Demand Forecasting and Supplier Selection for Incoming Material in RMG Industry: A Case Study
ABSTRACT RMG sector is the single most important manufacturing industry in Bangladesh. Almost all of the raw materials in this sector are being imported from abroad. Hence, incoming material management is of paramount importance for effective and efficient management of the supply chain in this sector. This paper deals with incoming material management of a 100% export-oriented knit composite factory. Demand forecasting and supplier selection are two major components of incoming material management. Different techniques of demand forecasting have been implemented to find the best suitable model for a particular raw material. In supplier selection, AHP technique has been implemented to select the best supplier of the concerned raw materials. Successful implementation of the recommendations of this paper can significantly improve the level of material management and thereby increase overall profit by reducing waste.
International Journal of Business and Management May, 2009
Demand Forecasting and Supplier Selection
for Incoming Material in RMG Industry: A Case Study
Executive, Lean Management, Viyellatex Group
297, Khortoi, Satais Road, Gazipur 1712, Bangladesh
Tel: 880-2-8912-870 E-mail: firstname.lastname@example.org
Abdullahil Azeem (Corresponding author)
Department of Industrial and Production Engineering
Bangladesh University of Engineering & Technology
Dhaka 1000, Bangladesh
Tel: 880-2-9665-611 E-mail: email@example.com
RMG sector is the single most important manufacturing industry in Bangladesh. Almost all of the raw materials in this
sector are being imported from abroad. Hence, incoming material management is of paramount importance for effective
and efficient management of the supply chain in this sector. This paper deals with incoming material management of a
100% export-oriented knit composite factory. Demand forecasting and supplier selection are two major components of
incoming material management. Different techniques of demand forecasting have been implemented to find the best
suitable model for a particular raw material. In supplier selection, AHP technique has been implemented to select the
best supplier of the concerned raw materials. Successful implementation of the recommendations of this paper can
significantly improve the level of material management and thereby increase overall profit by reducing waste.
Keywords: Demand forecasting, Supplier selection, Analytical Hierarchy Process (AHP)
Ready-made Garments (RMG) industry is the major export-base for Bangladesh. It has considerable impact on national
economy, as well as high of degree of social implications, as a large number of female workers are employed in this
labor-intensive industry. In the current post MFA era, international competition in this sector has been increased a lot.
Therefore, garments companies in Bangladesh need to become more competitive and efficient to survive, to retain
market position and increase market base. The foreign competitors have upper-hand basically in three areas: stronger
backward linkage, more skilled manpower, and better methodology of manufacturing. Among these three, backward
linkage is the most important factor as almost all of the raw materials needed in this sector are being imported from
different countries. Proper management of supply chain is of outmost importance for smooth operations of the
manufacturing processes in this sector that can help in maintaining the delivery schedule.
Highly competitive market is currently forcing every factory to think globally. Survival becomes increasingly difficult
and critical to find new ways to grow. Looking back at North American and European business trends, it seems that
strategies have been changing and updating frequently. “How to do more” was emphasized in 60s; “How to do it
cheaper” became important in 70s; “How to do better quality” was in the 80s and “How to do quicker” was the key in
the 90s. All of those are still important in our business, however meeting the increasing time demands of customers will
become important. Shorter lead time and will be the strategic focus for at least the next decade. Time – the number of
seconds, minutes, hours, days, months or years – is the yardstick by which we increasingly judge around us –
particularly organizations providing manufacturing services.
Demand forecasting is an integral part of any kind of supply chain management and very important to sustain
profitability. Improving demand forecasting performance has long been a concern of people involved in any kind of
industry (Armstrong & Grohman, 1972). To this end, researchers have developed and disseminated increasingly
sophisticated forecasting techniques, believed to be more accurately model the fluctuating demand patterns (Fildes &
Vol. 4, No. 5 International Journal of Business and Management
Hastings, 1994). However, improved forecasting techniques are useful mainly for different management practices
including decision making and planning processes (Winklehofer, 1996). Surveys of sales forecasting practice have
consistently shown that qualitative methods are more widely used than quantitative methods; however an extensive
body of research supporting the superiority of quantitative forecasting techniques in most situations (Dalrymple, 1987).
