PowerPoint Presentation

What Drives Value in a Value Chain?

Speakers

Lora Cecere, Founder of Supply Chain Insights

Dave Goldsman School of ISyE Georgia Tech

Disclosure: This work is open source jointly funded by Kinaxis and Supply Chain Insights.

SECTION 1

Introduction

I Am A Social Scientist

Who Is Lora?

2 years Partner at Altimeter Group (leader in open research)

15 Years Leading Teams in Manufacturing and Distribution for Clorox, Kraft/General Foods, Nestle/Dreyers Grand Ice Cream and Procter & Gamble.

Founder of Supply Chain Insights (13 years) “LinkedIn Influencer”, Guest blogger for Forbes, Author: Bricks Matter (2012), Supply Chain Metrics That Matter (2014), and Shaman’s Journal (2014-23)

8 years as an Analyst at Gartner and AMR Research

8 years Experience in Marketing and Selling Supply Chain Software at Descartes Systems Group and Manugistics (now Blue Yonder)

SECTION 2

Background

Our Journey

Wrote the Book Bricks Matter

Published Supply Chains to Admire 2016-2022

First Project with Arizona State 2014

Target End Date Winter 2024

2013

Started Work with Georgia Tech

Published the Supply Chain Index

Wrote Metrics That Matter 2015

Spring 2024

Spring 2015

1996

A Look at the Shifts

Organizational

Value Chain

• Alignment issues grew 3-fold for brand owners. • Focus on efficient supply chains. Lack of recognition of supply chain flows. • Supply chain became a function within a functional organization focused on supply.

• 40-50% of products are no longer forecastable due to distortion of history. • 8 out of 10 companies are degrading the forecast with current practices. • Increase in demand latency. • Growth of bullwhip impact and supply variability.

1 1

Food Industry Operating Margin vs. Inventory Turns (2013 - 2022)

Best Scenario

7.0

2015

2014

2016

2013

2018

2020

2019

6.0

2017

Food Industry Average: Margin: 0.10 Inventory Turns: 6.19

2021

2022

5.0

0.08

0.09

0.10

0.11

Operating Margin Food Industry

◆ Average (Operating Margin, Inventory Turns)

Mondelez Operating Margin vs. Inventory Turns (2013 - 2022)

Best Scenario

2015

7.0

MDLZ 0.14, 6.23

2021

2016

6.5

2017

2014

2019

2022

2020

6.0

2018

2013

5.5

0.11

0.12

0.13

0.14

0.15

0.16

0.17

0.18

Operating Margin Mondelez

◆ Average (Operating Margin, Inventory Turns)

Source: Supply Chain Insights LLC, Corporate Annual Reports 2013-2022 from YCharts

The Industry Struggles with Inventory

SECTION 3

Methodology

Data Processing

• Data sourced from Y-Charts for the period of 1982-2022 • For each company and each year… • 114 dependent and independent factors (performance metrics) to examine • Many of these factors aren’t really independent • 26 Different industry sectors (as defined in Supply Chains to Admire) • Examples: Chemical sector has 37 companies, Food has 31, etc. • Data is laid out well and is very clean

Method

• Pick three industry groups (Chemical, Food, Pharma)

• Consider 40 years of data as a whole and in 10-year increments • Conduct correlation analysis to develop “independent” clusters of the 114 factors for our models

• Settled on 5 independent clusters for now (possibly dependent on industry group, but that’s for later)

• Pick representative factor from each of the 5 clusters

• Perform regression analysis using the 5 factors to model the 3 output quantities one-at-a-time (market cap, fundamental score, price to book ratio). • Actually, for a particular model, e.g., market cap, can also use the other two output quantities as model input factors.

• “Elementary” regression first (to make sure to get things right), then…

• …more -advanced techniques to squeeze out improvements

• Rinse and Repeat for other sectors.

Model Analysis/Reduction

Start

Step 1: Select Predicting Variables

Step 2: Remove Outliers

Methodology:

• Conducted a correlation analysis to identify 'independent' clusters among the 114 factors. • Determined 5 independent clusters for the initial model framework • Run forward stepwise regression to see which predicting variables are insignificant

• Utilize Cook's Distance to pinpoint influential points in the dataset. • Establish a threshold for Cook's Distance ( 𝐷 𝑖 ) such that data points with 𝐷 𝑖 > 4 𝑛 (where 𝑛 is the sample size) are classified as outliers. • Remove identified outliers from the dataset to mitigate the risk of skewed analysis results.

Note: Clusters may vary by industry group, which will be considered in subsequent analyses.

Model Analysis/Reduction

Step 3: Check Predicting Variable Goodness of Fit (Linear)

Step 4: Check for Multicollinearity

Methodology:

Determine R-squared ( 𝑅 2 ) Value: •

• For each predictor, compute the Variance Inflation Factor ( 𝑉𝐼𝐹 ) to quantify the inflation in variance caused by correlations with other predictors. • Establish a VIF threshold using the greater of two values: 10 , or ൗ 1 1−𝑅 2 , to discern significant multicollinearity. • Evaluate the VIF results to confirm that no predicting variables exhibit multicollinearity.

Calculate the proportion of variance in the dependent variable that's predictable from the predicting variable. Examine the residual plots for patterns that indicate deviations from the linearity assumption.

Analyze Residuals: •

Conduct F-test: •

Use the F-test to check the overall significance of the regression model.

Inspect p-values: •

For each predictor, assess the p-value to determine its statistical significance.

Model Analysis/Reduction

Step 6: Modify Dataset to Meet Residual Assumptions

Step 5: Transform Response Variable

Methodology:

• Fitted value standardized residual plot • Remove redundant response variables

Diagnostic Tools Applied: •

QQ plot to assess the normality of residuals. Residual histogram to identify skewness in the data distribution.

