Hundreds of products, a global network of suppliers and stocks, reducing orders when demand drops can be difficult as stopping a production line. It is important to have the information to detect shifts demand early so they can adjust for trends and send the right message to your suppliers. With a forecasting planning they can quickly respond when demand changes and make a plan on what they think it should happen in the future.
The first step in planning production is forecasting future demand. To describes all the planning necessary for making a product. Organize the resources needed to make the products. The production is base on forecast, since orders must be filled from stock. It is important to remember that the future will not be like the past and a prediction is preferable to a forecast.
Forecasting methodologies have been existed since the nineteenth century and is based on what we think will happen in the future, and allows business to create, modify and track their financial.
Sitater???
AIM
Project aim for this report is:
Comparing Qualitative forecasting methods with Quantitative forecasting methods, and identify forecasting techniques for production planning.
OBJECTIVES
The objectives are:
Investigating The Delphi technique.
Investigating Market Research technique.
Investigating Exponential Smoothing technique.
Investigating Box- Jenkins technique.
Comparing all the techniques together.
Comparing Qualitative forecasting methods and Quantitative forecasting methods with each other.
Identify forecasting application for production planning.
In which situation do we use the various forecasting methods?
CHAPTER 2 RESEARCH
2.1
WHAT IS FORECASTING
Forecasting can be considered as a method or a technique for estimates many future aspects of a business or other operation.
Forecasting is based on experiences of what has happened in the past and then make a forecast about future performances, future stocks and a determination on future trends or to plan in advances. It is a translating on past experiences into estimates of the future.
2.2
WHY USE FORECASTING
Forecasting is used to prepare for the future and to answer vital questions such as:
When and how will borrowed fund be repaid?
How much profit will the business make?
How much demand will there be for a product?
How much will it cost to produce the product?
How much money will the company need to borrow?
Referanse
Forecasting lead businesses to a financially successful and it is important for the business to develop new products, or new products line and decide whether the product or product line will be successful. Forecasting prevents the business from spending time and money, manufacturing, and marketing a product that will fail.
2.3
WHERE DO WE USE FORECASTING
Forecasting is used for all aspects of planning and every other management decision, where the selection will become effective at some point in the future.
2.4
STEPS IN FORECASTING
Figur 1 ( Vonderemse and White 2004: 136
2.5
FORECASTING TECHNIQUES
Forecasting techniques are divided into two categories:
2.6
QUALITATIVE FORECASTING
Qualitative forecasting methods are based on opinions of appropriate individuals and these techniques are used to forecast future trends and demand for a product.
Qualitative forecasting techniques involved primarily judgment of experts in the appropriate field to generate forecasts, intuition and subjective evaluation. The qualitative forecasting methodology does not require mathematical formal and statistics model.
Qualitative forecasting techniques are divided into more precise methods such as;
2.6.1 PERSONAL INSIGHT METHOD
Personal Insight method uses a single person who is well known with the situation to produce a forecast based on this person judgment, personally opinions, prejudices and ignorance. This is the most widely used forecasting method.
Weakness of this method it is unreliability, the method uses their own experiences and opinions to forecast will consistently produce worse forecasts than someone who knows nothing about the situation.
2.6.2 PANEL CONSENSUS METHOD
Panel consensus method collects together a group of people and gives a consensus. If there is no secrecy and the people on the panel talk friendly and openly, a genuine agreement can be the result. There may be difficult to combine the views of different people when a consensus cannot be found.
Weakness with the Panel Consensus is that everyone makes mistakes and problems of team working. One and all try to make the best decisions to please the boss and not all is comfortable to speak in groups.
2.6.3 HISTORICAL ANALOGY METHOD
Historical analogy method is based on life-cycles of similar products, services, or processes. This method uses likely demand from the actual demand to make the forecasting.
Weakness with this method is the extent of the analogy between the model and the forecasting is often not evident.