Not all the techniques are suitable for each category of the materials used in production. Out of several techniques,
proper implementation of the appropriate technique is very much important for accurate demand forecasting.
Analytic Hierarchy Process (AHP) is a multi-criteria decision-making approach and was introduced by Saaty (1977).
The AHP has attracted the interest of many researchers mainly due to the nice mathematical properties of the method
and the fact that the required input data are rather easy to obtain. The AHP is a decision support tool which can be used
to solve complex, unstructured decision problems (Putrus, 1990). It uses a multi-level hierarchical structure of
objectives, criteria, sub criteria, and alternatives.
Some of the industrial engineering applications of the AHP include its use in integrated manufacturing (Boucher &
McStravic, 1991), in the evaluation of technology investment decisions [Wabalickis, 1988], in location planning and
layout design (Cambron & Evans, 1991; Min, 1994], in software development (Finnie et. al., 1993), in project risk
assessment (Mustafa & Al-bahar, 1991) and also in other engineering problems (Wang & Raz, 1991; Shtub & Dar-el,
This paper focuses on selecting appropriate technique for demand forecasting of raw material in RMG sector. A
particular raw material, which is widely used in one selected factory of this sector is taken into consideration to
implement this technique. AHP technique has been implemented to find out the best suitable supplier of this raw
2. Forecasting Models
In this analysis, ten different techniques have been used to forecast the demand of raw materials used by the selected
factory. The techniques are simple average (SA), moving average (MA), weighting moving average (WMA), single
exponential smoothing (SES), single exponential smoothing with linear trend (SESLT), double exponential smoothing
(DES), double exponential smoothing with linear trend (DESLT), adaptive exponential smoothing (AES), linear
regression (LR), and holts-winters additive algorithm (HWAA).
Simple average (SA) method simply let the forecast equal to the average of all prior demand data. As time passed, our
forecasts would stabilize and converge towards the level of the series, because in the long run the noise terms will
cancel each other out because their mean is zero. The more data we include in the average, the greater will be the
tendency of the noise terms to sum to zero, thus revealing the true value. Sometimes the demand for an item in a
logistics system may be essentially “flat” for a long period but then undergo a sudden shift or permanent change in level.
In moving average (MA) technique, the forecast would be calculated as the average of the last “few” observations. If
number of observation is "small", the forecast will quickly respond to any "step", or change in level when it does occur;
the "averaging out" effect is lost which would cancel out noise when many observations are included. The optimal value
in any given situation depends in a fairly complicated way upon the level, the noise variance, and the size and frequency
of occurrence of the step or steps in the demand process.
It might seem more reasonable to assume that historical observations actually lose their predictive value "gradually",
rather than so "abruptly" as in the moving average. As a given data point becomes older and older, it becomes
progressively more likely that it occurred before the step change in level happened, rather than after it did. It therefore
might improve the accuracy of the forecast if relatively more emphasis is placed on recent data and relatively less
emphasis on less current experience. This idea leads to the concept of a weighted moving average (WMA) forecast,
where the last observations are averaged together, but where they are not given equal weight in the average.
