Traffic Movement in Lufthansa Airlines: a Supply Chain Perspective

Journal of Services Research Volume 10 Number 2 October 2010 – March 2011 FORECASTING THE PASSENGER TRAFFIC MOVEMENT IN LUFTHANSA AIRLINES: A SUPPLY CHAIN PERSPECTIVE Aniruddh Kr Singh Faculty of Management Studies University of Delhi, India. Debadyuti Das Associate Professor, Faculty of Management Studies University of Delhi, India. The Journal of IIMT FORECASTING THE PASSENGER TRAFFIC MOVEMENT IN LUFTHANSA AIRLINES: A SUPPLY CHAIN PERSPECTIVE Aniruddh Kr Singh Debadyuti Das
The present paper attempts to find out the forecasted passenger traffic movement of Lufthansa Airlines on quarterly basis at a global level by employing four forecasting methods namely moving average, exponential smoothing, Holt’s model and Winter’s model with the help of published data pertaining to passenger traffic movement of Lufthansa Airlines. The study has also found out the forecasting errors of all the four methods through Absolute error (AE), Mean squared error (MSE), Mean absolute deviation (MAD) and Mean absolute percentage error (MAPE).
The study also carried out the comparative analyses of the above forecasting methods in the light of the available data. The findings reveal that the forecasting errors are the least in case of Winter’s model. Further the forecasted values suggested by Winter’s model more closely resemble the observed data of passenger traffic movement of Lufthansa Airlines. This provides a valuable insight to the top management as regards formulation of suitable strategies for addressing the varying demand of passenger traffic movement.

Few strategies in respect of both demand side and supply side options have been suggested with a view to improving the overall supply chain profit of Lufthansa Airlines. INTRODUCTION irlines industry across the globe is currently undergoing recession due to severe financial crisis faced by the major economies of the world. As per the estimates of International Air Transport Association (IATA), globally air travel has declined by 2. 9% and 1. 3% during September and October, 2008 respectively compared to the same months in the previous year.
Segment-wise passenger traffic estimates provided by IATA further reveal that the Asia Pacific Carriers and North American Carriers registered a decline in passenger traffic flow by 6. 1% and 0. 9% respectively in October, 2008 compared to the same month in the previous year. African Carriers recorded the largest decline in traffic flow by 12. 9% in October, 2008 Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) ©2010 by Institute for International Management and Technology. All Rights Reserved. A 4 Forecasting the Passenger compared to the same month in the previous year. The remaining segments namely European, Latin American and Middle Eastern Airlines experienced a moderate growth in its traffic flow to the tune of 1. 8%, 4. 5% and 3. 5% respectively in October, 2008 (IATA International traffic statistics, 2008a, 2008b). However, the financial crisis sweeping across the globe does not appear to have much negative impact on Lufthansa Airlines in respect of its passenger traffic flow till September, 2008 as revealed from the data provided in table 2a.
A cursory observation into the table 2 further demonstrates that the passenger traffic flow in Lufthansa Airlines has been following a very systematic pattern since October, 2006 to September, 2008. There has been hardly any departure from the pattern observed in passenger traffic movement during the above period. Despite difficult market conditions, Lufthansa passenger Airlines was able to achieve a sales growth of 4. 2% and 0. 7% in September and October, 2008 respectively.
It registered an increase in its passenger traffic flow in three major markets namely America (North/South), Asia/ Pacific, and Middle East & Africa both during September and October, 2008. American segment recorded a growth rate of 6. 9% and 1% during September and October, 2008 respectively. Asia/Pacific region exhibited an increasing trend of 8. 8% and 6% while Middle East and African region recorded an increasing trend of 2. 5% and 11% during September and October, 2008 respectively. Only European market experienced a declining trend to the tune of 0. 4% and 3% during the above periods (Lufthansa Investor Info, page 1, 2008).
The above phenomenon has motivated us to apply the most popular and well-established forecasting methods with a view to finding out the forecasted demand of passenger traffic movement of Lufthansa Airlines for future periods. The main objective of the paper is to find out the quarterly forecasted demand of passenger traffic flow in Lufthansa Airlines at a global level with the help of moving average (MA), exponential smoothing (ES), Holt’s model and Winter’s model by making use of published data pertaining to passenger traffic movement in Lufthansa Airlines.
