SlideShare ist ein Scribd-Unternehmen logo
1 von 7
SUPPLY CHAIN PLANNING
DEMAND ANALYSIS PROJECT
G. Lauson, M.S., P.E.
January 31, 2017
Purpose
This paper presents the findings of an analysis of four (4) pharmaceutical products. Four (4)
forecasting methods were applied and three (3) error measurements were applied to those
forecasting methods to determine which method most accurately estimated next-future-month
demand for each product. Monthly data was selected at the request of the manufacturing
division, as this was determined by them as optimum for their production schedule planning.
Note that all forecasting and error methods / descriptions were taken from a Coursera / Rutgers
University online course titled “Supply Chain Planning” which the author completed in January
2017. The course is part of a five-course offering within Coursera’s “Supply Chain Management
Specialization.”
Forecasting Methods
The following forecasting methods were evaluated for accuracy with respect to each of the four
pharmaceutical products. A brief explanation of the rationale behind each method is provided,
along with how the method is computed.
1. Naïve Method. The simplest of the four, this method uses the prior-period actual demand
as the forecast for the next-period demand. The computation is:
Ft = Dt-1
2. Cumulative Mean Method. This method averages all demand from prior periods – that
average becomes the forecast. Based on information provided in the above-referenced
Coursera course, this is a very stable forecast that averages out all noise; however, that
stability may be contrary to current demand conditions. The assumption behind this method
is that “all prior data is equally useful.” The computation is:
Ft = Σ i = 1 to t–1 [Di] / (t – 1)
3. Simple Moving Average (SMA) Method. An adaptable forecasting method that can be
made reactive or stable. Companies with stable demand tend to like this indicator. The
computation is:
Ft = Σ i = (t-N+2) to t–1 [Di] / N
4. Exponential Smoothing (or Exponential Moving Average (EMA)) Method. A weighted
average of all past demand; most of the weight is on the latest observed period and the
remainder of the weights decline exponentially. α changes the weights and varies how much
weight (in the weighted average) is applied to each period. The computation is:
Ft = α * [Dt-1] + (1 – α) * Ft-1
Error Estimation Methods
The following methods were used to estimate the error associated with each of the above
forecasting methods.
1. Mean Error. This is the simplest forecast accuracy measure, and is more a measure of
bias than accuracy. This error measurement averages the difference between true demand
and associated forecast for the forecasted time periods. The computation is:
ME = Σ i = 1 to N [Di - Fi] / N
2. Mean Absolute Percentage Error. The next-simplest forecast accuracy measure
measures accuracy (as opposed to bias). Averages the absolute value of the difference
between true demand and associated forecast (divided by true demand) for the forecasted
time periods. The computation is:
MAPE = Σ i = 1 to N ABS[(Di - Fi) / Di] / N
3. Mean Squared Error. Probably the most important error measurement; it averages the
squared differences between true demand and associated forecast for the forecasted time
periods. Squaring gives more weight to large errors (which are the ones we want to avoid;
small errors can be tolerated). Large errors surprise us and make life and planning much
more difficult. The computation is:
MSE = Σ i = 1 to N (Di - Fi)2
/ N
Application of VBA/Excel
Visual BASIC for Applications (VBA) is a programming language that is attached to all Microsoft
applications, including Excel. VBA programming was used in this project to optimize the Simple
Moving Average span (i.e., N, the number of demand values used to compute the Moving
Average), as well as to optimize the weight parameter, α, that is the controlling part of
Exponential Smoothing. The VBA program applied trial span and alpha values and evaluated
each value according to the following selection criterion:
Minimize [1 * ABS(ME) + 2 * MAPE + 3 * MSE]
