STOCKS MANAGEMENT THROUGH
APPLICATION OF
DEMAND FORECAST
METHODS: A CASE STUDY
Lucas Lopes Filholino Rodrigues
Fatec Guarulhos, Brasil
E-mail: lucasfilholino@hotmail.com
Igor Henrique Inácio de Oliveira
Fatec Guarulhos, Brasil
E-mail: igoroliveira06@outlook.com
Maurílio Fagundes Alexandre
Fatec Guarulhos, Brasil
E-mail: maurilio.alexandre@yahoo.com.br
Rodrigo Rodrigues Castorani
Fatec Guarulhos, Brasil
E-mail: rodrigocastorani@hotmail.com
Celso Jacubavicius
Fatec Guarulhos, Brasil
E-mail: jacubavicius@uol.com.br
Submission: 31/03/2016
Accept: 31/03/2016
ABSTRACT
The
present study consists in assessing the feasibility of implementing demand
forecasting techniques due to the optimization of inventory management, so that
it is objective the reduce storage costs and to have the least amount of
stationary material stock in a certain period. Data analysis was for
application of techniques based on the real case of a multinational company in
the segment of electronic and digital systems in the infrastructure area, which
operates in the metropolitan region of São Paulo.
The
study aims to evaluate the behavior of the studied company demand, in order to
demonstrate some forecast models for it and thus being able to identify the
most appropriate method to get the highest possible degree of assertiveness.
At a
first moment, there will be a survey of data concerning the company's
historical demand, including the forecast used at the latest period, and then
to survey the state of the art discussed topic, in order to clarify the reader,
and as a result: the analysis of the collected data and the implementation of
demand forecasting techniques presented in bibliographic references.
After
performing an analysis of the naive method demand forecasting practiced by the
company, was carried out the application of different forecasting methods and
found out that the method that best suits the given demand was the moving
average, which provided the optimization of cost of storage in approximately
63% of the one presented by the naive method and also a gain of approximately R
$ 2,000,000.00 during the studied period, thus proving the effectiveness of
demand forecasting application for inventory management.
Keywords: Forecast; Demand; Management;
Inventory; Optimization.
1. INTRODUCTION
Currently is possible to see that the set of
uncertainties in a business segment is proportionally linked to the chances of
obtaining negative results, both financially as to the level of service. This
is a factor that is evident even to organizations working in the design and
production of consumer goods. Due to these aspects, demand forecasting
techniques are of fundamental importance to reduce that risk in order to
achieve positive results and profits.
The impacts caused by the development of a demand
forecasting are crucial to the survival of businesses: on this account,
mistakes can cause the end of its existence; and accurate results can leverage
profits and ensure its expansion and market gain. Due to this fact that there
are several techniques to achieve the demand forecast and several technological
features that make accurate results, based whenever possible on data analysis
using quantitative methods.
This is a research in which will be presented, on
implementation and evaluation of different quantitative methods of demand
forecasting, for an organization that produces electrical durable goods in the
construction industry, so that the central objective of the study is to
understand the demand behavior and identify the most suitable method to get the
forecast with the highest degree of assertiveness as possible.
ABC Group will be used as the base of this search, so
that its operation happens in approximately 180 countries, where electronic and
digital systems are developed in the building infrastructure area.
2. PROBLEM
The instability of the relation of excess inventory
and the lack of resources is a intrinsically detrimental factor to the finances
of a company, and in respect of a particular product the company studied, it
was realized that there is, in various times, material in excess due to the
need to meet the demand, a situation that leads to inventory build-up resulting
in impacting of costs in the financial statement for the period.
3. JUSTIFICATION
Due to the presented problem, it is of considerable
importance to develop a demand forecast which can stabilize the flow of
materials to greater assertiveness in meeting demand, since for the product in
question, after survey and analysis of the actual demand data and the expected
one, there was excess on average about 5.5% of the material required to meet
demand.
4. HYPOTESIS
The application of a demand forecasting method can
balance the ratio of surplus and shortages, providing to the company the ability
to spend less on stocks without the risk of not meeting the demand.
