1. INTRODUCTION

All issues of a supply chain planningsystem start from demand forecasting whichserves as the basis of every planning activity

(Chen & Blue, 2010). Fildes et al. (2009)stated that demand forecasting is the crucialaspect of a planning process in supply-chaincompanies. Sales forecasting forms the basisof all supply chain planning activities

FOCUS FORECASTING IN SUPPLY CHAIN: THE CASE STUDY OF

FAST MOVING CONSUMER GOODS COMPANY IN SERBIA

Zoran Rakićević* and Mirko Vujošević

Faculty of Organisational Sciences, University of Belgrade,Jove Ilica 154, 11000 Belgrade, Serbia

(Received 05 November 2014; accepted 14 January 2015)

Abstract

This paper presents an application of focus forecasting in a fast moving consumer goods (FMCG)supply chain. Focus forecasting is tested in a real business case in a Serbian enterprise. The data usedin the simulation refers to the historical sales of two types of FMCG with several different products.The data were collected and summarized across the whole distribution channel in the Serbian marketfrom January 2012 to December 2013. We applied several well-known time series forecastingmodels using the focus forecasting approach, where for the future time period we used the methodwhich had the best performances in the past. The focus forecasting approach mixes different standardforecasting methods on the data sets in order to find the one that was the most accurate during thepast period. The accuracy of forecasting methods is defined through different measures of errors. Inthis paper we implemented the following forecasting models in Microsoft Excel: last period, allaverage, moving average, exponential smoothing with constant and variable parameter α,exponential smoothing with trend, exponential smoothing with trend and seasonality. The mainpurpose was not to evaluate different forecasting methods but to show a practical application of thefocus forecasting approach in a real business case.

Keywords: focus forecasting, moving average, exponential smoothing, Holt's model, Winter’s model,fast moving consumer goods (FMCG)

* Corresponding author: zoran.rakicevic@fon.bg.ac.rs

S e r b i a n

J o u r n a l

o f

M a n a g e m e n t

Serbian Journal of Management 10 (1) (2015) 3 - 17

www.sjm06.com

DOI:10.5937/sjm10-7075

(Chopra & Meindl, 2007). The first step thatmanagers must take is to forecast whatcustomer demand will be. With adequateanticipation of sales, managers can plan thelevel of activities: production, capacity,inventory, transportation, distribution.Information on demand is one of the mostimportant parts in the whole supply chainplanning (Chen & Blue, 2010) and adequatesales forecast can prevent a bullwhip effect(Ramanathan & Muyldermans, 2010;Oyatoye & Fabson, 2011). It can beconcluded that sale forecasting is: (1) anactivity of Supply Chain Management-SCM(Warren Liao & Chang, 2010), (2) the mainplanning problem in SCM (Zamarripa et al.,2012) and (3) the key success factor of theSCM (Thomassey, 2010). In this paper weapplied several time-series forecastingmethods through the focus forecastingsystem. The idea was to point out theimportance of successful forecasting systemin supply chains and to indicate the need fordevelopment of adequate forecasting systemwith plenty of forecasting methods andforecast accuracy indicators.

This paper is organized as follows:Section 2 defines the term of forecasting andpresents a group of forecasting methods thatare based on time series. Section 3 presentsand defines focus forecasting approach withmeasures of forecasting accuracy i.e. error,which is important element of this concept.Section 4 presents a case study on focusforecasting applied on sales of two types ofFMCG; time-series forecasting methods areused on real data from practice andcompared using different measures offorecasting errors. Section 4 also presentsmain results and discussion. Section 5concludes the paper and discusses futureresearch.

2. FORECASTING METHODS

Forecasting is a process of buildingassumptions and estimates about futureevents that are generally unknown anduncertain (Vujošević, 1997). Forecasting isthe art and science of predicting futureevents. It implies taking historical data andprojecting them into the future, usingmathematical models and methods orintuition. Forecasts are used for manypurposes: marketing, sales,finance/accounting, production/purchasing,and logistics (Kerkkänen et al. 2009).Numerous factors are related to the salesforecast (Chopra & Meindl, 2007): pastsales, product lead time, planned advertisingor marketing efforts, state of economy,planned price discounts, competitors’actions.

