUTOMOBILE DEALERS
COMPUTERSSILICON CURCUIT BOARDS
8
Purchasing Strategy Questions….
Question 3: Who is going to supply your need?
How many suppliers will you work with?
9
Strategic Supplier Partnering
Many business firms in the world borrowed the Japanese concept of extremely close supplier interactions and cooperation.
This strategic partnering involves
Selecting the “best” suppliers, Working closely with them, and Entering into long-term relationship based on
mutual need and trust.
10
Purchasing Strategy Questions….
Question 4: How much and what time are you going to
buy?
11
Forecasting
When demand is uncertain, Someone should forecast the quantity to be purchased.
A forecast is an Inference of what is likely to happen in future.
Forecast can be wrong.
Businesses may use Forecasts in several subjects.
Some of the major forecasting areas are (1) Economic Forecasting, (2) Technological Forecasting, and (3) Demand Forecasting.
12
Economic forecast
Economic forecast is a prediction of what general business conditions will be in the future.
Some examples of economic forecasting are: inflation rates, gross national product, personal income, tax revenues, level of employment, and so on.
Economic forecast is usually made by government agencies, banks, and econometric forecasting services.
13
Technological forecast
Technological forecast predicts the probability and significance of possible future developments in technology.
What technology will the firm’s competitors incorporate into their products and processes?
Are there any technological advances with which the firm can create a competitive advantage?
14
Technological forecast…
For example, development of electric cars seems like a challenging shift for car manufacturers.
But what time and how will they be in the market is a concern of technological forecasting. Toyota Prius: A hybrid
(electric + oil) car
The forward-thinking 2005 gas/electric Prius with Hybrid Synergy Drive® offers fuel economy and cutting-edge available features like Bluetooth® technology -- all with the performance of a conventional car.
Plus, you never need to plug-in for recharging.
15
Demand Forecast
Demand Forecast predicts the quantity and timing of demand for a product or material.
Factors affecting the forecast are: Status of the general economy, Time of the year, Competitor’s actions, Advertising and sales promotions, New product entries to the market, etc.
16
Forecasting Methods
A forecast can be developed through either a subjective approach or an objective approach.
Subjective approaches are qualitative in nature and they are usually based on the opinions of people (that is why they are subjective).
Objective approaches involve quantitative methods and mathematical formulations.
(They can also be referred as statistical forecasting)
17
Subjective (Qualitative) Forecasting Methods
1) Executive Committee Consensus
2) Delphi Method
3) Sales Force Composite
18
1) Executive Committee Consensus
Here, a forecast is developed by asking a group of knowledgeable executives to discuss their opinions regarding the future values of the items being forecasted.
It provides forecast in a relatively short time.
But, presence of a powerful member in the group may prevent the committee from achieving a consensus.
In addition, it requires the valuable time of highly paid executives.
19
2) Delphi Method
The delphi method also involves a group of experts who eventually develop a consensus.
They usually make long range forecasting for future technologies or future sales of a new product.
The difference here is that, the panel members are located in different places and do not know each other.
This reduces the influence of powerful executives.
20
2) Delphi Method…
There is one coordinator who knows all the participants, and all participants only contact with the coordinator.
First, each member completes a questionnaire and returns it to the coordinator.
The results are summarized by the coordinator and a new questionnaire is developed based on these results.
This summary report is sent back to the participants.
21
2) Delphi Method…
The participants review this report and they either defend or modify their original views.
The process is repeated until a consensus is reached.
The quality of the consensus and final decision is largely dependent on the coordinator.
22
3) Sales Force Composite
Sales people are a good source of information regarding customers’ future intentions to buy.
They can help a firm obtain a forecast quickly and inexpensively.
Each sales representative is asked to estimate sales in his/her territory.
23
3) Sales Force Composite…
These individual estimates are then combined together by upper managers to develop regional sales forecast.
This method is more suitable for forecasting sales volume of a new product.
But still it is subject to opinion based terms.
24
Quantitative Forecasting Methods
Quantitative forecasting methods employ mathematical models and historical data to predict demand.