The above criterion applies judgment-based weights (i.e., the "1," "2," and "3" in the above
expression) that emphasize the relative importance of the Mean Error (ME), the Mean Absolute
2 / 6
Percent Error (MAPE), and the Mean Squared Error (MSE) in determining the Simple Moving
Average and Exponential Moving Average ability to forecast accurately. Span and alpha values
were then subjected to visual inspection for each of the four products (via graphed data). The
spans and alphas were visually acceptable from the following perspectives:
1. Minimized bias.
2. Minimized error.
3. Trend capture.
4. Responsiveness to non-random fluctuation.
Forecasting Decisions
The optimum forecasting method was decided on the basis of the smallest Mean Squared Error
(MSE) value. The standard deviation (s) was computed from the MSE by taking the square root
of the MSE; s was then used to determine the forecasting confidence interval. The confidence
interval computations are:
LCL = Fi – si * 1.64 (where 1.64 is the 90-percent confidence normal-distribution z-score)
UCL = Fi + si * 1.64
The 90-percent confidence z-score was selected on the basis of a visually-suitable width
confidence interval (i.e., a judgment was applied in selecting the 90-percent interval).
The following table provides the summary monthly data on which the best forecasting methods
for each product are based. Again, the criterion of relevance used to determine the best
forecasting method is the smallest Mean Squared Error value.
MONTHLY DATA ACCURACY ASSESSMENT
Product A Product B Product C Product D
Method E MAPE MSE E MAPE MSE E MAPE MSE E MAPE MSE
NM -2,040 16.8% 90,837,443 -49 23.4% 2,887,528 9 12.7% 1337 -41 20.0% 70,472,694
CMF -9,720 30.6% 206,087,159 94 15.5% 1,418,775 71 32.8% 8635 4,402 29.0% 138,201,620
MA -4,252 17.4% 68,944,942 -163 15.6% 968,698 13 11.2% 1146 -607 21.2% 66,255,361
EMA -2,888 16.4% 75,744,998 201 15.3% 1,296,510 13 11.2% 1118 -46 18.8% 64,604,584
Best Method NM EMA MA NM EMA MA NM MA EMA NM EMA EMA
The following table provides the next-future month forecasts for each product. The Mean is the
forecast, and the Lower-Control and Upper-Control Limits provide the 90-percent probable
extreme-value range for the next-future month demand. Note that Sales and Operations
Planning (S&OP) can be brought to bear on critical products, with an informed team selecting
the “best” forecast from within the 90-percent confidence interval; i.e., from within the estimated
future demand range: [LCL, UCL].
3 / 6
MSE-BASED MONTHLY DATA FORECAST
Product
A
Product
B
Product
C
Product
D
Mean 28,908 4,936 296 20,187
UCL 42,526 6,550 351 33,369
LCL 15,291 3,322 241 7,005
Sy/x 8,303 984 33 8,038
n 24 24 26 26
z90% 1.64 1.64 1.64 1.64
Finally, the following monthly graphs provide visual aids of the selected forecasting methods and
of the applied confidence interval statistics. True demand values are represented via blue line;
forecast values are represented via red line. Note that each chart’s present-time (zero) value is
at its right side; past-time (negative) values are to the left.
PRODUCT A (3m SMA) w/ 90% Confidence Interval
10,094
20,094
30,094
40,094
50,094
60,094
70,094
80,094
90,094
100,094
-26
-24
-22
-20
-18
-16
-14
-12
-10
-8
-6
-4
-2
0
Months (0 is Present Time)
Demand
PRODUCT A MA_A LCL UCL
4 / 6
PRODUCT B (3mSMA) w/ 90% Confidence Interval
3,094
4,094
5,094
6,094
7,094
8,094
9,094
-26
-24
-22
-20
-18
-16
-14
-12
-10
-8
-6
-4
-2
0
Months (0 is Present Time)
Demand
PRODUCT B MA_B LCL UCL
PRODUCT C (EMA a = 0.672) w/ 90% Confidence Interval
0
50
100
150
200
250
300
350
400
450
-26
-24
-22
-20
-18
-16
-14
-12
-10
-8
-6
-4
-2
0
Months (0 is Present Time)
Demand
PRODUCT C EMA_C LCL UCL
5 / 6
PRODUCT D (EMA a = 0.639) w/ 90% Confidence Interval
50
10,050
20,050
30,050
40,050
50,050
60,050
70,050
-26
-24
-22
-20
-18
-16
-14
-12
-10
-8
-6
-4
-2
0
Months (0 is Present Time)
Demand
PRODUCT D EMA_D LCL UCL
6 / 6
PRODUCT D (EMA a = 0.639) w/ 90% Confidence Interval
50
10,050
20,050
30,050
40,050
50,050
60,050
70,050
-26
-24
-22
-20
-18
-16
-14
-12
-10
-8
-6
-4
-2
0
Months (0 is Present Time)
Demand
PRODUCT D EMA_D LCL UCL
6 / 6