Generally, forecasting methods areclassified into two groups: quantitativemethods and qualitative methods (Chopra &Meindl, 2007; Vujošević, 1997; Heizer &Render, 2011). Qualitative forecastingmethods are based on experts’ estimates andjudgements as well as their experience.Quantitative forecasting methods usehistorical statistics and appropriatemathematical models to make futureprediction. The most common categorisationof quantitative methods draws a distinctionbetween projective and causal methods(Vujošević, 1997) i.e. time-series methodsand associative methods (Heizer & Render,2011). Projective methods (time seriesmethods) try to find rules in the data, wherethe forecast is a “picture” of historyprojected in future. Causal (associative)methods try to find and make causalrelationship between variables (Vujošević,1997).

4 Z.Rakićević / SJM 10 (1) (2015) 3 - 17

Time-series forecasting methods areprobably the most used techniques forprediction of sales data within a supply chain(Thomassey, 2010; Davydenko & Fildes,2013). Time series is a collection of data thatdescribes the values of variables during theobserved period of time. Some of thestatistical techniques for forecasting basedon time series are (Vujošević, 1997; Chopra& Meindl, 2007; Heizer & Render, 2011):last period estimation, arithmetic mean,moving average, weighted moving average,exponential smoothing, Holt’s model(exponential smoothing with trends),Winter’s model (exponential smoothing withtrend and seasonality), regression analysis.

We have presented here several most usedtime-series forecasting methods, with thefollowing notations of variables andconstants:

Dt - demand in period t;Ft - forecast for period t;Lt – level of forecast in period t; α – smoothing parameter for forecast

demand level, 0≤α≤1;β – smoothing parameter for trend,

0≤β≤1;γ - smoothing parameter for seasonality,

0≤γ≤1;Tt –exponential smoothed trend;St - exponential smoothed seasonality;FITt – exponential smoothing forecast

that includes trends in period t;FITSt - exponential smoothing forecast

that includes trend and seasonality in periodt;

m – number of observed periods w;ωi – weighted coefficient, ωi > ωi-1, ωi >

0, ωi < 1;Et – forecast error in period t;At – absolute forecast error in period t.

Sales forecasting methods:

Last period estimation is a simpleforecasting method (strategy) that assumesthat the demand in the next period will beequal to the demand in the most recent pastperiod. This method is applied when thevariations in actual values of data are small,from period to period:

(1)

Arithmetical average is a method forforecasting next period that is trying to makeaverage of all historical sales data.Arithmetical average ignores trend andseasonality patterns.

(2)

Moving average is a forecasting methodthat uses the average of several recent salesdata for the next period sales forecast. Thismethod follows trend patterns with a certaintime delay, but seasonal patterns cannot befollowed.

(3)

If there is clear trend patterns, movingaverage can be upgraded with weightedcoefficients. This method refers to weighted

moving average:

(4)

Increasing the size of an observed periodin history, this method better stabilizesfluctuations and becomes less sensitive toreal changes in the data. The trends alwaysremain in the shadow of its averageness.

Exponential smoothing is a widespreadforecasting time series method within supply

5Z.Rakićević / SJM 10 (1) (2015) 3 - 17

chain models (Ferbar et al., 2009).Exponential smoothing is a kind of weightedmoving average, where the weights areassigned using exponential function –forecast for the next period is equal to theforecast from one period earlier corrected bythe weighted difference between a realdemand and forecast in an earlier period.

(5)

The smoothing constant regulates theinfluence of historical values (Wallström &Segerstedt, 2010). A lower value of constantα gives bigger significance to historicalvalues of data. Higher value of constant αhas an opposite effect. Recommended valuesfor α constant, in literature, are between 0.1and 0.3 (Vujošević, 1997). A basic model ofexponential smoothing in long term forecastsdoes not take trend patterns into account.However, an exponential smoothing methodcan be modified for forecasting time serieswith a trend component.