The first step in developing a quantitative forecast model is to collect sufficient data on past levels of demand.
For example, data obtained for at least 2 to 3 years of past are desirable.
In addition, the effects of unusual or irregular events that caused a change in demand should be removed from the data (such as natural disasters, or Olympics).
25
Two types of quantitative forecast models There are two major types of quantitative forecast
models:
1) time series models2) causal models
The main difference between the two models is that:
In time series modeling technique, the only independent variable is the time.
In contrast, causal models may employ some factors other than time, when predicting forecast values.
26
Weather forecast is a causal model
Some factors: humidity, wind direction, and time of the year.
Time is a factor but not the only factor to determine the weather
27
Time Series Modeling
Time Series modeling involves plotting demand data on a time scale.
A time series is a sequence of chronologically arranged observations taken at regular intervals for a particular variable.
Daily, weekly, monthly sales data are examples for time series.
Time series are frequently analyzed to identify any 1) Trends, or 2) Seasonal Factors, or 3) Cyclical Factors.
28
Types of Time Series Models
There are two major types of time series models:
1- Smoothing Models
- Moving Average (Simple & Weighted)
- Single Exponential Smoothing
- Double Exponential Smoothing
2- Time Series Decomposition Models
- Additive Models
- Multiplicative Models
29
Exponential Smoothing
Exponential smoothing is an averaging technique that inherently assigns the highest weights to the most recent observation.
It successively assigns lower weights to older observations.
The value of the weights decreases exponentially.
30
Exponential Smoothing…
Basically, The forecast for the next period (Ft+1) is set equal to the Forecast for the current period (Ft) + a percentage of the forecast error in the current period, which is (At – Ft):
Ft+1 = Ft + (At - Ft)
The percentage is referred to as alpha () AND is chosen by the user. (0 1).
31
Example
We will calculate The monthly demand forecast for an Example by using two exponential smoothing models.
The first model uses an value of 0.2 and the second uses a value of 0.8
32
Example…
33
Example…
For = 0.2;
FMar = FFeb + 0.2 (AFeb – FFeb)
FMar = 1004 + 0.2 (920 – 1004) = 987.2
Note that; the first forecast value for January in Year 5 is chosen by substituting simply the Actual value in previous month (FJan = ADec = 1020).
Exponential Smoothing method ASSUMES an Initial Value for the first forecast period is given by the user.
34
Example…
For = 0.8;
FMar = FFeb + 0.8 (AFeb – FFeb)
FMar = 956 + 0.8 (920 – 956) = 927.2
Determination of An Alpha value is a critical issue in Exponential Smoothing.
The weights given to the data will be a function of alpha
Age of data
Weights Assigned to Data
=0.6
=0.4
=0.2
35
Single Exponential Smoothing Method
Since we use a single parameter (alpha) in this method , it is also called Single Exponential Smoothing Method.
Single Exponential smoothing is suitable for forecasting mean values that remain fairly Stable.
Because, single exponential smoothing does not anticipate a Trend in the data.
However, this model can be extended to include a Trend factor (or parameter).
36
Exponential Smoothing with a Trend
This extended version of Exp. Smoothing uses a second parameter (beta) to obtain a forecast that contains a Trend.
In this case, the Smoothed Forecast value for the next period (SFt+1) is:
SFt+1 = (At) + (1 - ) (SFt + Tt)
37
Exponential Smoothing with a Trend...
SFt+1 = (At) + (1 - ) (SFt + Tt)
Here, Tt is the Trend estimate for the current period t.
Therefore, the value of (SFt + Tt) in the Right Hand Side of the formula is actually a Trend-Adjusted Forecast value, which will be referred as TAFt.
SFt+1 = (At) + (1 - ) (TAFt)
38
Exponential Smoothing with a Trend...
The trend estimate for the next period is calculated by using (as a second parameter):
Tt+1 = (SFt+1 - SFt) + (1 - ) (Tt)
And, we will also need the trend-adjusted forecast for the next period:
TAFt+1 = SFt+1 + Tt+1
39
Example
Let’s recalculate the demand forecast for our regular example by using exponential smoothing method with trend values.