Weitere ähnliche Inhalte

Was ist angesagt?

Week 4 forecasting - time series - smoothing and decomposition - m.awaluddin.t
Week 4   forecasting - time series - smoothing and decomposition - m.awaluddin.tWeek 4   forecasting - time series - smoothing and decomposition - m.awaluddin.t
Week 4 forecasting - time series - smoothing and decomposition - m.awaluddin.tMaling Senk
 
Quotes on quality management
Quotes on quality managementQuotes on quality management
Quotes on quality managementselinasimpson331
 
Food quality management system
Food quality management systemFood quality management system
Food quality management systemselinasimpson2301
 
Project quality management tools
Project quality management toolsProject quality management tools
Project quality management toolsselinasimpson2901
 
IRJET- Overview of Forecasting Techniques
IRJET- Overview of Forecasting TechniquesIRJET- Overview of Forecasting Techniques
IRJET- Overview of Forecasting TechniquesIRJET Journal
 
Quality software project management
Quality software project managementQuality software project management
Quality software project managementselinasimpson1601
 
Healthcare quality management
Healthcare quality managementHealthcare quality management
Healthcare quality managementselinasimpson0601
 
Forecasting of electric consumption in a semiconductor plant using time serie...
Forecasting of electric consumption in a semiconductor plant using time serie...Forecasting of electric consumption in a semiconductor plant using time serie...
Forecasting of electric consumption in a semiconductor plant using time serie...Alexander Decker
 
A NOVEL PERFORMANCE MEASURE FOR MACHINE LEARNING CLASSIFICATION
A NOVEL PERFORMANCE MEASURE FOR MACHINE LEARNING CLASSIFICATIONA NOVEL PERFORMANCE MEASURE FOR MACHINE LEARNING CLASSIFICATION
A NOVEL PERFORMANCE MEASURE FOR MACHINE LEARNING CLASSIFICATIONIJMIT JOURNAL
 
A Novel Performance Measure for Machine Learning Classification
A Novel Performance Measure for Machine Learning ClassificationA Novel Performance Measure for Machine Learning Classification
A Novel Performance Measure for Machine Learning ClassificationIJMIT JOURNAL
 
Holtwinters terakhir lengkap
Holtwinters terakhir lengkapHoltwinters terakhir lengkap
Holtwinters terakhir lengkapZulyy Astutik
 
Statistical-Process-Control-Analysis-Unraveled_updated210
Statistical-Process-Control-Analysis-Unraveled_updated210Statistical-Process-Control-Analysis-Unraveled_updated210
Statistical-Process-Control-Analysis-Unraveled_updated210pbaxter
 
!!!!!!!!!!!!!!!!!!!!!!!!Optimal combinationofrealizedvolatilityestimators!!!!...
!!!!!!!!!!!!!!!!!!!!!!!!Optimal combinationofrealizedvolatilityestimators!!!!...!!!!!!!!!!!!!!!!!!!!!!!!Optimal combinationofrealizedvolatilityestimators!!!!...
!!!!!!!!!!!!!!!!!!!!!!!!Optimal combinationofrealizedvolatilityestimators!!!!...pace130557
 
Applying quality management in healthcare
Applying quality management in healthcareApplying quality management in healthcare
Applying quality management in healthcareselinasimpson1601
 
IRJET- GDP Forecast for India using Mixed Data Sampling Technique
IRJET- GDP Forecast for India using Mixed Data Sampling TechniqueIRJET- GDP Forecast for India using Mixed Data Sampling Technique
IRJET- GDP Forecast for India using Mixed Data Sampling TechniqueIRJET Journal
 

Was ist angesagt? (20)

Week 4 forecasting - time series - smoothing and decomposition - m.awaluddin.t
Week 4   forecasting - time series - smoothing and decomposition - m.awaluddin.tWeek 4   forecasting - time series - smoothing and decomposition - m.awaluddin.t
Week 4 forecasting - time series - smoothing and decomposition - m.awaluddin.t
 
Quotes on quality management
Quotes on quality managementQuotes on quality management
Quotes on quality management
 
Food quality management system
Food quality management systemFood quality management system
Food quality management system
 
Dt33726730
Dt33726730Dt33726730
Dt33726730
 
Quality management books
Quality management booksQuality management books
Quality management books
 
Project quality management tools
Project quality management toolsProject quality management tools
Project quality management tools
 
Forecasting
ForecastingForecasting
Forecasting
 
IRJET- Overview of Forecasting Techniques
IRJET- Overview of Forecasting TechniquesIRJET- Overview of Forecasting Techniques
IRJET- Overview of Forecasting Techniques
 
Quality software project management
Quality software project managementQuality software project management
Quality software project management
 
Healthcare quality management
Healthcare quality managementHealthcare quality management
Healthcare quality management
 
Forecasting of electric consumption in a semiconductor plant using time serie...
Forecasting of electric consumption in a semiconductor plant using time serie...Forecasting of electric consumption in a semiconductor plant using time serie...
Forecasting of electric consumption in a semiconductor plant using time serie...
 