Trend – corrected exponential

smoothing, Holt’s model, is a morecomplex model of exponential smoothingthat calculates an exponential smoothingforecast and then corrects it with a positiveor negative average value of a moving trend.

- Exponentially smoothed forecast ofsales level:

(6)

- Exponentially smoothed trend:

(7)

- Exponentially smoothed forecast thatincludes trend:

(8)

The forecast for the next period (FITt)represents a sum of exponential smoothingforecast (Lt) for the next period and acalculated trend for the next period (Tt).When the value of β parameter is high,forecasts are sensitive on recent changing intrend. Conversely, when the value ofparameter β is low, a forecast is less sensitiveto a recent value and emphasis the pasthistorical values. The value of β can beobtained testing different values andmeasuring forecast error, with the help ofoptimizing software such as MicrosoftExcel.

Trend and seasonality corrected

exponential smoothing, Winter’s model -Seasonal variations are regular changes inthe time series, up or down, related to theevents that are repeated (e.g. summer orwinter season, holiday season). A seasonalcharacteristic of variables can be identifiedbased on increased or decreased values ofdemand in a certain period of year. For thispurpose the seasonal indexes that represent aratio between demands in certain periods andthe average demand for a whole series areused.

Exponential smoothing method thatincludes trend and seasonality is:

(9)

- Exponentially smoothed forecast ofsales level:

(10)

- Exponentially smoothed trend:

(11)

6 Z.Rakićević / SJM 10 (1) (2015) 3 - 17

- Exponentially smoothed seasonality:

(12)

According to Thomassey (2010), theseabove presented statistical time seriesforecasting methods provide satisfactoryresults in real-life applications. Allmentioned forecasting methods havedifferent accuracy and success in salesforecasting. Chang and Lin (2010) stated: “itis very important that business andmanufacturing enterprises build flexible androbust supply chain forecasting systems thatallow them to reduce the negative impacts ofbullwhip effect”. Forecasting accuracy is acritical element of supply chainmanagement, but, at the same time, it is thearea that continues to challenge mostcompanies (Vayvay et al., 2012).

3. FOCUS FORECASTING

According to Min & Yu (2008) focusforecasting, is an expert system thatidentifies a simple forecasting rule-of-thumbmethod that worked best in the past, and usesit to make a short-term prediction for futureevents such as sales or customer demands.Focus forecasting often includes three steps:(1) simulate the past forecasts using a varietyof forecasting rules and methods; (2)evaluate the performances of theseforecasting methods with respect toforecasting errors; (3) select the forecastingmethod that performed best and apply it inthe next period’s demand. After realisingsales in the next period, the sales data in thedatabase are refreshed and the sameprocedure of focus forecasting is realized.This system of forecasting can be applied

involving experts' opinions (Figure 1).Focus forecasting allows application of

different forecasting time-series methods anduses methods with the best accuracy, i.e.minimum forecast error (Gardner et al.,2001). Chopra and Meindl (2007) stated thatusing multiple forecasting methods to createa combined forecast is more effective thanusing any method alone. Focus forecastingcan also be used for forecasting sales ofdifferent products.

Sales and demand forecasting is never100% accurate (Warren Liao & Chang, 2010;Oyatoye & Fabson, 2010). The mostimportant part in a focus forecasting systemis how to measure forecast accuracy or itssuccess. Typically, it is measured with a