5. OBJECTIVE
Evaluate the behavior of the demand in the studied
company, in order to demonstrate some forecast models for such and thus being
able to identify which method is the most appropriated one to provide the
highest possible degree of assertiveness.
6. METHODOLOGY
According to Gil (2008), this study is a quantitative
research, once that will be used mathematical data and models in order to get
the results, and is exploratory. Still according to the author, this method may
involve people with some experience in the subject, interviews and literature,
keeping in view the objective of analyzing the existing process without direct
participation in it, where the following steps are performed:
·
Survey of data regarding the company's demand in
concerning to the occurrences studied, including the forecast used by this one
in a given period;
·
Survey of the state of the art on the theme
approached, in order to elucidate the reader on the subject;
·
Description of the methodology used by the company
studied;
·
Demonstration of the relation between the technical
literature and quoted ones;
7. THEORETICAL FOUNDATION
7.1.
Demand Forecasting
Initially the horizon of the demand planning Buffa
& Sarin (1987) classifies that the forecasts can be short-term, probably
related to inventory controls; can be medium-term referring to planning
production; and long-term plans, usually used in the company's strategic level
to plan the growth of the business.
Considering the need for accuracy at the time of
acquisition and manufacturing, in addition to the uncertainties that involve
the demands in general, companies cannot rely on a chance to make possession of
their inputs, materials and goods, since they need to ensure that the amount
acquired is the closest to be used, limiting or preventing faults and
leftovers. Technological advances brought business development techniques,
philosophies and policies focused on increasing the degree of assertiveness on
the management of inventory, transportation and other activities of logistics
competence.
The forecast of demand is one of these techniques, so
that can be defined by the number of activities that estimate the amount of
resources for future needs based on past occurrences (DIAS, 1990), as agreed
Boland (1985) arguing that "forecasting can be understood as a projection
or extrapolation of past trends.”.
Well elaborate demand forecasts can provide to a
company successful integration of production processes, distribution and
inventory management with lower costs well as greater objectivity and front
flexibilities to extra searches. Diaz & Saucer (2003) say that the forecast
demand is a critical step for all players of a supply chain because of the
complexity and uncertainty of its activities, in addition to Makridakis (1988),
noting that it is through demand forecasting that companies make strategic
decisions, structure plans or take any action that relates future events.
Concerning the methodology, the forecast demands may
occur in qualitative ways considering expert opinions, capable of critical
analysis of a history of occurrences, or quantitative, that are more accurate
and based on mathematical and statistical systems, but without sensitivity to
differentiation of demand scenarios.
There are also researchers trying to combine the two
methodologies, developing a model based on historical data and then quantifying
this data, generating mathematical analyzes that are more accurate than any
individual forecast (MOREIRA, 2009). The next topics will demonstrate some of
the leading demand forecasting methods, which are used in this study.
7.2.
Linear Regression
According to Neufeld (2003), the linear regression
analyzes the relation between two variables. The correct data for this method
consists in observations with different measures each. So has been an
independent variable (variable x) and a dependent one (variable y), considering
that the first variable causes changes in the second. This method is one of the
most used ones. The authors Krajewski, Ritzman and Malhotra (2009) complemented
this claim and pointed out to the possibility of the trend line training that
enables statements within the studied horizon.
The application of the method is given by the formula
demonstrated below:
7.3.
Simple Average
This method although simple has certain degree of
assertiveness. He forecasts future demand as the average of the demands from
previous periods. This type of technique is not suitable for series that
present trend and / or seasonality and should only be used to forecast a period
ahead (WANKE; JULIANELLI, 2006). Because the old values have the same weight
the new values obtained in the course of time, this method does not respond
quickly to changes in the level of the sample.
Below is demonstrated the formula of this method:
7.4.
Simple Moving Average
Simple Moving Average (SMA) has strength in continuous
demand variations and therefore its disability is in cases of intermittent
demands. (EAVES; KINGSMAN, 2004; PORRAS; DEKKER, 2008; TEUNTER; DUNCAN, 2009).
This method is recommended for short-term forecasts where the components of the
trend and seasonality can be ignored or non-existent. (MAKRIDAKIS; WHEEL WRIGHT;
HYNDMA, 1998).