2.6.4 MARKET RESEARCH METHOD
Market research method takes a look of consumer opinions. Surveys are used to create potential demand. That involves constructing a questionnaire that gives information about the person, economic and marketing information. Market researchers collect such information in person at retail outlets and shops, where the purchaser. The researcher must be careful so people who are involve in the surveyed are representative of the desired consumer target. Market research method may be useful in discussing data sources and are based on good theory and information that are valuable for marketing decisions, panels and questionnaires.
Weakness with the market research is that will take time and efforts.
2.6.5 THE DELPHI METHOD
The Delphi method consists of a number of experts that are given a questionnaire. The replies from these questionnaires are analyzed and summaries are passed back to the experts. Each expert is then asked to reconsider their reply and is anonymous. This process repeated between three and six times. By this time, the range of opinions should be enough to help with a decision.
Weakness for the Delphi method is that if a quick reply is needed, the Delphi method will take too long.
2.7
QUANTITATIVE FORECASTING
Quantitative forecasting techniques involved primarily past data, non formal mathematical and statistics model, also called for intrinsic and extrinsic types. These methods are based on an analysis of historical data by using the time series of the specific variable of interest and are related to time series. A set of observations are measured at successive times or over successive periods and then make the past patterns in data and uses to forecast the future data. This is a forecasting method with multiple models that switches to the model that is currently doing the best forecast and as soon as another model becomes better, the forecast will switch model.
2.7.1 PROJECTIVE INTRINSIC METHOD
Projective intrinsic method uses historical values and demand to forecast the future. The simples of this method are to take an average of past demand and use this as a forecast for the future. It is vital to remember all observations older than some specified age can be ignored. Find a forecast from the average will give most value.
Weakness with the Projective Intrinsic method is depending on past value and stock of data is needed.
2.7.2 CAUSAL EXTRINSIC METHOD
Causal Extrinsic method uses a cause and a relationship between variables to forecast unknown values. Focus on long-term forecasts that use measurements to forecast the future and using additional related data past the time series data.
Weakness with the Causal Extrinsic method is not good on short-range to calculate separate past requirements.
Each of these judgment methods works best in different circumstances.
2.8
FORECASTING METHODS
2.8.1 THE DELPHI METHOD
The Delphi technique started in the 1963 by Henry Arnold, at the time the Delphi technique was called for the Rand Corporation to forecast of future technological capabilities that used in the military. Henry Arnold thought that human judgment as legitimate and that was useful in forecasts. The method for the collection of judgment was developed in the 1950s, where the Delphi method was developed. It was a set of procedures developed to improve methods of forecasting and to combine experts' opinions concerning the expected future. These days the Rand Corporation is known as the Delphi method.
The Delphi method is a qualitative technique to establish reliable consensus of opinion among a group of expert. It is a series of questionnaires send to a group of experts, these questionnaires are to elicit to the problems posed and to enable the experts to refine their views as the group works progress in their specified task. Responses are feedback to panel members where they may change their responses.
The method of creating a Delphi forecast is a variation of everyone submits a list of such items to the panel coordinator. Then the list is sent back to each panel members for evaluation and rating of likelihood of occurrence. Panel members may see something that they had not thought of and rate it highly. Furthermore, members may have second thoughts about items concerning themselves previously submitted. After a sufficient number of cycles, two or three times, the result is a list with high consensus.
PROCEDURE
The procedure begins with a planner and a research preparing the issue and the future. These are distributed to the respondents separately who are asked to rate and respond. The experts answer questionnaires in two or three rounds. After each round, a facilitator provides an anonymous summary of the experts' forecasts from the previous round and reasons they provided for their judgments.
The results are then tabulated and the issues raised are identified, the outcomes are then returned to the experts in a second round. They are asked to rank or assess the factors and justify why they made their choices. During a third round their ratings along with the group averages and lists of comments are provided. Furthermore, the experts are asked to evaluate one more time and the answer will decrease and the group will converge towards to the correct answer. This process will continue until an agreed is decided, where the final rounds determine the results.
Steps true Delphi method:
Identification of the problem: researcher identifies and articulates a problem.
Selection of experts: it is important to have a balance of people and a mix of individuals that no one is overly represented.