In addition, the paper has also attempted to find out the most suitable forecasting model for the above problem by comparing the forecasting errors of the above four forecasting models obtained through absolute error (AE), mean squared error (MSE), mean Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 65 Singh, Das absolute deviations (MAD) and mean absolute percentage error (MAPE). The following section provides a brief review of literature. Section 3 provides a brief overview of Lufthansa Airlines along with the recent data on passenger traffic movement.
It contains a thorough analysis of forecasted passenger traffic movement by employing four forecasting methods and the comparative analysis of the same. Section 4 suggests few strategies for absorbing the varying nature of demand. The paper is concluded with a brief summary, potential contribution and limitations of the same. REVIEW OF LITERATURE Forecasting literature is replete with a number of studies ranging from simple time-series forecasting models to econometric models as also the forecasting models employing artificial intelligence techniques etc.
Researchers have employed the forecasting models with a view to finding out the forecasted demand of traffic for a particular period. However, the study findings reveal that there does not exist a single model which consistently outperforms other models in all situations. Quantitative forecasting methods can be categorized under three broad heads: (1) time-series modeling, (2) econometric models and (3) other quantitative models (Song and Li, 2008). Under time-series models, several techniques are available, e. g.
Moving Average, Exponential Smoothing, Holt’s Model, Winter’s Model, ARIMA etc. (Makridakis et al, 2003). In time-series model, particular attention is paid to exploring the historic trends and patterns of the time-series involved and to predict the future of this series based on trends and patterns identified in the model. Since time-series models require only historical observations of a variable, it is less costly in data collection and model estimation. However, these models cannot account for the changes in demand that might occur in different periods.
The major advantages of econometric models over time-series models lie in their ability to analyze the causal relationships between the demand and its influencing factors (Song and Li, 2008; Makridakis et al, 2003). It is possible for econometric models to take into consideration several variables together, for example, air fare charged by an airline, competitive fare offered by other airlines, promotional campaign, perceived security threat, price and income elasticity of Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 6 Forecasting the Passenger demand etc. However, it is difficult and costly to collect data on each individual variable, incorporate the same into the model and explain its contribution towards the dependent variable. A number of new quantitative forecasting methods, predominantly Artificial Intelligence (AI) techniques, have emerged in forecasting literature. The main advantage of AI techniques is that it does not require any preliminary or additional information about data such as distribution and probability (Song and Li, 2008).
Table 1 provides a brief overview of some related works pertaining to forecasting and traffic movement in airlines. Table 1: Brief Overview of Few Works Relating to Traffic Movement in Airlines Author Choo and Mokhtarian (2007) Contribution Developed a conceptual model in a comprehensive framework, considering causal relationships among travel, telecommunications, land use, economic activity and socio-demographics and explored the aggregate relationships between telecommunications and travel using structural equation modeling of national time-series data pning 1950-2000 in the US.
Proposed an artificial neural network (ANN) structure for seasonal time-series forecasting. Results found by the proposed ANN model were compared with the traditional statistical models which reveal that the prediction error of the proposed model is lower than the traditional models. The proposed model is especially suitable when the seasonality in time-series is very strong. Developed a methodology for assessing the future route network and flight schedule at a medium-sized European airport.
The existing origin and destination demand from the base airport across the world is considered. In addition, the growth rates by country or region is also taken into account. The future origin and destination demand in then converted into route traffic subject to a threshold for direct service. Where demand falls below this level, traffic is reallocated via various appropriate hubs. Applied Static-regression trend-fitting model for the purpose of forecasting future tourism demand in North Cyprus.
Applied different types of time-series forecasting modeling with reference to China and compared the forecasting accuracy of the models. Applied different types of time-series forecasting modeling with reference to Australia for the purpose of forecasting business tourism and compared the forecasting accuracy of the models. Employed autoregressive distributed lag model (ADLM) for the purpose of forecasting tourism demand at Greece.