This technique is not indicated to the series having
seasonality or trend, because the forecast tend to become outdated due to this
method, when used in fewer periods, shows higher impact by including new data,
and the weights assigned to these are the same for all sample data.
This is the most used model by companies in general, because
of the simplicity and the need for a reduced history data, analyzing a very
short period of the process so that the number of observations in each average
calculation remains constant and is stipulated in a way that tries to eliminate
as best as we can the trend and seasonality components. "(CHAMBERS;
MULLICK; SMITH, 1971; ARCHER, 1980; MAKRIDAKIS; WHEELWRIGHT; HYNDMAN, 1998).
Below is demonstrated the formula of this method:
7.5.
Double Moving Average
According to the authors Wanke and Julianelli (2006),
the dual moving average method can be considered more efficient when compared
to the simple moving average, by adhering to samples with a tendency. However,
both methods assume that the recent values are just as important as the
earliest to define the future value, a fact that contradicts practical
situations. In this case, the simple moving average method is applied twice,
once in the original data and then the data resulting from this first
application.
Below is demonstrated the formula of this method:
7.6.
Method Of Simples Exponential Damping
According to Corrêa and Corrêa (2006), the Simple
Exponential method is the weighted average of past data, so that the weight
decreases according to the antique of data: the older the value, less is its impact
in the calculation.
This is a different method than the others explained,
because here is considered a differentiation among the values in each period
analyzed. Thus, through the Damping Coefficient is possible to assign light or
heavier weights to the most recent values, but as recent is the occurrence the
bigger will be its impact in the calculation. (SOUZA; SAMOHYL; MIRANDA, 2008).
The Mathematical formula for this method is
demonstrated below:
7.7.
Holt Method Or Double Exponential Damping
According to Gaither and Frazier (2001), this method
is indicated to be used with series and tendencies, possessing two Damping
coefficients utilized for tendency and level sample. The first coefficient is
represented by α and is referent to the level, and the second is represented by
β and refers to the series of trend.
The Mathematical formula for this method is
demonstrated below:
8. CASE STUDY
The ABC Group is one of the largest companies in the
electronic and digital systems segment in the infrastructure area, owns several
brands with more than four thousand patents and one hundred seventy thousand
items, is present in over one hundred and eighty countries. It is emphasized
that due to the high flow material handling, it can be said that inventory management
is crucial to the organization of products and control of entrance and exit of
goods, thus avoiding losses and obtaining the profits of such stock.
8.1.
Case
The product chosen for study holds about 20% of the
total turnover of the company, however, comparing with other items developed by
the company, this one has a low value at the time of sale, which refers to R $
2.89, besides high costs of storage due
to volume, representing the value of R $ 1.20 per piece within one month. By
fidelity issues, the company adopts policies that force to meet all periodic
supply for this product, so in situations of lack of products it is necessary
to provide extra production of the items, a process that represents additional
cost of 15% of the original value of the production cost that is, each unit
that need to be added to the amount provided will have the cost of R $ 1.02.
The company adopted a simple quantitative method, so
that to anticipate the demands of the next period, the management was based on
the same period last year, but with the adding 10% of the total value of items
sold. It will be shown below the mathematical model used by the company to make
such a prediction. This method is defined as naive method and is characterized
by simple use of historical data as happened in the previous period or simply a
correction factor for the period. (BARBIERI; MACHILINE, 2009).
From the table below will be shown the actual demand
of the period to which it tried find, besides the forecast made by the company
based on the previous calculation and the error rate caused by the prediction.