Complete the first questionnaire anonymously or independently: responses of the question are submitted to researcher.
The results of the first questionnaire are listed.
After reviewing the results, members submit new solutions.
They may make new estimates.
Researcher summarizes responses: key factors suggested by experts are complied and listed.
Feedback: the lists are returned to the experts.
Figure the Delphi method (European Journal of Innovation Management 2009)
http://www.emeraldinsight.com/journals.htm?articleid=1770777HYPERLINK "http://www.emeraldinsight.com/journals.htm?articleid=1770777&show=html"&HYPERLINK "http://www.emeraldinsight.com/journals.htm?articleid=1770777&show=html"show=html
One of the main parts in the Delphi forecasting is the selection of experts. The persons invited must be knowledgeable about the issue and represent a variety of backgrounds, around than ten to fifteen of experts can provide a good base for the forecast.
ADVANTAGE
The Delphi forecasting is one of the most significant versatility; the technique can be used in a wide range of environments, as well it is anonymity. Therefore, everyone feel better protected from criticism over their solutions and from the group members of different views in order to appear in agreement.
The main point with the Delphi method is to overcome the disadvantages of conventional committee action.
The Delphi method is useful in technological forecasting, in predicting the state of the market, economy, or technology advances five or more years from now. The experts never need to be brought together physically and the process does not require complete agreement by all panellists.
DISAVANTAGE
This technique is time consuming, which it ineffective when fast answers are needed. People are acting together in a group benefits from other ideas, which might be more insightful and pragmatic resolution to problems offered by people in interactive settings. A further drawback to use the Delphi technique is may be difficult for researcher to design an effective study, a survey and other respondent dependent research designs. The results are determined in larger part by how they are framed and conducted.
2.8.2
MARKET RESEACH METHOD
Market Research method is about analysis and recording of data about customers, competitors, the market and issues to marketing products and service. Used to determine which of the population will purchase the product and is based on age, gender, location and income level. This method collect data from sample of customers analyses their views and make inferences about the population. The analyst examines the sales behaviour in the research and uses it to predict sales in other markets. The results are then to be more accurate because the consumers in a test market actually use the product. The Market Research is able to answer where the market is for your idea and that may be a successful idea.
True the market research you can find information about:
Market Information is the quotation, lest sale data, and number of shares. How the customers are, where they are located, the quantity and quality they want.
PROCEDURE
Customers are asked to make their own forecast about their usage and buying, customer intentions will be based on judgment for the future requirements. The forecast period, and the population must be specified clearly and the Market Research is discovering what people want, need, or believe.
Market Segmentation is the division of the market or population. Personality differences, demographic- and geographic -differences and psychographic differences.
Research methods can be the SWOT analysis, Marketing Plans, Competitive Analysis etc.
Market Trends is the up- or downward of a market, during a period of time.
Market Size is to estimate the number of potential customers.
Market Analysis is information about the target market, competitor, risk analysis and advertising research.
Demand usually comes from existing customer, where the customers will be contacted by the organization. Mostly by email, because is low-cost method for staying in touch with your customers, or in the shopping centre.
ADVANTAGE
The good part with the Market Research is to identify customer behaviour and give useful information for the future. It is more accurate than a survey, because we can see the consumers use the product. The user customers have the best information on what the forecast should be base on. The forecast is trustable when the customers, or at least the major customer, are few in number.
DISAVANTAGE
Information about the Market Research is expensive and time consuming, it is difficult when a sensitive purchase decision is involved and those may be reluctant to provide information
2.8.3
THE BOX-JENKINS METHOD
The Box-Jenkins was founded by the statisticians George Box and Gwilym Jenkins in 1976. Back in days the procedure used to select from a group forecasting models that best fit to the set of a time series data.
The Box-Jenkins is based in a time series analysis and the method applies three models; Autoregressive AR, Moving Average MA and Autoregressive Integrated Moving Averages (ARIMA). These models represent the processes that illustrate the forecasts are stationary or non stationary. A stationary process is statistical properties over time, where the time series fluctuates around a fixed value and no trend involved. Non stationary is where the trends are changes or seasonal changes. The result is to find the best of a time series on past values of time series and to identify underlying time series that fit the best model.