Quality management wiki
Quality management wikiQuality management wiki
Quality management wiki
 
A NOVEL PERFORMANCE MEASURE FOR MACHINE LEARNING CLASSIFICATION
A NOVEL PERFORMANCE MEASURE FOR MACHINE LEARNING CLASSIFICATIONA NOVEL PERFORMANCE MEASURE FOR MACHINE LEARNING CLASSIFICATION
A NOVEL PERFORMANCE MEASURE FOR MACHINE LEARNING CLASSIFICATION
 
A Novel Performance Measure for Machine Learning Classification
A Novel Performance Measure for Machine Learning ClassificationA Novel Performance Measure for Machine Learning Classification
A Novel Performance Measure for Machine Learning Classification
 
Holtwinters terakhir lengkap
Holtwinters terakhir lengkapHoltwinters terakhir lengkap
Holtwinters terakhir lengkap
 
Statistical-Process-Control-Analysis-Unraveled_updated210
Statistical-Process-Control-Analysis-Unraveled_updated210Statistical-Process-Control-Analysis-Unraveled_updated210
Statistical-Process-Control-Analysis-Unraveled_updated210
 
Juran quality management
Juran quality managementJuran quality management
Juran quality management
 
!!!!!!!!!!!!!!!!!!!!!!!!Optimal combinationofrealizedvolatilityestimators!!!!...
!!!!!!!!!!!!!!!!!!!!!!!!Optimal combinationofrealizedvolatilityestimators!!!!...!!!!!!!!!!!!!!!!!!!!!!!!Optimal combinationofrealizedvolatilityestimators!!!!...
!!!!!!!!!!!!!!!!!!!!!!!!Optimal combinationofrealizedvolatilityestimators!!!!...
 
Applying quality management in healthcare
Applying quality management in healthcareApplying quality management in healthcare
Applying quality management in healthcare
 
IRJET- GDP Forecast for India using Mixed Data Sampling Technique
IRJET- GDP Forecast for India using Mixed Data Sampling TechniqueIRJET- GDP Forecast for India using Mixed Data Sampling Technique
IRJET- GDP Forecast for India using Mixed Data Sampling Technique
 

Andere mochten auch

PARA QUE SEPAS QUIÉN TE CANTA POR LAS MAÑANAS
PARA QUE SEPAS QUIÉN TE CANTA POR LAS MAÑANASPARA QUE SEPAS QUIÉN TE CANTA POR LAS MAÑANAS
PARA QUE SEPAS QUIÉN TE CANTA POR LAS MAÑANASdegatitos
 
Diploma a pachi
Diploma a pachiDiploma a pachi
Diploma a pachidegatitos
 
Barga Data Science lecture 2
Barga Data Science lecture 2Barga Data Science lecture 2
Barga Data Science lecture 2Roger Barga
 
Lo Stretto digitale - 0
Lo Stretto digitale - 0Lo Stretto digitale - 0
Lo Stretto digitale - 0anucita
 
Barga Data Science lecture 9
Barga Data Science lecture 9Barga Data Science lecture 9
Barga Data Science lecture 9Roger Barga
 
Barga DIDC'14 Invited Talk
Barga DIDC'14 Invited TalkBarga DIDC'14 Invited Talk
Barga DIDC'14 Invited TalkRoger Barga
 
Отчет куратора Школы, 2015 год
Отчет куратора Школы, 2015 годОтчет куратора Школы, 2015 год
Отчет куратора Школы, 2015 годnizhgma.ru
 
Francesco Micali : Dal sito internet al network diocesano - Mediabeta srl
Francesco Micali : Dal sito internet al network diocesano - Mediabeta srlFrancesco Micali : Dal sito internet al network diocesano - Mediabeta srl
Francesco Micali : Dal sito internet al network diocesano - Mediabeta srlf.micali
 
Data Driven Engineering 2014
Data Driven Engineering 2014Data Driven Engineering 2014
Data Driven Engineering 2014Roger Barga
 

Andere mochten auch (12)

PARA QUE SEPAS QUIÉN TE CANTA POR LAS MAÑANAS
PARA QUE SEPAS QUIÉN TE CANTA POR LAS MAÑANASPARA QUE SEPAS QUIÉN TE CANTA POR LAS MAÑANAS
PARA QUE SEPAS QUIÉN TE CANTA POR LAS MAÑANAS
 