Table 1: Average error obtained
by the naive method
Real |
Forecast |
Error (%) |
|
550000 |
585000 |
5,98 |
|
430000 |
450000 |
4,44 |
|
390000 |
430000 |
9,3 |
|
400000 |
450000 |
11,11 |
|
330000 |
355000 |
7,04 |
|
300000 |
310000 |
3,23 |
|
310000 |
340000 |
8,82 |
|
400000 |
428000 |
6,54 |
|
420000 |
500000 |
16 |
|
425000 |
400000 |
-6,25 |
|
450000 |
460000 |
2,17 |
|
400000 |
390000 |
-2,56 |
|
Média |
5,49 |
As noted, in
one of the periods there was no excess material needed, a situation which
consequently limited the company about the efficiency and accuracy of its
stock, in addition to providing impactful cost of such storage. The costs of
the previous forecast errors are listed in the table below:
Table 2: Total Cost storage
obtained
Real |
Forecast |
Accumulated
Stock |
Storage
Cost |
550000 |
585000 |
35000 |
R$
42.000,00 |
430000 |
450000 |
55000 |
R$
66.000,00 |
390000 |
430000 |
95000 |
R$
114.000,00 |
400000 |
450000 |
145000 |
R$
174.000,00 |
330000 |
355000 |
170000 |
R$
204.000,00 |
300000 |
310000 |
180000 |
R$
216.000,00 |
310000 |
340000 |
210000 |
R$ 252.000,00 |
400000 |
428000 |
238000 |
R$
285.600,00 |
420000 |
500000 |
318000 |
R$
381.600,00 |
425000 |
400000 |
293000 |
R$
351.600,00 |
450000 |
460000 |
303000 |
R$
363.600,00 |
400000 |
390000 |
293000 |
R$
351.600,00 |
Total |
R$ 2.802.000,00 |
As a result, in
order to have a better observation of the problem was developed a graphic
representation with the relation between the demand and the forecast made by
the company.
Figure
1: Graphical representation of the relationship between forecast and actual
demand.
After analyzing the results obtained by the company in
total period observed, it’s possible to see that only due to the high volume
stored during the year there was a total cost of R $ 2,802,000.00, that is
a significantly value, since the amount is about 40% of all
profits obtained in the period on this product.
8.2.
Application
As explained before, the company needs a new demand
forecasting technique able to reduce the alarming costs to better economic
performance. Thus, the aforementioned techniques were applied in order to find
a better solution to the problem.
The application of such techniques took place in
accordance with the formulas shown above, so that how to get the results of
storage costs and their impacts, is the same that tab of the presentation of
the method used by the company, as Table 3.
After the application of the concept and the
mathematical techniques for each method of forecast, was possible to develop
the below table, with the results and others data about the operation.
Table 3: General demonstrative
of results.
The
figure shows the recipe (R) consisting of the period billing amount; the Total
Storage Cost (TAC) that indicates the reason of unitary cost to storage for all
goods that must be kept in inventory in function of the determined forecast;
the Cost Total production (CPT), established by multiplying the suggested
volume in the given forecast, for the unit cost of production, besides the
total profit of the period (LPT) that quantifies the value obtained in the
difference between revenue and costs presented for each of the related
forecasts; and, finally, the average error of each forecast (EM). It is
noteworthy that, after application of the linear regression method, it was
noted an average of 16.9% error, a result that makes it completely not workable
to use this technique to the problem in question. In this way, this one was
removed of the demonstration of analysis.
Analyzing the results, it is observed that the
techniques of simple average and simple exponential represent results below
what was presented by the naive technique, put into practice by the studied
company, in such a way representing respectively 68% and 89% of the total
profit obtainable from such applications, conditions that make their use
impractical.
As for the Holt technique, there is a small positive
change in the result as compared to naive technique. In percentage, was
obtained approximately 2% gain in total income for the period, but also
presents considerable amounts of products that would be stopped in stock, so
that the storage value represents approximately 41% of profit.
Considering the proposed methods, the one that
presents the best solution for forecasting is the method of moving average,
because besides presenting an average error in
just 1.47%, reduced by 63% storage costs, when compared with the
technique used by the company and thus increased profit by approximately R $
2,000,000.00. Participation of storage costs in the profit obtainable from the
application of this technique corresponds to only about 12%, is significantly
lower than those achieved by the studied company.
In figure 3, one can see graphically that the method
of moving average has greater adhesion to the company's demand, since the
correlation between these figures is shown with a high degree of similarity,
thus demonstrating that with the fall of the error forecast, more assertive
will be the method. The reduction of about 4% in the average for in this demand
forecasting method boosted the company's profit, without any change in the
price of the product, so it may be said that the organization's procurement
process has become more efficient.