A popular way to capture the benefit of the weighted moving average approach, while keeping the forecasting
procedure simple and easy to use, is called single exponential smoothing (SES), or occasionally, the “exponentially
weighted moving average”. In its simple computational form, a forecast is made for the next period by forming a
weighted combination of the last observation and the last forecast using various coefficients and taking the forecasting
error in consideration. An upward and downward trend in data collection over a sequence of time periods causes the
exponential forecast to always lag behind (be above or below) that actual occurrence. If such a trend is observed in
single exponential smoothing then exponentially smoothed forecasts can be corrected somewhat by adding linear trend
To develop a smoothing procedure that will separate the trend component from the noise in the series and forecast
trended data without a lag is called Double Exponential Smoothing (DES). Given a smoothing coefficient of a, a simple
smoothed average of the data is first calculated. This series would follow the slope of the original data while smoothing
out some of the noise. A second series is then formed by smoothing the second series will also tend to capture the slope
of the original data while further smoothing the noise. An upward and downward trend in data collection over a
International Journal of Business and Management May, 2009
sequence of time periods causes the exponential forecast to always lag behind (be above or below) that actual
occurrence. If such a trend is observed in double exponential smoothing then exponentially smoothed forecasts can be
corrected somewhat by adding a new adjustment (DESLT). A quantitative forecasting method (AES) in which averages
derived from historical data are smoothed by a coefficient, which is allowed to fluctuate with time in relation to changes
in demand pattern. The larger the coefficient, the greater the smoothing effect.
One way to deal with trended demand data is to fit the historical data to a linear model with an "ordinary least squares"
regression (LR). This procedure is an attempt to decompose the demand data observations into an initial level, a trend
component, and noise components, which are modeled as the errors in the regression estimates. Once established, the
model can be used for several periods, or it could be updated and re-estimated as each new data point is observed. It
would often be the case that items in a logistics system exhibit demand patterns that include both trend and seasonality.
It is possible to combine the logic of Holt’s procedure for trended data and the seasonal index approach so as to forecast
level, trend, and seasonality. This approach is embodied in Winter’s Model for Trended/Seasonal Data (HWAA). Each
component term of the forecast is estimated with exponential smoothing, and separate smoothing coefficients.
Out of many forecasting models discussed above, no single model is appropriate to forecast the demand of different
products in the market. Best suitable of these techniques need to be selected for each individual product and the raw
materials used for the respective product. In this paper, a particular type of yarn, maral combat, has been picked up to
forecast its future consumption by using all the above mentioned techniques. The actual consumption and forecasted
data for each of the techniques are shown in Table 1.
Few criteria have been chosen to select the most suitable technique for the particular yarn. Values of Cumulative
forecast error (CFE), Mean absolute deviation (MAD), Mean square error (MSE) and Mean absolute percent error
(MAPE) and Tracking signal (TS) are shown in Table 2 that are being used to select the best model to suit the material.
From the table, it is obvious that MAD, MSE and MAPE values are the minimum for the Adaptive Exponential
Smoothing model. In addition, CFE and TS values are also considerably lower for Adaptive Exponential Smoothing.
Thus, it can be ascertained that, for the chosen yarn, maral combat, Adaptive Exponential Smoothing is the most
suitable forecasting technique to be used. Figure 1 shows the forecasting trend of few techniques those give better result
in forecasting the demand of the selected material. The trends also support the Adaptive Exponential Technique among
all the models.
3. Supplier Selection
Since a decision maker bases judgments on the knowledge and experience, then makes decisions accordingly, the AHP
approach agrees well with the behavior of the decision maker. The strength of this approach is that it organizes tangible
and intangible factors in a systematic way, and provides a structure yet relatively simple solution to the decision making
Decision making process needs to consider multiple criteria, which are often qualitative and conflicting as well in
nature. This requires multi-criteria evaluation using Analytical Hierarchy Process (AHP) technique developed by Satty
. Analytical Hierarchy Process (AHP) presents a different approach for the situations in which ideas, feelings &
emotions are quantified to provide a numeric scale for prioritizing decision alternatives. Figure 2 shows the process
flow of AHP technique for the supplier evaluation.
