The Future of Forecasting MI MIT SCM M Capst ston one Proje [PDF]

SLIDE 1

The Future of Forecasting

MI MIT SCM M Capst ston

ne Proje

ject ct

Evan Humphrey | Federico Laiño Advisor isor | Inma Borrella

SLIDE 2 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Agenda

2

1

Introduction

1. Company Background 2. Motivation 3. Objective and Scope

2

Methodology

1. Overview 2. Discovery 3. Diagnosis 4. Demand Sensing Approaches

3

Findings

1. Characterization of Demand 2. Current Forecasting Process 3. Forecast Accuracy Analysis 4. Suggestions for Implementing Demand Sensing

4

Conclusions

1. Takeaways 2. Future work

SLIDE 3

Introduction

1

SLIDE 4 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Company Background

4

1960s

Company ny Foun unde ded

Originally ‘Frontier Contact Lenses’ from Buffalo, New York. Later moved to Jacksonville, Florida 1970s

Develope loped d Etafil ilcon

n

Chief Optometrist develops new material, Etafilcon, that allowed production of soft lenses. 1980s

Acquired ed by J&J

Division was renamed to ‘Vistakon’. Developed automated production system, leading to the creation

f the Acuvue brand.

1990s

1 Day Lenses

Created first low-cost, daily disposable lens. Expanded globally to Brazil, Japan, Singapore and UK. Changed name to JJVC. 2000+

Market Leader

JJVC gains and maintains leadership in the contact lens market.

SLIDE 5 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Company Background

5 3% of revenue 1% of volume

LATAM

38% of revenue 32% volume

US and Canada

15% of revenue 16% of volume

Asia Pacific

21% of revenue 26% volume

Japan

21% of revenue 22% volume

Europe, Middle East and Africa

Production + DC

Jacksonville, FL

DC

EVC – London, UK

DC

HDC - Tokyo

$3Bn Business 4Bn Lenses 22.000 SKUs

Key Insights

Production

Limerick, Ireland

SLIDE 6 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño

We must constantly strive to reduce our cost in order to maintain reasonable prices. Customers' orders must be serviced promptly and accurately.

Lines 3 and 4 of the J&J Credo

“

6

SLIDE 7 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Motivation

7

Market Context

Contact lens global leader by market share but faces competition from other large companies and disruptive entrants.

Cost Efficiency

Driven to continuously improve forecast accuracy and capitalize

n lower inventory costs and

higher service levels.

Production Capacity

Owns high-end manufacturing lines that are at near-maximum utilization with expansion requiring considerable CAPEX and time.

Forecast Accuracy

Wants to explore the potential of demand sensing as a means to improve forecast accuracy

SLIDE 8 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Objective and Scope

8

Brand

Acuvue

Production Facility

Jacksonville, FL

Region

Continental United States

Brand Family

1-Day Moist (1DM) 1-Day Moist for Astigmatism (1DM-A)

Production + DC

Jacksonville, FL Analyze the current J&J Vision Care forecasting process and propose suggestions for improvement, paying special attention to Demand Sensing approaches.

Scope Objective

SLIDE 9 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño

Methodology

2

SLIDE 10 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Overview

10

Sep Oct Nov Dec Jan Feb Mar Apr May Jun

Tasks Phas ase I Objective and Scope Discovery Demand Sensing Approaches Phas ase II Diagnosis Capstone Write-up

SLIDE 11 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Discovery

11

01 02

Discovery

03

Forecasting Demand Planning S&OP

Interview erviews 02

Jacksonville, FL Production Facility Distribution Center S&OP Interviews

Site Visit it 03

SCM History Forecasting Techniques Forecasting Measures Demand Sensing Case Studies

Litera rature ture Review ew 01

SLIDE 12 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Diagnosis

12

1 2 3

Demand Characterization

Pareto Analysis Time Series Distribution Statistics

Forecasting Process Mapping Forecast Accuracy Analysis

Cycle Time Framework Data Inputs Forecasts Pareto Breakdown Comparison with Alternative Forecasts

SLIDE 13 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Demand Sensing Approaches

13

Record and measure the impact of demand shaping events such as promotions, price changes, product launches and forward-buy arrangements.

Meas asurin ring the Impac act t of Demand Shaping Actio ions 03

Reduce cycle time between forecasts to take advantage of latest demand information updates.

Latenc tency Reduct ction 01

Include downstream supply chain data, such as POS data, in the demand forecasting model.

Downstream stream Data a Integ egrat ratio ion 02

SLIDE 14 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño

Findings

3

SLIDE 15 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Characterization of Demand

15

SLIDE 16 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño

Characterization of Demand

Shipments Time Series by Quarters

16

1 Day Mois

ist 90-Pack

1 Day Mois

ist for Asti

tigm gmati tism 90-Pac ack 1 Day Mois

ist 30-Pack

1 Day Mois

ist for Asti

tigm gmati tism 30 30-Pac ack

SLIDE 17 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño

Characterization of Demand

Pareto Curves

17 % of Total SKUs 0% 100% 10% 30% 40% 50% 60% 70% 80% 90% 20% % of Total SKUs 0% 100% 10% 30% 40% 50% 60% 70% 80% 90% 20%

1 Day Mois

ist 90-Pack

1 Day Mois

ist for Asti

tigm gmati tism 90-Pac ack

% of Total SKUs 0% 100% 10% 30% 40% 50% 60% 70% 80% 90% 20% % of Total SKUs 0% 100% 10% 30% 40% 50% 60% 70% 80% 90% 20%