7Z.Rakićević / SJM 10 (1) (2015) 3 - 17

Figure 1. The diagram of focus forecastingconcept, modified source (Vujоšević, 1997)

forecast error. The literature provides severaldifferent measures for forecast errors(Vujošević, 1997; Chopra & Meindl, 2007;Kerkkänen et al. 2009; Wallström &Segerstedt, 2010; Heizer & Render, 2011). Ingeneral, the forecast error is defined as adifference between the forecast and actualsales. Forecasts that are larger/smaller thanthe actual sales are referred as over/under-forecasts. According to Bratu-Simionescu(2013), some of the most popular measuresof forecast accuracy are: Mean AbsoluteDeviation (MAD), Average of absolutedeviation over all periods, Mean AbsolutePercentage Error (MAPE), Mean SquaredError (MSE), Root Mean Square Error(RMSE), Cumulative error, Average error orBias, Tracking Signal (TS). In production orservice industries sales forecasting errorshave significant impacts on total costs,schedule instability, lost capacity,uneconomical use of capacity, excessinventory, inventory holding cost,obsolescence, reduced margin, lost “sales”cost (Kerkkänen et al., 2009).

The forecast error in period t is thedifference between the actual demand inperiod t and the forecast for period t:

(13)

Mean absolute deviation (MAD) is ameasure of total forecast errors for themodel:

(14)

where Ft is sales forecast in period t and Dt issales in period t.

Mean Squared Error (MSE) is a bettermeasure of error than MAD, since squaredfunction gives bigger weights to a bigger

difference between real sales and forecasts.Thus, one big difference that is in absolutesum equal to several small differences is notequal in square function. The MSE can berelated to the variance of the forecastingerror.

(15)

Root mean squared error (RMSE) can beused instead of MSE:

(16)

MAD, MSE and RMSE express errorvolume, but do not point out at errordirection, i.e. over-forecasting or under-forecasting. For this purpose bias is used.Bias is a sum of differences between realsales and forecasts. Bias is also famous as acumulated forecast error (CFE) (Wallström& Segerstedt, 2010). CFE divided with thenumber of forecast periods gives averagebias.

(17)

Ideal forecasting methods have zerovalues for MAD and bias. Positive biasshows tendency of under-forecasts, whilenegative bias shows tendency of over-forecast (Vujošević, 1997).

Mean Absolute Percentage Error (MAPE)has advantage over MAD and MSE becauseits value is not dependent of product’smeasurement units. MAPE is an averageabsolute error, between forecasted valuesand a real sales value, presented as apercentage of demand.

8 Z.Rakićević / SJM 10 (1) (2015) 3 - 17

(18)

Tracking signal (TS) is the ratio betweenthe cumulated forecast error (CFE), i.e. bias,and the mean absolute deviation (MAD).

(19)

Tracking signal shows whether to reviseparameters of forecasted model or not. It is asimple method that checks whether theforecast error is in the limits of tolerance, andis usually between -3% and +3% or -8% and+8%, according to Vujošević (1997).Forecast errors that are within the limits oftolerance are acceptable, but if they are outof the limits, it is necessary to reviseparameters of forecasting models.

Another interesting indicator that can beused for evaluation of several forecastmethods is the Percentage better indicator.Percentage better is the percentage of timeperiods in which a certain forecast methodperformed better than another one(Wallström & Segerstedt, 2010).

In the next section of the paper we appliedthe most usable time-series forecastingmethods on a real forecasting problem inbusiness practice in Serbia.

4. CASE STUDY

Many industries such as Fast MovingConsumer Goods (FMCG) seeking moreaccurate demand forecasting (Sayed et al.2009). Focus on FMCG industry is not thatstrong from academic researchers. The mainreasons are: characteristics of FMCG, poor

forecast accuracy, unspecialized forecasters,unsuitable forecasting techniques, highlydynamic markets with lots of data aboutsales, short-term business activity planning,marketing activities that push product insales (Chopra & Meindel, 2007). On theother hand, successful demand forecastinghave crucial important to support supplychain processes and distributor and retailerinventory planning (Huang et al. 2014). Itdirectly affects the key performanceindicators in supply chain such as: stockcover, out-of-stock, order-fill ratio, capacityutilization, and inventory levels (Sayed et al.2009).