The Box-Jenkins is a procedure which used a variable past data to select the best forecasting. Capture the past pattern and forecast the future.
The Box-Jenkins involved three basic activities:
Identifying the tentative model
Determining the models parameters
Testing the model
If the model developed in step 2 and 3 does not give the result that they expected, the process is repeated and a new model is chosen and tested.
PROCEDURE
Identifying the tentative model and make the data stationary, by differencing the data. Analyzing the autocorrelation and partial autocorrelations of the stationary data. Discover seasonality and use layers of the autocorrelation and partial autocorrelation functions of time series to find which autoregressive or moving average method is best. The goal is to detect seasonality.
Identify the order of regressive and moving average. Determining the parameters of the model and estimates the parameters in regression analysis.
Application of the model by testing the estimated conforms to the specifications of a stationary process. The results should be independent of each other and constant and variance over time. If the estimation is error, they have to go start again and try to make a better model.
The observation represented by a linear of previous observations, known as the autoregressive and an error term, known as the moving average associated with the current observation. A linear of error terms associated with previous observations. The error terms have no meaning, constant variance and are uncorrelated. Furthermore, the process will determine the number of terms in the autoregressive, moving average parts and determination of values for the parameters. By determining the number of parameters are estimated in the model can be reduced. This parameter reduction is really important for estimate procedure.
The Box- Jenkins method that assuming a tentative pattern that is fitted to the data so that the error will be minimized. The forecaster with explicit information, on theoretical, will determine whether the assumed pattern is appropriate or not. If the correct pattern is not found, the method provides additional clues to the analyst so the correct pattern is founded. When the correct pattern is selected, the analyst can be use to make forecasts.
Regression analysis observation has two components; the first part is what is explainable by the model and the second part is the error. The expected value of the error term is zero, and the terms are assumed to be uncorrelated with each other.
AUTOREGRESSIVE MODEL
An autoregressive AR model is based on linear function of past data and with a mean or constant term of zero can have an order of one, two, or it could exclude some of the lower order term. Is an extension of the regression model, the only difference between two is that, in the AR model the independent variable are simply lagged values of the dependent variable.
MOVING AVERAGE MODEL
The moving average MA model involves linear combination of past errors and with current value of the time series to random errors that have occurred in previous time periods. And is a direct and predictable result of past random errors. Mean of an MA model is constant term in the model as we assume the expected value of the error term to be zero. The moving average is a way to calculate the next period. A forecast is made by averaging a number of periods to predict the next period.
AUTOREGRESSIVE INTEGRATED MOVING AVERAGE
Autoregressive Integrated Moving Average ARIMA the third model of Box- Jenkins is a combination of AR and MA into one, the autoregressive integrated moving average. Use a combination of past values and past errors. The past values and errors are used to make future forecasts.
ARIMA describes linear stochastic model, for understanding the data better, to predict future points in the series mode, easier to observable the process.
ADVANTAGE
The Box- Jenkins provides some of the most accurate short term up to six months. Estimates are easily constructed and the method is able to capture data patterns, give a set of data that is the best model and this forecasting approach can handle complex data patterns.
DISADVANTAGE
The method requires a very large amount of data and more complicated than the other time series models.
2.8.4
EXPONENTIAL SMOOTHING METHOD
Exponential Smoothing or more accuratley, single Exponential Smoothing was developed by C.C Holt, Brown and Holt-Winters in 1965. Holt and Brown applied this technique to the forecasting of demand of inventroy control problem and developed Exponential Smoothing model for constant process, processes with linear trends and for seasnoal data. Expoenential smoothing are generally used by three basic variations; Simple Exponential Smoothing, Tren- Corrected Exponential Smoothing and Holt-Winters' method.( International Journal Instiuate of Forecasters:2010)
Exponential smoothing is a way to take some of the random effects out of a time series by using all time series values up to the current period.