Figure
2: Graphic representation of forecast x moving average
9. FINAL CONSIDERATIONS
At the end of the survey, was possible to note that
the company had high levels of inventory, which represents an important problem
at its financial health. The company still had used inefficient methods to
promote the demand forecasting, supply chain and to guide the production, which
explain the existence of a lack in the supply needs and the supply storage.
There are techniques to forecasting the warehouse
demand, with proved efficacy and knew by the scientific community. Several
techniques were applied to the real data, and the results were analyzed to best
fit the supply chain itself.
As shown in this survey, was
possible to indicate that the moving average was the best-fitted method to
forecast the needed level of storage, which decreased the levels of the error
compared with the naive technique. The usage of the moving average was able to
reduce the cost of storage in 63% and increase the profit in R$ 2.000.000,00,
when compared with the naïve method at the time of the study.
Finally, can be emphasized
that the choice of a technic as the moving average does not eliminate the
possible choice and use of another forecast technic, however, in this case, the
moving average technic demonstrates that it is adequate to fit the company
demand and be a good solution to increase the company profit.
REFERENCES
BARBIERI, J.
C.; MACHLINE, C. (2009) Logística Hospitalar: teoria e prática. 2.ed.
rev. e atual. São Paulo:
Saraiva.
BOLAND, J. J. (1985) Forecasting water use:
A tutorial. Computer
Application sinwater Resouces.
Edited by H. C. Torno,
Buffalo,
Nova
York,
p. 907-916.
BUFFA, E.
S.; SARIN, R. K. (1987) Modern production/operations management.
8. ed. Los Angeles: John
Wiley & Sons.
CORRÊA, H. L.;
CORRÊA, C. A. (2006) Administração da
Produção
e Operações. São Paulo: Atlas.
DEKKER,
R.; VAN DONSELAAR, K.; OUWEHAND,
P. (2004) How to use aggregation
and combined
forecasting
to improve seasonal demand
forecasting. International
Journalof
Production
Economics, v. 90, p.151-167.
DIAS, M. A. P. (1990) Administração de
materiais: uma abordagem logística.
São Paulo: Atlas.
DIAZ, C. A. P.;
PIRES,
S. I. R. (2003) Variação da
Demanda ao Longo da Cadeia
de Suprimentos:
O Efeito da Amplificação da
Demanda, XXIII
Encontro Nacional
de Engenharia
de Produção. Ouro Preto/MG.
EAVES, A.; KINGSMAN, B.
(2004) Forecasting for
the ordering and
stock-holding
old spare
parts. Journal of
the Operation Research Society, v.
50, p.431-437.
GIL, A. C. (2008) Como
elaborar
projetos de
pesquisa.
4. ed.
São Paulo: Atlas.
KRAJEWSKI,
L. J.;
RITZMAN, L.
P.; MALHOTRA, M.
(2009) Administração da
produção e operações.
8. ed.
São Paulo: Pearson
Prentice Hall.
MAKRIDAKIS,
S. (1988) Meta forecasting:
Waysof Improving
Forecasting Accuracy
and Usefulness.
International Journal
of Forecasting,
v. 4,
1988, p.467-491.
MAKRIDAKIS, S.; WHEELWRIGHT, S. C.;
HYNDMAN,
R. J. (1998) Forecasting: methods and
applications. 3. ed.
New York, USA: John Wiley
& Sons, p. 641.
NEUFELD, J. L.
(2003) Estatística Aplicada à Administração usando
Excel.
1ª ed., São Paulo:
Prantice Hall.
PORRAS,
E.; DEKKERR. (2008) An Inventory
control system
for
spare parts at a refinery:
an empirical comparison
of different re-order
point
methods.
European
Journal of
Operational
Research, v. 184,
p. 101-132.
REID, R. D.; SANDERS,
N. R. (2005) Gestão de operações. Rio
de Janeiro:
LTC.
SAMOHYL, R. W.; SOUZA,
G. P.; MIRANDA, R. G.(2008) Métodos simplificados de previsão
empresarial. Rio
de Janeiro:
Editora
Ciência
Moderna
Ltda.
WANKE, P.; JULIANELLI,
L. (2006) Previsão de
vendas: processos organizacionais e métodos quantitativos
e qualitativos. São Paulo:
Atlas.