In this section of paper, we present a realcase study from a company that producestwo types of FMCG - cereals for breakfastand cereal bars for snack. FMCG areproducts that are characterized by: quickbuying decisions of customers; short shelflife because of high consumer demands andrapid product deterioration; low costproduction and low profit per item, althoughthe profit of selling large quantities can belarge. All these characteristics point out theimportance and difficulties of adequate salesforecasting. Yang and Burns (2003) statedthat a forecast and planning became verycomplex because FMCG needs to beavailable on customer requests and also thereis a high risk of stock out or overstocking.

The Company’s branch in Serbia thatcollaborates with an exclusive distributoruses the sales data from the distributor tomake a successful system of forecastingsales, which on the other hand helps thedistributor to have better supply chainmanagement. The distributor collected theproduct sales data from small and big retailshops across the Serbian market fromJanuary 2012 to December 2013. The keyproblem consists of several questions and

9Z.Rakićević / SJM 10 (1) (2015) 3 - 17

dilemmas: Which forecasting methodsshould be used for sales forecasting? Do wealways have to use the same forecastingmethods that gave good results in earlierperiods? Do we have to use the sameforecasting methods for all different variantsof one product or not?

The idea was to use the focus forecastingmethodology for sales forecasting and bettermanagement of distributor’s supply chain. Inthis paper we considered sales data base ofsix different products divided into twocategories: breakfast cereals (Product A to E)with one product with different variants (B1-B5) and cereal bar snacks (Product F). Timeseries forecasting methods we used in focusforecasting were: a) Last period estimation;b) Arithmetical average of all past periods; c)Moving average for six, four, three and twomonths; d) Weighted moving average -weighted coefficient as a variable withoptimal value for error functions; e)Exponential smoothing – with constant valueof coefficient α and optimal value forcoefficient α optimised on error function; f)Trend corrected exponential smoothing(Holt's model) – with constant value ofcoefficients α and β and optimal value forcoefficients α and β optimised on error

function; g) Trend and seasonality correctedexponential smoothing (Winter’s model) -with constant value of coefficients α, β and γand with optimal value for coefficients α, βand γ.

Tables 1 to 6 present the results of focusforecasting with different forecastingmethods. For evaluating results offorecasting we used three measures of error:Root Mean Squared Error (RMSE), MeanAbsolute Percentage error (MAPE) andTracking signal (TS). We omitted MSE,MAD and Bias because they were verysimilar to the previous ones, giving the sameresults. For some forecasting methods weused Microsoft Excel Solver for determiningoptimal values of coefficients α, β and γ, forminimum of RMSE and MAPE errorfunctions. It is important to notice that wecould also use a lot of different errorfunctions in optimization, but it was not thesubject of this paper. The simulation resultsof focus forecasting are presented in Tables 1to 6. Bold values in Tables 1-6 mark the bestvalues of forecasting error, i.e. the bestapplied sales forecasting methods for thespecified criteria of forecasting error.

Table 1 presents the results of salesforecasting of products A to F using the

10 Z.Rakićević / SJM 10 (1) (2015) 3 - 17

Table 1. Sales forecasting error indicators for the following methods: last period estimation;all average; moving average for six and four months

following methods: last period estimation,all average, moving average for six and fourmonths, analysed through three errorfunctions (RMSE, MAPE, TS).

Table 2 presents the results of salesforecasting of products A to F using thefollowing methods: moving average for threeand two months, weighted moving averagefor two months optimized on RMSE andMAPE error function, analysed throughthree error functions (RMSE, MAPE, TS).

Table 3 presents the results of salesforecasting of products A to F using the

following methods: weighted movingaverage for three and four months optimizedon RMSE and MAPE error function,analysed through three error functions(RMSE, MAPE, TS).

Table 4 presents the results of salesforecasting of products A to F using thefollowing methods: weighted movingaverage for six months optimized on RMSEand MAPE error function, exponentialsmoothing with constant α and optimizedvalue of α on RMSE, analysed through threeerror functions (RMSE, MAPE, TS).