The crux of AHP is the determination of the relative weights to rank the decision alternatives. Assuming that there are n
criteria at a given hierarchy, the procedure establishes a
decision maker’s judgment of the relative importance of the different criteria. The numerical results of attributes are
presented to the decision maker(s) to assign relative importance according to a predefined scale. A judgment matrix is
then prepared to evaluate the criteria and the suppliers. Normalized weights of each of the criteria and suppliers have
been calculated using equation (1).
pair-wise comparison matrix, A, that reflects the
a1= priority of criterion/supplier 1 to criterion/supplier n
= priority of criterion/supplier n to criterion/supplier 1
n xx ?
= overall priority vectors of the selected criteria/suppliers
1 13 12
Vol. 4, No. 5 International Journal of Business and Management
In supplier qualification evaluation for the mentioned yarn, maral combat, three different suppliers have been
short-listed to be evaluated based on 11 different criteria. The chosen criteria are: quality, quantity, versatility, lead time,
cost, reputation, experience, relationship, transportation, payment flexibility & bureaucracy advantage. The decision
maker has decided the priority to be assigned to each criterion compared to the others. In Table 3, the values of the
top-most row indicate which criterion is given how much preference compared to the each of the other criteria. The
remaining cells are automatically calculated from the cells of the top-most row using a simulation model that confirms
the consistency of priority matrix.
From Table 3, a normalized matrix has been calculated to find out the average priority for each of the criteria. This
calculated priority factors is shown in Figure 3.
After prioritization of each of the criteria taken into consideration, all selected suppliers are then prioritized based on
each category. For example, supplier A is 0.5 times preferred to supplier B and 3 times preferred to supplier C with
respect to quality. Similar preference matrices for different suppliers with respect to few selected criteria are shown in
Finally combining the priority matrices of both criteria and suppliers for each criterion, an overall priority matrix has
been generated using the mathematical model of Analytical hierarchy Process (AHP). Table 5 represents these final
overall priority values of each pre-qualifying supplier for our case study.
Twelve different criteria and three alternative pre-qualified suppliers have been considered in this study. Acceptance is
checked for each priority matrix. Acceptability of alternative and attribute is measured in terms of consistency ratio
(C.R.) which is the ratio between consistency index (C.R.) and randomly generated consistency index (R.I.). Here both
qualitative and quantitative criteria are considered. The qualitative criteria are judged by expert opinion and quantitative
criteria are judged against the collected and calculated quantitative data. By analyzing, the overall priority values are
calculated for different suppliers. Supplier B has the highest overall priority value, then supplier A and supplier C
respectively. So, Supplier B should be selected from three different qualified suppliers. In figure 4 below, overall
priority values of the pre-qualifying suppliers for our case study is represented graphically. It is associated with its
quantitative portion for its easy understanding. From the diagram and overall priority table (Table 5), we can easily find
that supplier B should be selected due to its highest overall priority.
A detailed analysis has been done to customize the best forecasting model for a selected raw material used in the
concerned factory. The obtained result shows that the adaptive exponential smoothing method can forecast the future
demand of the particular raw material very precisely. Similarly, demand of any other raw material used in the factory
can be forecasted using the most suitable out of different forecasting models through similar analysis.
Later, Analytical Hierarchy process (AHP) has been implemented to select the best supplier based on few important
criteria. The same raw material that has been selected for demand forecasting was chosen to serve the purpose.
Combining the priority factors of each criterion over the other as well as of each supplier over another, the best supplier
has been selected for the concerned raw material. Through converting subjective judgment into quantitative form, AHP
provides a better solution for selecting the best suitable supplier with less effort.
Armstrong, J. S. & Grohman, M. C. (1972). A comparative study of methods for long range market forecasting,
Management Science, 19(2), 211-221.
Boucher, T. O. & McStravic, E. L. (1991). Multi-attribute Evaluation within a Present Value Framework and its
Relation to the Analytic Hierarchy Process. The Engineering Economist, 37, 55-71,
Cambron, K. E. & Evans, G. W. (1991). Layout Design Using the Analytic Hierarchy Process. Computers and Industrial
Engineers, 20, 221-229.