Diploma a pachi
Diploma a pachiDiploma a pachi
Diploma a pachi
 
Barga Data Science lecture 2
Barga Data Science lecture 2Barga Data Science lecture 2
Barga Data Science lecture 2
 
Lo Stretto digitale - 0
Lo Stretto digitale - 0Lo Stretto digitale - 0
Lo Stretto digitale - 0
 
Barga Data Science lecture 9
Barga Data Science lecture 9Barga Data Science lecture 9
Barga Data Science lecture 9
 
Barga DIDC'14 Invited Talk
Barga DIDC'14 Invited TalkBarga DIDC'14 Invited Talk
Barga DIDC'14 Invited Talk
 
Отчет куратора Школы, 2015 год
Отчет куратора Школы, 2015 годОтчет куратора Школы, 2015 год
Отчет куратора Школы, 2015 год
 
Francesco Micali : Dal sito internet al network diocesano - Mediabeta srl
Francesco Micali : Dal sito internet al network diocesano - Mediabeta srlFrancesco Micali : Dal sito internet al network diocesano - Mediabeta srl
Francesco Micali : Dal sito internet al network diocesano - Mediabeta srl
 
Data Driven Engineering 2014
Data Driven Engineering 2014Data Driven Engineering 2014
Data Driven Engineering 2014
 
Present progressive
Present progressivePresent progressive
Present progressive
 
World Cancer Day Awareness In UAE, 2017
World Cancer Day Awareness In UAE, 2017World Cancer Day Awareness In UAE, 2017
World Cancer Day Awareness In UAE, 2017
 
Proyecto final
Proyecto final Proyecto final
Proyecto final
 

Ähnlich wie Supply Chain Planning Paper

ForecastingBUS255 GoalsBy the end of this chapter, y.docx
ForecastingBUS255 GoalsBy the end of this chapter, y.docxForecastingBUS255 GoalsBy the end of this chapter, y.docx
ForecastingBUS255 GoalsBy the end of this chapter, y.docxbudbarber38650
 
Demand forecasting methods 1 gp
Demand forecasting methods 1 gpDemand forecasting methods 1 gp
Demand forecasting methods 1 gpPUTTU GURU PRASAD
 
Forecasting
ForecastingForecasting
Forecasting3abooodi
 
Sheet1HCM3002 Economics of HealthcareChoose a Health Plan (Part 2).docx
Sheet1HCM3002 Economics of HealthcareChoose a Health Plan (Part 2).docxSheet1HCM3002 Economics of HealthcareChoose a Health Plan (Part 2).docx
Sheet1HCM3002 Economics of HealthcareChoose a Health Plan (Part 2).docxmaoanderton
 
IEOR 265 Final Paper_Minchao Lin
IEOR 265 Final Paper_Minchao LinIEOR 265 Final Paper_Minchao Lin
IEOR 265 Final Paper_Minchao LinMinchao Lin
 
Demand Estimation Model Sensitivity Analysis
Demand Estimation Model Sensitivity AnalysisDemand Estimation Model Sensitivity Analysis
Demand Estimation Model Sensitivity AnalysisGary Lauson, M.S., P.E.
 
Marketing Engineering Notes
Marketing Engineering NotesMarketing Engineering Notes
Marketing Engineering NotesFelipe Affonso
 
Determining Measurement Uncertainty Parameters for Calibration Processes
Determining Measurement Uncertainty Parameters for Calibration ProcessesDetermining Measurement Uncertainty Parameters for Calibration Processes
Determining Measurement Uncertainty Parameters for Calibration Processestheijes
 
Operations Management VTU BE Mechanical 2015 Solved paper
Operations Management VTU BE Mechanical 2015 Solved paperOperations Management VTU BE Mechanical 2015 Solved paper
Operations Management VTU BE Mechanical 2015 Solved paperSomashekar S.M
 
A Comparative Study for Anomaly Detection in Data Mining
A Comparative Study for Anomaly Detection in Data MiningA Comparative Study for Anomaly Detection in Data Mining
A Comparative Study for Anomaly Detection in Data MiningIRJET Journal
 
3122999029.pptx
3122999029.pptx3122999029.pptx
3122999029.pptxMyoZawOo10
 
IRJET- Error Reduction in Data Prediction using Least Square Regression Method
IRJET- Error Reduction in Data Prediction using Least Square Regression MethodIRJET- Error Reduction in Data Prediction using Least Square Regression Method
IRJET- Error Reduction in Data Prediction using Least Square Regression MethodIRJET Journal
 
Pain Procedure Valuation as a Competitive Strategy
Pain Procedure Valuation as a Competitive StrategyPain Procedure Valuation as a Competitive Strategy
Pain Procedure Valuation as a Competitive StrategyJAMES B. MACON, M.D.
 