Hamzacebi (2008) Dennis (2002) Bicak, Altinay and Jenkins (2005) Kulendran and Shan (2002) Kulendran and Witt (2003) Dritsakis and Athanasiadia (2000) THE CASE OF LUFTHANSA AIRLINES Deutsche Lufthansa (Lufthansa), the third largest airlines of Europe, is the world’s fifth largest airline in terms of overall passengers carried and operating services to 209 destinations in 81 countries. It has the 6th largest passenger airline fleet in the world.
Lufthansa is headquartered in Cologne, Germany with its main base and primary traffic hub at Frankfurt International Airport in Frankfurt and a second hub at Munich International Airport. Lufthansa has built a premium brand synonymous with quality, innovation, reliability, competence and safety despite operating in a tough market where cost cutting is commonplace. Lufthansa founded the world’s first multilateral airline grouping, ‘Star Alliance’ along with Air Canada, SAS, Thai Airways and United Airlines.
At the same time, the airline invested in the most advanced passenger aircrafts and in 1999 it embarked on a vast IT programme that would transform the revenue and profit of its passenger Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 67 Singh, Das airline business (Lufthansa, Wikipedia, 2008). However, estimating the demand of passenger traffic for a particular period has always been the principal determinant in generating revenue for the airline. Table 2a shows the passenger traffic movement in Lufthansa (excluding the number in Swiss Airlines) Airlines for the period during October, 2006 to September, 2008.
Table 2 (a): Monthly Traffic Flow for the Last Two Years Traffic Year – Month Oct-06 Nov-06 Dec-06 Jan-07 Feb-07 Mar-07 Apr-07 May-07 Jun-07 Jul-07 Aug-07 Sep-07 Oct-07 Nov-07 Dec-07 Jan-08 Feb-08 Mar-08 Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Passenger traffic (in thousands) 4936 4327 3969 3851 3820 4668 4635 4991 5003 5241 5067 5193 5241 4604 4132 4141 4223 4625 5031 5152 5203 5171 4883 5164 2006 Q- 4 2007 Q- 1 2007 Q- 2 2007 Q- 3 2007 Q- 4 2008 Q- 1 2008 Q- 2 2008 Q- 3 13232000 12339000 14629000 15501000 13977000 12989000 15386000 15218000
Table 2 (b): Quarterly Data of Passenger Quarters Passenger traffic Source of data: Key data, Lufthansa Investor Relations, 2008; Lufthansa Investor Info, page 2, 2008 The monthly passenger traffic shown in table 2 (a) has been utilized to calculate the quarterly data of passenger traffic for the last two years Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 68 Forecasting the Passenger (from Quarter 4, 2006 to Quarter 3, 2008) which has been shown in table 2 (b).
With the help of these quarterly data of passenger traffic for the last two years, we have attempted to find out the forecasted values of passenger traffic movement by employing four forecasting methods namely 4-period Moving Average, Simple Exponential Smoothing, Holt’s Model and Winter’s Model. Table 3 presents the forecasted values through 4-quarter moving average while table 4 shows the forecasted data through simple exponential smoothing. Table 5 and 6 shows the forecasting through Holt’s model along with forecasting errors.
Table 7 through 10 reveals, in detail, the forecasted demand of the passenger traffic flow by employing Winter’s Model. Table 10 also includes the forecasting errors. The exercise reveals that the forecasting errors are the lowest in case of Winter’s Model which are indicated by the values of AE, MSE, MAD and MAPE. Moreover, the quarterly forecasted values suggested by Winter’s Model closely follow historical pattern which is clearly depicted in figure 1. FORECASTING THROUGH 4-PERIOD MOVING AVERAGE (MA) Moving Average method is generally employed in a situation in which only level, i. e. eseasonalized demand is present and neither trend nor seasonality is observed. We took the average traffic flow of four quarters starting from the 4th quarter of 2006 and continued the exercise till the 3 rd quarter of 2008 for the purpose of finding out the forecasted passenger traffic movement in the immediate following quarter. Table 3 presents the forecasted values of passenger traffic movement through four-quarter MA method. In the same table, the values of forecasting errors measured in terms of AE, MSE, MAD and MAPE are also shown. Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 9 Singh, Das Table 3: Forecasting through 4-Period Moving Average & Forecasting Errors Period(t) 1 2 3 4 5 6 7 8 Quarters Traffic (D) Level (L) Forecast (F) Four Period Moving Average Method Absolute Error Mean Squared Error Error (E) (AE) (MSE) Mean Absolute Deviation (MAD) 2006 Q- 4 13232000 2007 Q- 1 12339000 2007 Q- 2 14629000 2007 Q- 3 15501000 13925250 2007 Q- 4 13977000 14111500 13925250 2008 Q- 1 12989000 14274000 14111500 2008 Q- 2 15386000 14463250 14274000 2008 Q- 3 15218000 14392500 14463250 -51750 1122500 -1112000 -754750 51750 1122500 1112000 754750 2678062500 6. 31342E+11 8. 3076E+11 7. 67219E+11 51750 587125 762083. 3333 760250 % Error MAPE Forecasted Traffic F9=F10=F11=F12=14392500 0. 37025113 0. 37025113 8. 64192779 4. 50608946 7. 22734954 5. 41317615 4. 95958733 5. 29977895 Formula used Systematic demand = Level Lt= (Dt + Dt-1+….. Dt-n+1)/N Ft+1=Lt Ft+n=Lt (Chopra and Meindl, 2007) FORECASTING THROUGH EXPONENTIAL SMOOTHING (ES) Like moving average method, exponential smoothing is also used in a situation, in which only level is observed. However, ES attempts to smoothen the fluctuations observed in demand data of different periods through smoothing constant (alpha).
We first calculated the level of passenger traffic flow of the initial period by taking the average of actual traffic flow for the last eight quarters, which has been considered as the forecasted value of passenger traffic flow for quarter 1. Table 4 demonstrates the forecasted values through simple ES. The same table also contains the values of forecasting errors expressed in terms of AE, MSE, MAD and MAPE. Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 70 Forecasting the Passenger Table 4: Forecasting through Simple Exponential Smoothing & Forecasting Errors Period(t) 0 1 2 3 4 5 6 7 8 % Error 7. 0479897 13. 9977916 5. 02789835 9. 89599461 1. 02611209 8. 60018261 9. 04478131 7. 12621269 2006 Q- 4 2007 Q- 1 2007 Q- 2 2007 Q- 3 2007 Q- 4 2008 Q- 1 2008 Q- 2 2008 Q- 3 MAPE 7. 00479897 10. 5012953 8. 67682963 8. 98162087 7. 39051912 7. 5921297 7. 79965136 7. 71547153 Formula used Systematic demand = Level Ft+1=Lt Ft+n=Lt Lt+1=alpha(Dt+1)+(1-alpha)Lt alpha=0. 1 Forecasted Traffic F9=F10=F11=F12=14241980 13232000 12339000 14629000 15501000 13977000 12989000 15386000 15218000 Quarters Traffic (D) Level (L) 14158875 14066187. 5 13893468. 75 13967021. 8 14120419. 69 14106077. 72 13994369. 95 14133532. 95 14241979. 66 14158875 14066187. 5 13893468. 75 13967021. 88 14120419. 69 14106077. 72 13994369. 95 14133532. 95 926875 1727187. 5 -735531. 25 -1533978. 1 143419. 688 1117077. 72 -1391630. 1 -1084467 926875 1727187. 5 735531. 25 1533978. 125 143419. 6875 1117077. 719 1391630. 053 1084467. 048 8. 59097E+11 1. 92114E+12 1. 46109E+12 1. 68409E+12 1. 35139E+12 1. 33413E+12 1. 42021E+12 1. 38969E+12 926875 1327031. 25 1129864. 583 1230892. 969 1013398. 313 1030678. 214 1082242. 762 1082520. 98 Forecast (F) Simple Exponential Smoothing Method Absolute Error Error (E) (AE) Mean Squared Error (MSE) Mean Average Deviation (MAD) (Chopra and Meindl, 2007) FORECASTING THROUGH HOLT’S MODEL We carried out a regression analysis wherein Time period was considered on X-axis and passenger traffic data was taken on Y-axis in order to find out the initial level and trend. Holt’s model, also known as trend-corrected exponential smoothing, is applicable in a situation, in which level and trend are observed in the demand data. However, seasonality is not considered in Holt’s model.
