PepsiCo’s Practical Application of Supply Chain Resilience Strategies and Inventory Optimization
Tim Rowell, CPIM
PepsiCo, Supply Chain Sr. Manager
#APICS2016
Irma
Tuesday Morning Forecast Wednesday Morning Forecast Thursday Morning Forecast
#APICS2016
IrmaSaturday Morning Forecast Sunday Evening
= My House
#APICS2016
Agenda
Overview / Background
Supply Chain Resilience
Inventory Deep Dive
Case Studies
#APICS2016
OVERVIEW / BACKGROUND
#APICS2016
PepsiCo Overview
Global Beverages
Global Snacks
Global Nutrition
Brands
22billion-dollar
brands
Performance
More than $63 billion
revenue
Scale
>200 countries
& territories
People
More than 250,000
employees
PepsiCo is a global food and beverage powerhouse.
Our broad range of delicious products offers consumers convenient,
nutritious and affordable options in nearly every country around the world
#APICS2016
PepsiCo Overview
Good For You
Better For You
Fun For you
#APICS2016
Specialized Supply Chain
As PepsiCo proactively adapts to emerging market trends for healthy &
nutritious beverages, our supply chain becomes more global and complex
Global Nutrition Segment
Good for You Products
Chilled Supply Chain
#APICS2016
Case Study - International Supply
Natural, Functional Beverage
#APICS2016
PepsiCo Coconut Water Supply Chain
• Copackers in Southeast Asia
• Packaging Materials from Germany,
Pakistan, Serbia, Singapore
• Warehouses near East/West ports
• Redistributed to regional DCs
• Backup US Copackers
Southeast Asia
#APICS2016
Supply Chain Starts Here
#APICS2016
SUPPLY CHAIN RESILIENCE
#APICS2016
Supply Chain Resilience Concepts
Presented at 2013 APICS International Conference
(Orlando)
Michigan State PHD, Steven Melnyk
Supply Chain Resilience – The ability of a supply chain to anticipate, create plans to
avoid or mitigate, and/or to recover from disruptions to supply chain functionality
APICS Dictionary, 14th Edition, ©2013
#APICS2016
Supply Chain Resilience Strategies
#APICS2016
Several strategies for SC Resilience: a) early warning, b) buffers, c)
supply chain configuration d) brand equity…etc.
Relevant strategies applied to managing our coconut water supply chain
a) Early Warning1) The earlier you are aware of the disruption, the sooner you can start working on the recovery
2) Improve frequency of communications and data quality
b) Buffers – 3 kinds of buffers
1) Stock Buffers
- Pre-build prior to high risk SE Asia supply timeframe and prior to peak seasonal demand
in US
- Optimize inventory targets to balance fill rate and expiration risk
2) Capacity Buffers
- SE Asia: how much more can we get above the contractual commitments when needed
- US copack: how much can we get from US copackers when needed
3) Lead Time Buffers
- SE Asia – flexibility to change short term schedule; dual sourcing where possible (500ml
Pure)
- US copackers – stage long lead time materials for quicker response when SE Asia supply
disrupted (or demand spikes)
Application to Coconut Water Supply Chain
#APICS2016
Seasonal Inventory DOH
Targets
Supply & Demand
SeasonalityLow Sales Moderate Sales Peak Sales Moderate Sales Low Sales
Demand Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Supply Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct
Typhoons Moderate Risk Low Risk Moderate Risk Peak Risk
Holidays HolidaysChinese
NY
Holy
WeekRamadan
Understand and Address Seasonal Risks
#APICS2016
Production Lock Hold Transit, Customs, Port Dist
Line Time, Ingredients Hold Dist
8 2 8 2
8 2 2
FG lead time from SE Asia
is 20 weeks
2nd SE Asia copacker *
improves reliability**, but
same 20 week lead time
US contingency lead time is
8 weeks (less if line time &
ingredients available)
Notified of
Supply
Disruption
Order More from
SE Asia Copack
#1 or #2
Schedule US
Contingency
Supply
Production Lock Hold Transit, Customs, Port Dist
8 2 8 2
Arrives 8
Weeks Sooner
vs SE Asia
Example Supply Scenario and Lead Times
(weeks)
* Dual SE Asia sourcing currently applies only to 500ml Pure. For others, scheduler will increase plan at
Copack #1 in subsequent months after disruption has subsided.
** It is important to quantify (and optimize) both average (m) and standard deviation (s) of supply (lead time and quantity).