11Z.Rakićević / SJM 10 (1) (2015) 3 - 17

Table 2. Sales forecasting error indicators for the following methods: moving average forthree and two months; weighted moving average for two months

Table 3. Sales forecasting error indicators for the following methods: weighted movingaverage for three and four months

Table 5 presents the results of salesforecasting of products A to F using thefollowing methods: exponential smoothingwith optimized value of α on MAPE, Holt’smodel with constant values of α, β andoptimized values of α, β for RMSE andMAPE, analysed through three errorfunctions (RMSE, MAPE, TS).

Table 6 presents the results of salesforecasting of products A to F using thefollowing methods: Winter’s model withconstant values of α, β, γ and optimizedvalue of α, β, γ for RMSE and MAPE,analysed through three error functions

(RMSE, MAPE, TS).All these applied forecasting methods

gave different results in forecasting productsales. As we see, there is not one forecastingmethod that is common for forecasting salesof every product. Furthermore, it isinteresting that the best forecasting methodsfor summarized sales of variants of productB (ΣB) are not the same as forecastingmethods for every product B1 to B5

separately. Also, our analysis shows thateven for very similar products (B1 to B5)different forecasting methods offer the bestforecasting accuracy.

12 Z.Rakićević / SJM 10 (1) (2015) 3 - 17

Table 4. Sales forecasting error indicators for the following methods: weighted movingaverage for six months and exponential smoothing

Table 5. Sales forecasting error indicators for the following methods: exponential smoothingand Holt’s model

After all, each of the mentioned forecastmethods gives good results for salesforecasting of some products, while for theothers results are not very good.

Table 7 presents the aggregate results ofsimulation for different products anddifferent error functions. The followingforecasting methods, which compare eachother, gave the best results in terms ofminimum error functions:

- Three months weighted movingaverage optimized for RMSE (products: A,B2, B4); Six months weighted movingaverage optimized for RMSE (products: B1,

B3, B5, Sum of B, C, D, E); All average forRMSE (product F);

- Four months weighted movingaverage optimized for MAPE (products: A,B3, D, E); Six months weighted movingaverage optimized for MAPE (products: A,C); Holt’s method, αopt, βopt for MAPE(products: B1, B4, Sum of B, F); Winter’smethod, αopt, βopt, γopt for MAPE, (products:B3, B5); Exponential smoothing αopt forMAPE, (product: F);

- The same analysis was done fortracking signal (TS), as shown in Table 7.

13Z.Rakićević / SJM 10 (1) (2015) 3 - 17

Table 6. Sales forecasting error indicators for the Winter’s model

Table 7. The best forecasting methods for comparing product error function

Each forecasting method is intended for aspecific forecasting case and is under astrong influence of criteria function chosenby a decision maker, i.e. manager. However,if a manager does not have enoughexperience, he/she can “play” through thefocus forecasting system with a group offorecasting methods and a group offorecasting accuracy indicators and functionsto come to a better decision.

5. CONCLUSION

The paper presents the research of a realforecasting sales problem in a FMCG supplychain. The aim was not to explore whetherone sales forecasting method is better thatthe other, but to present the focus forecastingmethodology and describe the possibility ofmanagers’ decision support in forecastingsales.

We analysed a real business practiceproblem of forecasting sales of two differenttypes of products, using the following salesforecasting methods: Last period estimation,all average, moving average, weightedmoving average, exponential smoothing,Holt's model and Winter’s model. Theevaluation of forecasting success was doneby the following forecasting error indicators:RMSE, MAPE and TS.

Obtained results in focus forecastinganalysis have showed that there is not onesales forecasting method which gave the bestresults of forecasting (i.e. smallestforecasting error) for all observed products.Different methods give different results thatdepend on the past sales of products and theway managers observe the forecastingaccuracy. So, it is necessary that forecastingtechniques are being monitored and

evaluated continuously. Now days, it ispossible by using modern informationsystems like business intelligence andbusiness analytics systems. This case studyalso indicates the need for it.