Case StudiesMemorial HospitalMemorial Hospital is a privately .docx
Case StudiesMemorial HospitalMemorial Hospital is a privately .docxCase StudiesMemorial HospitalMemorial Hospital is a privately .docx
Case StudiesMemorial HospitalMemorial Hospital is a privately .docxtidwellveronique
 
Industrial engineering unit 2 new
Industrial engineering unit 2 newIndustrial engineering unit 2 new
Industrial engineering unit 2 newmohsinmmm
 
MovingAverage (2).pptx
MovingAverage (2).pptxMovingAverage (2).pptx
MovingAverage (2).pptxbrahimNasibov
 
Operations management forecasting
Operations management   forecastingOperations management   forecasting
Operations management forecastingTwinkle Constantino
 
International journal of applied sciences and innovation vol 2015 - no 2 - ...
International journal of applied sciences and innovation   vol 2015 - no 2 - ...International journal of applied sciences and innovation   vol 2015 - no 2 - ...
International journal of applied sciences and innovation vol 2015 - no 2 - ...sophiabelthome
 

Ähnlich wie Supply Chain Planning Paper (20)

ForecastingBUS255 GoalsBy the end of this chapter, y.docx
ForecastingBUS255 GoalsBy the end of this chapter, y.docxForecastingBUS255 GoalsBy the end of this chapter, y.docx
ForecastingBUS255 GoalsBy the end of this chapter, y.docx
 
Forecasting
ForecastingForecasting
Forecasting
 
Demand forecasting methods 1 gp
Demand forecasting methods 1 gpDemand forecasting methods 1 gp
Demand forecasting methods 1 gp
 
Forecasting
ForecastingForecasting
Forecasting
 
Sheet1HCM3002 Economics of HealthcareChoose a Health Plan (Part 2).docx
Sheet1HCM3002 Economics of HealthcareChoose a Health Plan (Part 2).docxSheet1HCM3002 Economics of HealthcareChoose a Health Plan (Part 2).docx
Sheet1HCM3002 Economics of HealthcareChoose a Health Plan (Part 2).docx
 
IEOR 265 Final Paper_Minchao Lin
IEOR 265 Final Paper_Minchao LinIEOR 265 Final Paper_Minchao Lin
IEOR 265 Final Paper_Minchao Lin
 
Demand Estimation Model Sensitivity Analysis
Demand Estimation Model Sensitivity AnalysisDemand Estimation Model Sensitivity Analysis
Demand Estimation Model Sensitivity Analysis
 
Marketing Engineering Notes
Marketing Engineering NotesMarketing Engineering Notes
Marketing Engineering Notes
 
Determining Measurement Uncertainty Parameters for Calibration Processes
Determining Measurement Uncertainty Parameters for Calibration ProcessesDetermining Measurement Uncertainty Parameters for Calibration Processes
Determining Measurement Uncertainty Parameters for Calibration Processes
 
Operations Management VTU BE Mechanical 2015 Solved paper
Operations Management VTU BE Mechanical 2015 Solved paperOperations Management VTU BE Mechanical 2015 Solved paper
Operations Management VTU BE Mechanical 2015 Solved paper
 
A Comparative Study for Anomaly Detection in Data Mining
A Comparative Study for Anomaly Detection in Data MiningA Comparative Study for Anomaly Detection in Data Mining
A Comparative Study for Anomaly Detection in Data Mining
 
3122999029.pptx
3122999029.pptx3122999029.pptx
3122999029.pptx
 
IRJET- Error Reduction in Data Prediction using Least Square Regression Method
IRJET- Error Reduction in Data Prediction using Least Square Regression MethodIRJET- Error Reduction in Data Prediction using Least Square Regression Method
IRJET- Error Reduction in Data Prediction using Least Square Regression Method
 
Pain Procedure Valuation as a Competitive Strategy
Pain Procedure Valuation as a Competitive StrategyPain Procedure Valuation as a Competitive Strategy
Pain Procedure Valuation as a Competitive Strategy
 