•

Transformation Technique: • In cases where the residual histogram indicated right skewness: • Executed a power transformation on the response variable using 𝑦 𝜆 • Determined the optimal 𝜆 , where 0< 𝜆 <1 , to achieve a more normalize the distribution of residuals.

Final Model

First Pass Model Initial Model : ො𝑦 = መ 𝛽 0 + መ 𝛽 1 𝑋 1 +⋯+ መ 𝛽 11 𝑋 11 Where 𝑋 1 ,…, 𝑋 11 are all 10 predicting variables + time

Model Results: • F Statistic P-value: • Adjusted 𝑅 2 :

Residual assumptions violated: • ϵ i ’s are normally distributed with… • … mean 0 and constant variance • ϵ i ’s are independent

Preliminary residual analysis →

Correlation Matrix of the

Selected Elements

Results

Chemical Sector: Predicting Market Capitalization

Dependent Variable

Coefficients

ෝ𝒚

𝑷𝒓𝒆𝒅𝒊𝒄𝒕𝒆𝒅 𝑴𝒂𝒓𝒌𝒆𝒕 𝑪𝒂𝒑

෡ β

4332.031

Independent Variables (Predictors)

෡ β

𝑿 𝟏

Profit Margin Annual

-1.393e+04

෡ β

𝑿 𝟐

Net Income Annual

15.506

෡ β

𝑿 𝟑

Finished Goods Annual

-2.950

෡ β

𝑿 𝟒

Annual Revenue per Employee

-0.003

Key Statistics

෡ β

𝑿 𝟓

Retention Ratio Annual

-5353.538

R-squared

0.70

෡ β T

𝑻

𝑻𝒊𝒎𝒆 (𝒀𝒆𝒂𝒓)

206.4631

F-statistic 122 Jarque-Bera (JB): 27.851 Prob (F-statistic) 0.000

ෝ𝒚 =෡𝜷 𝟎 + ෡𝜷 𝟏 𝑿 𝟏 + ෡𝜷 𝟐 𝑿 𝟐 + ෡𝜷 𝟑 𝑿 𝟑 + ෡𝜷 𝟒 𝑿 𝟒 + ෡𝜷 𝟓 𝑿 𝟓 +෡𝜷 𝑻 𝑻

Food Sector: Predicting Market Capitalization

Dependent Variable

ෝ𝒚

𝑷𝒓𝒆𝒅𝒊𝒄𝒕𝒆𝒅 𝑴𝒂𝒓𝒌𝒆𝒕 𝑪𝒂𝒑

Coefficients

෡ β

2.3619

Independent Variables (Predictors) 𝑿 𝟏 Annual Inventories Net 𝑿 𝟐 Annual Retention Ratio 𝑿 𝟑 Annual Revenue per Employee 𝑿 𝟒 Annual Profit Margin 𝑻 𝑻𝒊𝒎𝒆 (𝒀𝒆𝒂𝒓)

෡ β

0.0002

෡ β

-0.2606

෡ β

-1.412e-08

෡ β

6.5285

෡ β

-5.91e-06

Key Statistics

R-squared

0.70

F-statistic 176.4 Jarque-Bera (JB): 13.51 Prob (F-statistic) 0.000

ෝ𝒚 𝟎.𝟓 = ෡𝜷

𝟎 + ෡𝜷 𝟏 𝑿 𝟏 + ෡𝜷 𝟐 𝑿 𝟐 + ෡𝜷 𝟑 𝑿 𝟑 + ෡𝜷 𝟒 𝑿 𝟒 +෡𝜷 𝑻 𝑻

Pharmaceutical Sector: Predicting Market Capitalization

Dependent Variable

ෝ𝒚

𝑷𝒓𝒆𝒅𝒊𝒄𝒕𝒆𝒅 𝑴𝒂𝒓𝒌𝒆𝒕 𝑪𝒂𝒑

Coefficients

෡ β

198.815

Independent Variables (Predictors) 𝑿 𝟏 Annual Revenue per Employee 𝑿 𝟐 Annual Finished Goods 𝑿 𝟑 Annual Retention Ratio 𝑿 𝟒 Annual Debt to Assets 𝑿 𝟓 Annual Profit Margin 𝑻 𝑻𝒊𝒎𝒆 (𝒀𝒆𝒂𝒓)

෡ β

0.0002

෡ β

0.0901

෡ β

-219.236

෡ β

-135.621

෡ β

749.043

Key Statistics

෡ β T

R-squared

0.75

-3.970

F-statistic 104.4 Jarque-Bera (JB): 10.707 Prob (F-statistic) 0.000

ෝ𝒚 𝟎.𝟓 = ෡𝜷

𝟎 + ෡𝜷 𝟏 𝑿 𝟏 + ෡𝜷 𝟐 𝑿 𝟐 + ෡𝜷 𝟑 𝑿 𝟑 + ෡𝜷 𝟒 𝑿 𝟒 + ෡𝜷 𝟓 𝑿 𝟓 +෡𝜷 𝑻 𝑻

Wrap-up

Summary:

• Traditional supply chain approaches reduce costs, but do not necessarily drive improvements in value. • Many actions in the last decade reduced value decreasing enterprise resilience. • The market capitalization model is actionable by the supply chain leader.

Questions?

Thank You…

Page 1 Page 2 Page 3 Page 4 Page 5 Page 6 Page 7 Page 8 Page 9 Page 10 Page 11 Page 12 Page 13 Page 14 Page 15 Page 16 Page 17 Page 18 Page 19 Page 20 Page 21 Page 22 Page 23 Page 24 Page 25 Page 26 Page 27 Page 28 Page 29 Page 30 Page 31

Made with FlippingBook Ebook Creator