1 Day Mois

ist 30-Pack

1 Day Mois

ist for Asti

tigm gmati tism 30 30-Pac ack 116 SKUs 1528 SKUs

SLIDE 18 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño

Characterization of Demand

Mean Shipments vs Coefficient of Variation for each Pareto segment

18

SLIDE 19 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Forecast Accuracy Analysis

19 — Bias — MAPE — MAPV — PVE — RMSE — 3-Month Average — 4-Month Average — 5-Month Average — 6-Month Average — Simple Exp. Smoothing — Double Exp. Smoothing — Brand — Pack Size — Pareto to — SKU — Month — Quarters

1

Aggregation Levels

3

Accuracy Metrics

2

Forecasts Compared — J&J Vision

n Care - Statist

stic ical — J&J Vision Care - Lag 03 — J&J Vision Care - Lag 02 — J&J Vision

n Care - Lag 01

01 — Naïve — 2-Mont nth h Average ge

SLIDE 20 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño

Forecast Accuracy Analysis

Naïve vs Lag 01 Forecast Comparison Results

20

SLIDE 21 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño

Forecast Accuracy Analysis

2-Month Average vs Lag 01 Forecast Comparison Results

21

SLIDE 22 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Current Forecasting Process

22

Input Output Process Shipments Data 1 Statistical Forecast Consensus Forecast SKU Level Forecast Executive S&OP 2.2 Production Planning 1.1 JDA Software 1.2 Regional S&OP 2 3 2.1 SKU Breakdown 2 weeks 1-Month Cycle Lag 03 Lag 02 Lag 01

SLIDE 23 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Suggestions for Implementing Demand Sensing

23

Record and measure the impact of demand shaping events such as promotions, price changes, product launches and forward-buy arrangements.

Meas asurin ring the Impac act t of Demand Shaping Actio ions 03

Reduce cycle time between forecasts to take advantage of latest demand information updates.

Latenc tency Reduct ction 01

Include downstream supply chain data, such as POS data, in the demand forecasting model.

Downstream stream Data a Integ egrat ratio ion 02

SLIDE 24 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Suggestions for Implementing Demand Sensing

24

Input Output Process Executive S&OP 3 Shipments Data 1 2.2 Production Planning 1.1 JDA Software 1.2 Regional S&OP 2 2.1 SKU Breakdown 1 week 1-Month Cycle Shipments Data 1 2.2 Product Planning 1.1 JDA Software 1.2 Regional S&OP 2 2.1 SKU Breakdown 1 week

Latenc tency Reduct ction 01

Statistical Forecast Consensus Forecast SKU Level Forecast Statistical Forecast Consensus Forecast SKU Level Forecast

SLIDE 25 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Input Output Process Shipments Data 1 Executive S&OP 2.2 Production Planning 1.1 DS Model 1.2 Regional S&OP 2 3 2.1 SKU Breakdown 2 weeks 1-Month Cycle Statistical Forecast Consensus Forecast SKU Level Forecast Downstream Data

Suggestions for Implementing Demand Sensing

25

Downstream stream Data a Integra egratio tion 02

SLIDE 26 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Input Output Process Shipments Data 1 Executive S&OP 2.2 Production Planning 1.1 DS Model 1.2 Regional S&OP 2 3 2.1 SKU Breakdown 2 weeks 1-Month Cycle Statistical Forecast Consensus Forecast SKU Level Forecast Demand Shaping Data

Suggestions for Implementing Demand Sensing

26

Meas asurin ring the Impac act t of Demand Shaping Actio ions 03

SLIDE 27 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Suggestions for Implementing Demand Sensing

27 Latency Reduction

Challenges

— Cost-benefit tradeoff — Cross-functional coordination — Cost-benefit tradeoff — Systems integration — Data access — No guaranteed benefit in accuracy — Cost-benefit tradeoff — Data structuring — No guaranteed benefit in accuracy

Opportunities

— Simplest solution — Fastest to Implement — Guaranteed improvement in accuracy — Greater accuracy potential — Real time updates — More responsive to change

Downstream Data Integration Demand Shaping Actions

SLIDE 28 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño

Conclusions

4

SLIDE 29 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Takeaways

29

We recommend J&J Vision Care consider the use of simpler forecasting techniques for the 30-Pack pack size category and, more specifically, for the 1-Day Moist 30-Pack product segment.

Fo Forecast cast Accuracy uracy Impro rovemen vements ts

01 We recommend J&J Vision Care consider the Demand Sensing initiatives we provided. Latency Reduction should be implemented first.

Demand and Sens nsing ing Initiat atives ves

02

SLIDE 30 MIT Supply Chain Management Program Evan Humphrey | Federico Laiño I N T R O D U C T I O N M E T H O D O L O G Y F I N D I N G S C O N C L U S I O N S

Future Work

30

Me Measu suring ring the Impac act t of Demand and Shaping ng Actions

— Propose a system to capture demand shaping events in a structured manner. — Measure the impact of past initiatives. — Develop predictive system to forecast future events.

Down wnst strea ream m Data a Integra gration

n

— Collect data at different echelons in the Supply Chain. — Develop predictive system to forecast demand based on variations in downstream supply chain data.

SLIDE 31

31

Thank You

Questions?

Bachelor of Science Biotechnology Indiana University

Evan Humphrey

Bachelor of Science Industrial Engineering Buenos Aires Institute of Technology

Federico Laiño