The technique is to produce a smoother data, a time series data. A Time Series is a sequence of observation which are a collection of data over time and have tree updating equations, each with a constant that ranges from zero to one.
The time series are consist of four components: trend, seasonal, irregular and noise. However, in smoothing tenchnique is only applied to irregular component, to providing specific values for the time series. The goal is to smooth out the irregular component of the time series by using an averaging process. Once the time series is smoothed, it is used to estimates for forecasts. Exponential Smoothing is applied to financial market, systems engineering, educational psychology and economic data, to reduce random fluctuations in the series data and give an effective means of predicting future values of the forecasting.
Figure Element of a Time Series (Data & Analysis Services for Education and Industry 2006)
http://www.brighton-webs.co.uk/time_series/concept.asp
The trend is the variable that changing over time fluctuates and might be the subject of random variations.
The seasonality is based on quarterly fluctuations and identifies the seasonal variations in weeks or months.
The noise is random deviation from the expected value and the effect of noise is reduced by averaging.
Exponential smoothing is based on the moving averages method and averaging past values. On the idea that as data gets older it becomes less relevant and should be given less weight. Uses the forecast for the first period based on the actual value for the most recent period. The forecast for the second period is equal to the actual value of the previous period, overlapping observations to produce averages. This method make the long term of a time series clearer and to smooth out past data by averaging the last several periods and analytical that view forward. The equations are intended to give more weight to recent observations and less weights to observationss further in the past.
PROCEDURE- THE MOVING AVERAGES MODEL
The forecaster would drop the first observations and calculate the average of the next three observations. The process continue intul three periode averages are calculated.
The forecaster moves up or down the time seris to pick observations to calculate an average of a number observations. The moving averages method would use the average of the most recent three observations of data tn the time series as the forecast for the next periode. The measurements can be taken in every hour, day, week moth or year.
The Moving Averages Formula
Ft = A t-1 + At-2+At-3+....+At-n / n
Figure the Moving Average ( Forecasting 2009)
PROCEDURE - BASIC EXPONENTIAL SMOOTHING
In the basic Exponential smoothing model, the base for the current period St is estimated by modifying the previous base by adding or subtracting to it a alpha (α) of the difference between the actual current demand Dt and the previous base St-1. The estimate of the new base is then:
New base = Previous base + α (New demand - Previous base)
The smoothing constant, alpha α, is between zero and one, with commonly used values from 0.01 to 0.03. to be specific, a new forecast is calculated from a section, alpha α, of the latest demand and a section, 1- α, of the previous forecast.
CALCULATED THE BASIC EXPOENENTIAL SMOOTHING
Basic Exponential smoothing is one period ahead forecaster
Ft= Ft-1+ α (At-1-Ft-1)
Where:
Ft= Forecast value for the coming time period
Ft-1= Forecast value in one past time period
At-1= actual occurrence in the past time period
Α = Alpha smoothing constant
In basic Exponential smoothing there are one more smoothing parameters to be determined. The forecast will be a constant which is the smoothened value of the last observation. The forecasting error is compared to the error in forecasts obtained
ADVANTAGE
Exponential smoothing is easier to implement and more efficient to compute, as it does not require maintaining a history of previous data values. Used for production control and error adjustment and give the best estimate of what the next values will be thats leading to more realistic data about the future
DISADVANTAGE
Using Exponential smoothing for forecasting calculates only the average all the observations are given equal weight, we axpect the more recent observations to be a better indicator of the future and only use recent observations
2.8.4.1
DOUBLE EXPONENTIAL SMOOTHING
This technique uses when the data is a trend, is similar to the basic Exponential Smoothing formula, by adding a second equation with a second constant, α, the trend component, which must be chosen in conjunction with alpha α. Double Exponential Smoothing applies basic Exponential Smoothing twice, it is useful where the historical data series is not stationary.
Double exponential smoothing
Xt + k = At + K Bt K = 1, 2, 3, ...
At and Bt = calculated by using linear regression