The limitation of this paper is reflected inusing specific type of FMCG supply chain,where one exclusive distributor control thewhole supply chain and collect theinformation about retail sales. That case ismuch easier for implementing successfulforecasting system in supply chain systemthan in systems with lot of distributors.Limitation of this paper and presented focusforecasting system is also reflected in thefollowing: We use short range of FMCG(two type cereals for breakfast and cerealsbars), who may had specific characteristicsthat influences sales in compared maybesome to another type of FMCG. We didn’tuse information about phases of product lifecycle, which can have an impact on movingsales of one product. The sales of productsthat are being forecasted in this analysis arethe basic ones without any assumptions ofmarketing activity influences that may haveinduced strong and rapid sales fluctuations insupply chain.

The implications of this research shouldpoint out to new possibilities for upgradingenterprise forecasting systems, especially inenterprises that are in a supply chain withFMCG. Further research on this topic caninclude more time-series forecastingmethods and methods that are in a group ofassociative methods. The overall purpose is acontinuous improvement of enterpriseperformances.

14 Z.Rakićević / SJM 10 (1) (2015) 3 - 17

15Z.Rakićević / SJM 10 (1) (2015) 3 - 17

References

Bratu-Simionescu, M. (2013).Improvements in assessing the forecastsaccuracy: A case study for Romanianmacroeconomic forecasts. Serbian Journal ofManagement, 8 (1), 53-65.

Chang, P.C., & Lin, Y.K. (2010). Newchallenges and opportunities in flexible androbust supply chain forecasting systems –Editorial. International Journal of ProductionEconomics, 128 (2), 453-456.

Chen, A., & Blue, J. (2010). Performanceanalysis of demand planning approaches for

aggregating, forecasting and disaggregatinginterrelated demands. International Journalof Production Economics, 128 (2), 586-602.

Chopra, S., & Meindl, P. (2007). Supplychain management. Strategy, planning &operation. New Jersey: Pearson, PrenticeHall.

Davydenko, A., & Fildes, R. (2013).Measuring forecasting accuracy: The case ofjudgmental adjustments to SKU-leveldemand forecasts. International Journal ofForecasting, 29 (3), 510-522.

Ferbar, L., Čreslovnik, D., Mojškerc, B.,& Rajgelj, M. (2009). Demand forecasting

16 Z.Rakićević / SJM 10 (1) (2015) 3 - 17

ФОКУСНО ПРЕДВИЂАЊЕ У ЛАНЦУ СНАБДЕВАЊА:

СТУДИЈА СЛУЧАЈА КОМПАНИЈЕ СА РОБОМ ШИРОКЕ

ПОТРОШЊЕ У СРБИЈИ

Зоран Ракићевић, Мирко Вујошевић

Извод

Овај рад представља примену фокусног предвиђања у ланцу снабдевања робе широкепотрошње. Фокусно предвиђање је тестирано на стварном примеру из пословања једногпредузећа у Србији. Подаци који су коришћени у симулацији односе се на историјске податкео продаји две групе робе широке потрошње са неколико различитих производа. Подаци о овимпроизводима су прикупљени и сумирани кроз читав дистрибутивни канал на тржишту Србијеу периоду од јануара 2012. до децембра 2013. године. На подацима о продаји је примењенонеколико познатих метода за предвиђање временских серија, коришћењем приступа фокусногпредвиђања. Приступ фокусног предвиђања комбинује различите методе предвиђања навременској серији података у циљу проналажења методе која је била најпрецизнија упредвиђању протеклих периода, за предвиђање у будућем периоду. Тачност предвиђања једефинисана помоћу неколико индикатора којима се мере грешке предвиђања. У овом раду, узпомоћ “Microsoft Excel-а”, имплементиране су следеће методе предвиђања продаје:предвиђање по последњем периоду продаје, просек продаје из свих претходних временскихпериода, покретни просек, експоненцијално изравнање са константним и варијабилнимпараметром α, експоненцијално изравнање са укљученим трендом, експоненцијалноизравнање са укљученим трендом и сезоналношћу. Основна сврха у раду није била евалуацијаразличитих метода предвиђања, већ приказ практичне примене фокусног предвиђања нареалном пословном примеру.