Case StudiesMemorial HospitalMemorial Hospital is a privately .docx
Case StudiesMemorial HospitalMemorial Hospital is a privately .docxCase StudiesMemorial HospitalMemorial Hospital is a privately .docx
Case StudiesMemorial HospitalMemorial Hospital is a privately .docx
 
Industrial engineering unit 2 new
Industrial engineering unit 2 newIndustrial engineering unit 2 new
Industrial engineering unit 2 new
 
Chapter-3_Heizer_S1.pptx
Chapter-3_Heizer_S1.pptxChapter-3_Heizer_S1.pptx
Chapter-3_Heizer_S1.pptx
 
MovingAverage (2).pptx
MovingAverage (2).pptxMovingAverage (2).pptx
MovingAverage (2).pptx
 
Operations management forecasting
Operations management   forecastingOperations management   forecasting
Operations management forecasting
 
International journal of applied sciences and innovation vol 2015 - no 2 - ...
International journal of applied sciences and innovation   vol 2015 - no 2 - ...International journal of applied sciences and innovation   vol 2015 - no 2 - ...
International journal of applied sciences and innovation vol 2015 - no 2 - ...
 

Supply Chain Planning Paper

  • 1. SUPPLY CHAIN PLANNING DEMAND ANALYSIS PROJECT G. Lauson, M.S., P.E. January 31, 2017 Purpose This paper presents the findings of an analysis of four (4) pharmaceutical products. Four (4) forecasting methods were applied and three (3) error measurements were applied to those forecasting methods to determine which method most accurately estimated next-future-month demand for each product. Monthly data was selected at the request of the manufacturing division, as this was determined by them as optimum for their production schedule planning. Note that all forecasting and error methods / descriptions were taken from a Coursera / Rutgers University online course titled “Supply Chain Planning” which the author completed in January 2017. The course is part of a five-course offering within Coursera’s “Supply Chain Management Specialization.” Forecasting Methods The following forecasting methods were evaluated for accuracy with respect to each of the four pharmaceutical products. A brief explanation of the rationale behind each method is provided, along with how the method is computed. 1. Naïve Method. The simplest of the four, this method uses the prior-period actual demand as the forecast for the next-period demand. The computation is: Ft = Dt-1 2. Cumulative Mean Method. This method averages all demand from prior periods – that average becomes the forecast. Based on information provided in the above-referenced Coursera course, this is a very stable forecast that averages out all noise; however, that stability may be contrary to current demand conditions. The assumption behind this method is that “all prior data is equally useful.” The computation is: Ft = Σ i = 1 to t–1 [Di] / (t – 1) 3. Simple Moving Average (SMA) Method. An adaptable forecasting method that can be made reactive or stable. Companies with stable demand tend to like this indicator. The computation is: Ft = Σ i = (t-N+2) to t–1 [Di] / N 4. Exponential Smoothing (or Exponential Moving Average (EMA)) Method. A weighted
  • 2. average of all past demand; most of the weight is on the latest observed period and the remainder of the weights decline exponentially. α changes the weights and varies how much weight (in the weighted average) is applied to each period. The computation is: Ft = α * [Dt-1] + (1 – α) * Ft-1 Error Estimation Methods The following methods were used to estimate the error associated with each of the above forecasting methods. 1. Mean Error. This is the simplest forecast accuracy measure, and is more a measure of bias than accuracy. This error measurement averages the difference between true demand and associated forecast for the forecasted time periods. The computation is: ME = Σ i = 1 to N [Di - Fi] / N 2. Mean Absolute Percentage Error. The next-simplest forecast accuracy measure measures accuracy (as opposed to bias). Averages the absolute value of the difference between true demand and associated forecast (divided by true demand) for the forecasted time periods. The computation is: MAPE = Σ i = 1 to N ABS[(Di - Fi) / Di] / N 3. Mean Squared Error. Probably the most important error measurement; it averages the squared differences between true demand and associated forecast for the forecasted time periods. Squaring gives more weight to large errors (which are the ones we want to avoid; small errors can be tolerated). Large errors surprise us and make life and planning much more difficult. The computation is: MSE = Σ i = 1 to N (Di - Fi)2 / N Application of VBA/Excel Visual BASIC for Applications (VBA) is a programming language that is attached to all Microsoft applications, including Excel. VBA programming was used in this project to optimize the Simple Moving Average span (i.e., N, the number of demand values used to compute the Moving Average), as well as to optimize the weight parameter, α, that is the controlling part of Exponential Smoothing. The VBA program applied trial span and alpha values and evaluated each value according to the following selection criterion: Minimize [1 * ABS(ME) + 2 * MAPE + 3 * MSE] The above criterion applies judgment-based weights (i.e., the "1," "2," and "3" in the above expression) that emphasize the relative importance of the Mean Error (ME), the Mean Absolute 2 / 6