Assume that = 0.8 and = 0.5
Before beginning, we should set the initial values of SFDec and TDec.
Let’s assume SFDec = Actual demand in December = 1020
and TDec = 0.
40
Example…
Now we can continue by calculating the forecast values in January:
SFJan = (0.8) (ADec) + (1 – 0.8) (TAFDec)
Since, TAFDec = SFDec + TDec ; Then,
TAFDec = 1020 + 0 = 1020.
Therefore,
SFJan = (0.8) (1020) + (0.2) (1020) = 1020
41
Example…
Now, we should calculate the trend estimate in January:
TJan = (0.5) (SFJan – SFDec) + (1 – 0.5) (TDec)
= (0.5) (1020 – 1020) + (0.5) (0) = 0
Therefore, trend-adjusted forecast for January is:
TAFJan = SFJan + TJan = 1020 + 0 = 1020
42
Example…
Now we can continue by calculating the forecast values in February:
SFFeb = (0.8) (AJan) + (1 – 0.8) (TAFJan)
TAFJan = SFJan + TJan = 1020 + 0 = 1020.
SFFeb = (0.8) (940) + (0.2) (1020) = 956
Now, we should calculate the trend estimate in February:
TFeb = (0.5) (SFFeb – SFJan) + (1 – 0.5) (TJan) = (0.5) (956 – 1020) + (0.5) (0) = - 32 (Negative Trend)
Trend-adjusted forecast for February is:
TAFFeb = SFFeb + TFeb = 956 - 32 = 924
43
Example…
Now we can continue by calculating the forecast values in March:
SFMar = (0.8) (AFeb) + (1 – 0.8) (TAFFeb)
TAFFeb = 924
SFMar = (0.8) (920) + (0.2) (924) = 920.8
Now, we calculate the trend estimate in March:
TMar = (0.5) (SFMar – SFFeb) + (1 – 0.5) (TFeb) = (0.5) (920.8 – 956) + (0.5) (-32) = -33.6
Trend-adjusted forecast for March is:
TAFMar = SFMar + TMar = 920.8 – 33.6 = 887.2
44
Example…(forecast results)
45
Further predictions into the future
By assuming that The same Trend will be valid for the future,
We can use this method to make further predictions into the future values of demand.
This can be achieved by replacing the following formula for the Trend-adjusted Forecast:
TAFt+p = SFt + (p) Tt
Where, (t+p) denotes the pth period beyond the most recent period.
46
Further predictions into the future..
For example, if we wanted to forecast the demand for July; by assuming that the Trend value will not change after March:
We would use a p value of 4 because we would predict 4 months beyond March.
Therefore, TAFJul = TAFMar+4 = SFMar + (4) TMar = 920.8 + (4) (-33.6) = 786.4
But, as we can see from the previous Table, This prediction for July would not be a good forecast,
Because there is an obvious change in the direction of Trend on March.(from a negative Trend to a positive Trend).
47
Determination of alpha and beta
Again, here, determination of alpha and beta values is a critical concern in the Responsiveness and Correctness of the forecast.
Since there are No strict rules about selecting these parameters,
We should try many possible values of alpha and beta AND find the best range For our particular problem.
For example, a demand forecaster tries for alpha = 0.1 AND beta = 0.3
48
alpha = 0.1 ; beta = 0.3
Time
Demand
Actual
Forecast
Since alpha is low, forecast does not adjust (response) rapidly.
49
alpha = 0.3 ; beta = 0.1
Since alpha is higher, forecast follows the actual more closely.
But since beta is lower, trend had not been corrected at the end of the actual values.
Time
Demand
Actual
Forecast
50
alpha = 0.3 ; beta = 0.3
We can see a better estimate than the previous trials. Therefore, these values of alpha and beta are the best for this particular demand data.
But, don’t forget that, the best values of alpha and beta will be different for every other application.
Time
Demand
Actual
Forecast