We used the “Linest Function”of Microsoft Excel to calculate the values of L0 and T0, which is shown in table 5. Table 5: Regression to Find Initial Level and Trend for Holt’s Model x (Period) 1 2 3 4 5 6 7 8 270154. 7619 T0 y (Traffic) 13232000 12339000 14629000 15501000 13977000 12989000 15386000 15218000 12943178. 57 L0 Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 71 Singh, Das Once the initial values of level of trend are found, the subsequent values of the level and trend of each period are iteratively calculated following Holt’s model which is shown in table 6.
This finally helps in finding out the forecasted values of passenger traffic movement as per Holt’s model, which is shown in table 6. Table 6 also reveals the forecasting errors. Table 6: Forecasting through Holt’s Model Period(t) 0 1 2 3 4 5 6 7 8 2006 Q- 4 2007 Q- 1 2007 Q- 2 2007 Q- 3 2007 Q- 4 2008 Q- 1 2008 Q- 2 2008 Q- 3 13232000 12339000 14629000 15501000 13977000 12989000 15386000 15218000 Quarters Traffic (D) Trend(T) 270528. 095 Level (L) 13215200 Forecast (F) 13213333. 33 13485728. 1 13618648. 82 13987484. 49 14436906. 91 14679788. 95 14765767 15095251. 1 Error (E) -18666. 67 1146728. 1 -1010351 -1513516 459906. 91 1690788. 9 -620233 -122748. 1 Absolute Error (AE) 18666. 66667 1146728. 095 1010351. 181 1513515. 506 459906. 9118 1690788. 949 620232. 9957 122748. 0864 T8=269916. 6 15377443 15647360 15917276 16187193 Formula used Systematic demand = Ft+1=Lt+T t alpha =0. 1 Beta = 0. 2 Lt+1 = alpha(D t+1)+(1-alpha)(Lt+T t) T t+1= beta(Lt+1-Lt)+(1-beta)Tt Level + Trend Ft+n =Lt+nT t Mean Squared Error (MSE) 348444444. 4 6. 57667E+11 7. 78714E+11 1. 15672E+12 9. 67677E+11 1. 28286E+12 1. 15455E+12 1. 01211E+12 270154. 762 12943178. 7 247593. 533 13371055. 29 267800. 557 13719683. 94 298070. 867 14138836. 04 288872. 729 14390916. 22 255056. 95 267461. 61 14510710. 05 14827790. 3 269916. 571 15107526. 72 Mean Average Deviation (MAD) 18666. 66667 582697. 381 725248. 6476 922315. 3622 829833. 6721 973326. 2183 922884. 3294 822867. 299 % Error 0. 141072148 9. 293525369 6. 906495187 9. 763986233 3. 290455117 13. 0170833 4. 031151668 0. 806598018 MAPE 0. 141072148 4. 717298758 5. 447030901 6. 526269734 5. 879106811 7. 068769558 6. 634824146 5. 90629588 L8=15107527 F9 F10 F11 F12 Forecasted Traffic Chopra and Meindl, 2007) FORECASTING THROUGH WINTER’S MODEL Winter’s model, also known as trend and seasonality-corrected ES, is generally employed in a situation in which all characteristic features of demand data, i. e. level (Lt), trend (Tt) and seasonality (St) are observed. The actual demand (Dt), being seasonal in nature, is transformed into deseasonalized demand (Ddt ). The deseasonalized demand data and corresponding time periods are employed to run regression analysis in order to calculate the initial level (L0) and trend (T0) which is shown in table 7.
The values of L0 and T0 are then used to find out the estimated deseasonalized demand (Dt) of passenger traffic of different time periods. Seasonal factors for each period are calculated using the formula Dt /(Dt) as shown in table 8. Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 72 Forecasting the Passenger Table 7: Regression Analysis for Finding out the Deseasonalized Demand X (Period) 3 4 5 6 140439. 5 Y (Deseasonalized demand)(Ddt) 14018375 14192750 14368630 14427880 13619931 T0 L0 Table 8: Calculation of Seasonal Factors for Winter’s Model
Period(t) 0 1 2 3 4 5 6 7 8 2006 Q- 4 2007 Q- 1 2007 Q- 2 2007 Q- 3 2007 Q- 4 2008 Q- 1 2008 Q- 2 2008 Q- 3 13232000 12339000 14629000 15501000 13977000 12989000 15386000 15218000 14018375 14192750 14368630 14427880 13760370. 5 13900810 14041249. 5 14181689 14322128. 5 14462568 14603007. 5 14743447 0. 961602015 0. 887646116 1. 041858846 1. 093029187 0. 97590243 0. 898111594 1. 053618578 1. 032187385 Quarters Actual demand (Dt ) Deseasonalized demand (Ddt) Dt =L+Tt Seasonal factors (Dt / D t) Subsequently seasonality (St) is recalculated for each period as per Winter’s model which is shown in table 9.