Кључне речи: фокусно предвиђање, покретни просек, експоненцијално изравнање, “Holt“-овмодел, “Winter“-ов модел, роба широке потрошње

methods in a supply chain: Smoothing anddenoising. International Journal ofProduction Economics, 118 (1), 49-54.

Fildes, R., Goodwin, P., Lawrence, M., &Nikolopoulos, K. (2009). Effectiveforecasting and judgmental adjustments: anempirical evaluation and strategies forimprovement in supply-chain planning.International Journal of Forecasting, 25 (1),3-23.

Gardner Jr, E.S., Anderson-Fletcher, E.A.,& Wicks, A.M. (2001). Further results onfocus forecasting vs. exponential smoothing.International Journal of Forecasting, 17 (2),287-293.

Heizer, J., & Render, B. (2011).Operations Management. New Jersey:Prentice Hall.

Huang, T., Fildes, R., & Soopramanien,D. (2014). The value of competitiveinformation in forecasting FMCG retailproduct sales and the variable selectionproblem. European Journal of OperationalResearch, 237 (2), 738-748.

Kerkkänen, A., Korpela, J., & Huiskonen,J. (2009). Demand forecasting errors inindustrial context: Measurement andimpacts. International Journal of ProductionEconomics, 118 (1), 43-48.

Min, H., & Yu, W.B.V. (2008).Collaborative planning, forecasting andreplenishment: demand planning in supplychain management. International Journal ofInformation Technology and Management, 7(1), 4-20.

Oyatoye, E.O., & Fabson, T.V.O. (2011).A comparative study of simulation and timeseries model in quantifying bullwhip effectin supply chain. Serbian Journal ofManagement, 6 (2), 145-154.

Ramanathan, U., & Muyldermans, L.(2010). Identifying demand factors forpromotional planning and forecasting: A case

of a soft drink company in the UK.International Journal of ProductionEconomics, 128 (2), 538-545.

Sayed, H.E., Gabbar, H.A., & Miyazaki,S. (2009). A hybrid statistical genetic-baseddemand forecasting expert system. ExpertSystems with Applications, 36 (9), 11662-11670.

Thomassey, S. (2010). Sales forecasts inclothing industry: the key success factor ofthe supply chain management. InternationalJournal of Production Economics, 128 (2),470-483.

Yang, B., & Burns, N. (2003).Implications of postponement for the supplychain. International Journal of ProductionResearch, 41 (9), 2075-2090.

Vayvay, O., Dogan, O., & Ozel, S. (2012).Forecasting techniques in fast movingconsumer goods supply chain: a modelproposal. International Journal ofInformation Technology and BusinessManagement, 13 (1), 118-128.

Vujošević, M. (1997). Operationalmanagement: quantitative methods.Yugoslav Operational Research Society –DOPIS, Belgrade. (in Serbian)

Warren Liao, T., & Chang, P.C. (2010).Impacts of forecast, inventory policy, andlead time on supply chain inventory – anumerical study. International Journal ofProduction Economics, 128 (2), 527-537.

Wallström, P., & Segerstedt, A. (2010).Evaluation of forecasting errormeasurements and techniques forintermittent demand. International Journal ofProduction Economics, 128 (2), 625-636.

Zamarripa, M.A., Aguirre, A.M., Méndez,C.A., & Espuña, A. (2012). Improvingsupply chain planning in a competitiveenvironment. Computers & ChemicalEngineering, 42, 178-188.

17Z.Rakićević / SJM 10 (1) (2015) 3 - 17