  • 3. Percent Error (MAPE), and the Mean Squared Error (MSE) in determining the Simple Moving Average and Exponential Moving Average ability to forecast accurately. Span and alpha values were then subjected to visual inspection for each of the four products (via graphed data). The spans and alphas were visually acceptable from the following perspectives: 1. Minimized bias. 2. Minimized error. 3. Trend capture. 4. Responsiveness to non-random fluctuation. Forecasting Decisions The optimum forecasting method was decided on the basis of the smallest Mean Squared Error (MSE) value. The standard deviation (s) was computed from the MSE by taking the square root of the MSE; s was then used to determine the forecasting confidence interval. The confidence interval computations are: LCL = Fi – si * 1.64 (where 1.64 is the 90-percent confidence normal-distribution z-score) UCL = Fi + si * 1.64 The 90-percent confidence z-score was selected on the basis of a visually-suitable width confidence interval (i.e., a judgment was applied in selecting the 90-percent interval). The following table provides the summary monthly data on which the best forecasting methods for each product are based. Again, the criterion of relevance used to determine the best forecasting method is the smallest Mean Squared Error value. MONTHLY DATA ACCURACY ASSESSMENT Product A Product B Product C Product D Method E MAPE MSE E MAPE MSE E MAPE MSE E MAPE MSE NM -2,040 16.8% 90,837,443 -49 23.4% 2,887,528 9 12.7% 1337 -41 20.0% 70,472,694 CMF -9,720 30.6% 206,087,159 94 15.5% 1,418,775 71 32.8% 8635 4,402 29.0% 138,201,620 MA -4,252 17.4% 68,944,942 -163 15.6% 968,698 13 11.2% 1146 -607 21.2% 66,255,361 EMA -2,888 16.4% 75,744,998 201 15.3% 1,296,510 13 11.2% 1118 -46 18.8% 64,604,584 Best Method NM EMA MA NM EMA MA NM MA EMA NM EMA EMA The following table provides the next-future month forecasts for each product. The Mean is the forecast, and the Lower-Control and Upper-Control Limits provide the 90-percent probable extreme-value range for the next-future month demand. Note that Sales and Operations Planning (S&OP) can be brought to bear on critical products, with an informed team selecting the “best” forecast from within the 90-percent confidence interval; i.e., from within the estimated future demand range: [LCL, UCL]. 3 / 6
  • 4. MSE-BASED MONTHLY DATA FORECAST Product A Product B Product C Product D Mean 28,908 4,936 296 20,187 UCL 42,526 6,550 351 33,369 LCL 15,291 3,322 241 7,005 Sy/x 8,303 984 33 8,038 n 24 24 26 26 z90% 1.64 1.64 1.64 1.64 Finally, the following monthly graphs provide visual aids of the selected forecasting methods and of the applied confidence interval statistics. True demand values are represented via blue line; forecast values are represented via red line. Note that each chart’s present-time (zero) value is at its right side; past-time (negative) values are to the left. PRODUCT A (3m SMA) w/ 90% Confidence Interval 10,094 20,094 30,094 40,094 50,094 60,094 70,094 80,094 90,094 100,094 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 Months (0 is Present Time) Demand PRODUCT A MA_A LCL UCL 4 / 6
  • 5. PRODUCT B (3mSMA) w/ 90% Confidence Interval 3,094 4,094 5,094 6,094 7,094 8,094 9,094 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 Months (0 is Present Time) Demand PRODUCT B MA_B LCL UCL PRODUCT C (EMA a = 0.672) w/ 90% Confidence Interval 0 50 100 150 200 250 300 350 400 450 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 Months (0 is Present Time) Demand PRODUCT C EMA_C LCL UCL 5 / 6
  • 6. PRODUCT D (EMA a = 0.639) w/ 90% Confidence Interval 50 10,050 20,050 30,050 40,050 50,050 60,050 70,050 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 Months (0 is Present Time) Demand PRODUCT D EMA_D LCL UCL 6 / 6
  • 7. PRODUCT D (EMA a = 0.639) w/ 90% Confidence Interval 50 10,050 20,050 30,050 40,050 50,050 60,050 70,050 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 Months (0 is Present Time) Demand PRODUCT D EMA_D LCL UCL 6 / 6