Level and trend of each period are also iteratively calculated following Winter’s model which have been mentioned in detail in table 9. Finally table 10 demonstrates the forecasted data of passenger traffic flow along with forecasting errors. Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 73 Singh, Das Table 9: Determination of Level, Trend and Seasonal Factors (Winter’s Model) Period(t) Quarters Actual Traffic (Dt) Deseasonalized demand (Ddt) Estimated deseasonalized demand (Dt) 13760370. 5 13900810 14018375 14192750 14368630 14427880 14041249. 5 14181689 14322128. 14462568 14603007. 5 14743447 Seasonality St Level(L) Trend(T) 0 1 2 3 4 5 6 7 8 9 10 11 12 2006 Q- 4 2007 Q- 1 2007 Q- 2 2007 Q- 3 2007 Q- 4 2008 Q- 1 2008 Q- 2 2008 Q- 3 13232000 12339000 14629000 15501000 13977000 12989000 15386000 15218000 0. 968752222 0. 892878855 1. 047738712 1. 062608286 0. 968072702 0. 892415518 1. 047252432 1. 065603208 0. 968770988 0. 892874843 1. 047722994 1. 062255808 13619931 13755292. 34 13891430. 02 14027555. 72 14187811. 57 14334567. 79 14480348. 88 14626058. 49 14744278 140439. 5 139931. 6844 139552. 284 139209. 6254 141314. 2474 141858. 4444 142250. 709 142596. 999 140158. 8902 Table 10: Forecasting through Winter’s Model and the Forecasting Errors Forecast(F) 13330389. 5 12406751. 72 14700803. 33 15053722. 24 13871635. 54 12918987. 41 15313552. 98 15737526. 24 Error(E) 98389. 50148 67751. 71749 71803. 33314 -447277. 7569 -105364. 4571 -70012. 58968 -72447. 01855 519526. 2416 Absolute Error(AE) 98389. 50148 67751. 71749 71803. 33314 447277. 7569 105364. 4571 70012. 58968 72447. 01855 519526. 2416 Mean Squared Error (MSE) 9680494002 7135394612 6475502625 54870974917 46117113697 39247888533 34390843099 63830427174 Mean Average Deviation (MAD) 98389. 0148 83070. 60949 79314. 85071 171305. 5772 158117. 3532 143433. 226 133292. 3392 181571. 577 % Error 0. 743572411 0. 549085967 0. 490828718 2. 885476788 0. 753841719 0. 539014471 0. 470863243 3. 413893032 MAPE 0. 743572411 0. 646329189 0. 594495699 1. 167240971 1. 084561121 0. 993636679 0. 91895476 1. 230822044 L8=14407445 T8=3284577 Formula used Systematic component of demand =(level+demand)*seasonal factor Ft+1 = (Lt+T t)St+1 Ft+i=(Lt+iTt)St+i L t+1 = alpha (Dt+1/St+1)+(1-alpha)(Lt+Tt) T t+1= Beta (Lt+1 – Lt) + (1- Beta)T t St+p+1= gamma (Dt+1/Lt+1) + (1-gamma)St+1 Alpha = 0. 5 beta=0. 1 gamma=0. 1 Forecasted traffic F9 F10 F11 F12 14419610. 62 13415083. 6 15888462. 17 16257733. 32 (Chopra and Meindl, 2007) COMPARISON AMONG FOUR FORECASTING METHODS The following figure gives an interesting revelation regarding the behaviour of forecasted data by comparing the quarterly forecasted demand of passenger traffic obtained through all four methods. Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 74 Forecasting the Passenger Historical traffic Forecasted traffic Moving Average Simple exponential smoothing Holt’s Model Winter’s Model
Figure 1: Comparison among four forecasting methods The portion of the graph before the vertical line indicates historical data while the portion of the graph after the line is the forecasted data. The forecasted data of the model graph (Winter’s Model) replicates the historical data. It indicates a positive trend as well as seasonality. FORMULATION OF SUITABLE STRATEGIES FOR ABSORBING VARYING DEMAND Keeping in view the overall objective of improving the supply chain profit, the management should explore all possible alternatives of both demand side as well as supply side options.