Dalrymple, D. J. (1987). Sales Forecasting Practices results from a United States Survey. International Journal of
Forecasting, 3, 379-391.
Fildes, R. & Hastings, R. (1994). The Organization and improvement of Market Forecasting. Journal of the Operational
Research Society, 45(1), 1-16,
Finnie, G. R., Witting, G. E., & Petkov, D. I. (1993). Prioritizing Software Development Productivity Factors Using the
Analytic Hierarchy Process. System Software, 22(2), 129-139.
Min, H. (1994). Location Analysis of International Consolidation Terminals Using the Analytic Hierarchy Process.
Journal of Business Logistics, 15(2), 25-44.
Mustafa, M.A. & Al-bahar, J. F. (1991). Project Risk Assessment Using the Analytic Hierarchy Process. IEEE
Transactions on Engineering Management, 38(1), 46-52.
International Journal of Business and Management May, 2009
Putrus, P. (1990). Accounting for Intangibles in Integrated Manufacturing. Information Strategy, 6, 25-30.
Saaty, T. L. (1977). A Scaling Method for Priorities in Hierarchical Structures. Journal of Mathematical Psychology, 15,
Shtub, A. & Dar-el, E. M. A. (1989). Methodology for the Selection of Assembly Systems. International Journal of
Production Research, 27(1), 175-186.
Wabalickis, R. N. (1988) Justification of FMS with the Analytic Hierarchy Process. Journal of Manufacturing Systems,
Wang, L. & Raz, T. (1991). Analytic Hierarchy Process Based on Data Flow Problem. Computers and Industrial
Engineers, Vol. 20, pp. 355-365.
Winklehofer, H., Diamantopoulos, A. & Witt, S. F. (1996). Forecasting Practice: A Review of the Empirical Literature
and an Agenda for the Future Research. International Journal of Forecasting, 12(2), 193-221.
Table 1. Actual consumption (tons) and forecasted data (tons) in 12 month period
1 2 3 4 5 6 7 8 9 10 11 12
SA 20 21 20.5 20.7 21 21.822.824.6 25.6 26 26 25.5
MA 20 19.520 20.6721 22.6725 29.3332 32.33 29.3325
WMA 20.5 20 22 21 21.2 22.8 25.229.631.8 31.3 29.2 24.9
SES 20.3 21 20.85 20.8721.0421.64 22.5924.4525.73 26.22 26.19 25.26
SESLT 19.9 21 20.81 20.8 20.9921.7923.2025.9828.36 29.81 30.42 29.57
DES 20 21 20.98 20.96 20.9721.0721.3 21.7722.37 22.95 23.4323.71
DESLT 20.5 21 20.7 20.7721.12 22.324.1127.6 29.69 30.08 29.4327.09
AES 20.7 21 20.85 21.9224.73 27.16 24.9934.3 28.19 26.92 25.12 19.67
LR 23.27 25.58 23.9 24.2124.5324.84 25.16 25.4725.79 26.1 26.42 26.73
HWAA 20.8 22 20.9 22.3 21 20.7823.37 27.1430.01 31.48 34.5331.24
Actual 21 20 21 22 25 28 35 33 29 26 20 20
Table 2. Forecasting errors using different techniques
Model CFE MAD MSE MAPE TS
SA 23.54902 4.405283 32.15264 16.43619 5.345633
MA 0.6666679 5.333333 35.65432 20.79569 0.1250002
WMA -7.63E-06 5.55 36.7075 21.57678 -1.37E-06
SES 23.14943 4.409149 32.93404 16.37703 5.250317
SESLT 6.2733 5.0779 41.8572 20.1208 1.2354
DES 38.488 4.9788 41.712 17.6498 7.7304
DESLT 5.104 4.5186 32.2168 17.8278 1.1296
AES 4.1353 2.6562 10.0642 10.4376 1.5568
LR 0 4.0355 24.3199 16.1576 0
HWAA -3.5473 7.6212 75.7807 30.7724 -0.4655