It is observed that demand for passenger traffic movement is not uniform throughout the year. In order to level the demand, the management of the airlines can undertake the following well-established measures: • • Formulate suitable marketing strategies to create new demand in the lean period. During peak periods, when the demand will exceed capacity, the management needs to offer seats to the customers who will pay the highest fares. Of course, other customers need to be motivated and informed that they would probably be charged less fare, if they undertake their trip at some other period.
Shift some proportion of demand from peak period to lean period by offering the customers a reasonable rate of discount in the lean period. Of course, the cost/benefit analysis of this exercise has to be thoroughly examined beforehand. • Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 75 Singh, Das • Considering the lean periods of the airline in different routes and destinations, the top management needs to explore new destinations which may appear to be very attractive from the perspective of the customers.
Accordingly the management can withdraw some of the flights from the existing underloaded routes and ply the same in the new routes. Alternatively the management needs to examine the passenger traffic data of different routes on monthly/quarterly basis. If it is found that during the same period, some destinations experience very high demand while others have low demand, the management may withdraw some of the flights from underutilized routes and introduce the same in the heavily loaded routes. •
In all cases, the detailed cost/benefit analysis of different alternatives is to be thoroughly examined. Then a particular course of a strategy or a combination of strategies may be adopted by the management. CONCLUSION The present study has attempted to find out the quarterly forecasted demand of passenger traffic flow of Lufthansa Airlines by employing the four forecasting methods, viz. moving average, simple exponential smoothing, Holt’s model and Winter’s model. The forecasted data suggested by Winter’s model reflect the historical pattern in a better manner than three other forecasting methods.
This gives a valuable insight to the managers regarding formulation of appropriate strategies in order to absorb varying nature of demand in different quarters. The same kind of study can be replicated in other airlines with suitable modifications. Of course, the present work have not taken into consideration important factors, for example, the prevailing slowdown in the global economy, perceived security threat in the wake of terrorist strikes at different parts of the globe etc.
Moreover, the study has considered the total passenger traffic movement of Lufthansa as a whole and has not paid attention to an individual market segment. This may not provide a clear picture to the management regarding increase or decrease in traffic flow in a particular segment. Future study should take care of this aspect. Journal of Services Research, Volume 10, Number 2 (October 2010 – March 2011) 76 Forecasting the Passenger The implications of varying demand on supply side need to be thoroughly examined and accordingly suitable strategies should be adopted for improving the profit across the whole supply chain.
REFERENCES Bicak, H. A. , Altinay, M. & Jenkins, H. (2005) ‘Forecasting tourism demand of North Cyprus’, Journal of Hospitality and Leisure Marketing, Vol. 12, pp. 87-99. Chopra, S and Meindl, P (2007) Supply Chain Management: Strategy, Planning & Operation, 3rd edition, Pearson Education, New Delhi. Choo S. and Mokhtarian, P. L. (2007) ‘Telecommunications and travel demand and supply: Aggregate structural equation models for the US’, Transportation Research Part A, 41 pp. 4 -18. Dennis, N. P. S. 2002) ‘Long-term forecasts and flight schedule pattern for a medium-sized European airport’, Journal of Air Transport Management, Vol. 8, pp. 313-324. Dritsakis, N. and Athanasiadis, S. (2000) ‘An econometric model of tourist demand: The case of Greece’, Journal of Hospitality and Leisure Marketing, Vol. 7, pp. 39-49. Hamzacebi, C. (2008) ‘Improving artificial neural networks’ performance in seasonal time series forecasting’, Information Sciences, Vol. 178, pp. 4550-4559. IATA International traffic statistics, 2008a, Facts & Figures – 2008 Traffic Results, Montreal, Quebec, viewed 30 November,