For example, transit and port clearance times continually change; monitor and update as needed.
Decompose Lead Times and Optimize
#APICS2016
Certification
Supply – stable, on shelf, fresh
Promotion – new channels, drive awareness
Brand Equity and Reputational Resilience
#APICS2016
INVENTORY DEEP DIVE
#APICS2016
Inventory Policy is Part of SC Resilience
– Our work on applying these Supply Chain Resilience strategies within
PepsiCo’s global supply chain is leading to groundbreaking research,
particularly in the area of inventory buffers.
– International supply has long lead times and high variability. Inventory
required to meet customer service targets begins to introduce risk of
expiration, especially with organic products with shorter shelf lifes.
– We have derived formulas in conjunction with industry experts and
academia (and validated with historical data) to predict both the percentage
of stockouts as well as the percentage of expired product, and implemented
the solution that strikes the optimal balance.
– In doing so for the coconut water business, we found opportunities to
reapply this to other emerging product categories with supply
challenges such as probiotic beverages. These probiotic beverages have
the unique challenge of very short shelf life compared to other items in the
portfolio of premium juice beverages.
– Adapting the formulas for the product-specific characteristics of probiotic
beverages (supply & demand variability, lead time and shelf life) allowed us
to find the optimal balance for that product group as well and implement the
results to improve our supply
#APICS2016
The SawtoothInventory vs Time
Place order and the reorder quantity (ROQ) arrives one lead time
later. Safety stock buffers against uncertainty of supply and
demand.
#APICS2016
Widely Documented: Correlate Inventory Level with Percent Backorders/Cuts
Inventory vs Time
backorders/cuts ( ) across all replenishment
cycles throughout the year given variable
demand and supply
divide by annual demand to get percent backorders/cuts
fill rate % = 1 – percent backordered/cut
#APICS2016
Emerging Area of Research: Correlate Inventory Level with Percentage Expired
amount leftover at the end of the
replenishment cycles ( ) throughout the
year
divide by annual demand to get percent expired
Expected inventory
(safety stock) prior
to receiving next
replenishment
Expected amount of
each production lot
leftover at the end
of its shelf life
Inventory vs Time
* “Shelf Life” in this context is viewed from internal perspective and has subtracted the “remaining shelf life” guaranteed to external customers upon receipt. It is otherwise known as “Warehouse Life”.
#APICS2016
Consider the Boundary Cases
ROQ½
ROQ Zero
-½ ROQ RO
Q
Scenarios of expected inventory at the end of product shelf life
assuming no demand or supply variability
100% expired
0% backordered/cut
50% expired
0% backordered/cut
0% expired
0% backordered/cut
0% expired
50% backordered/cut
0% expired
100% backordered/cut
#APICS2016
Illustration of the Symmetric Relationship
-ROQ -½ROQ Zero ½ROQ ROQ
#APICS2016
Layer in Variability
-ROQ -½ROQ Zero ½ROQ ROQ
#APICS2016
Definition of TermsSafety Stock (SS) = Expected inventory position when reorder arrives
Lead Time (LT) = Time it takes from the point an order is placed until it arrives and is available to sell to customers. Lead Time
Demand (LTD) is the demand during lead time
Lock = Time required by manufacturer that production schedule be locked and no more changes accepted
Make = Time allotted for manufacturer to produce the scheduled quantity and product to clear incubation (if required)
Dist = Time it takes to distribute product from manufacturer to demand points
ROQt = Run cycle (or time between replenishments); the timeframe that the reorder quantity (ROQ) is expected to cover
Shelf Life* = Total internal shelf life from the time produced until the last date it can be sold to customer (consumer shelf life minus
time guaranteed to customers upon receipt).
eRSL* = Expected inventory remaining at the end of the internal shelf life
#APICS2016
Method DerivationArranging Terms
X* = Lock + Make + Shelf Life – ROQt
Y* = SS + LTD – eRSL*
Where: LTD = Lock + Make + Dist
Y* = Lock + Make + Dist + SS – eRSL*
Algebraic Reduction
Set X* = Y*
Lock + Make + Shelf Life – ROQt = Lock + Make + Dist + SS –
eRSL*
Lock + Make + Shelf Life – ROQt = Lock + Make + Dist + SS –
eRSL*
Shelf Life – ROQt = Dist + SS – eRSL*
SS = Shelf Life – ROQt - Dist + eRSL*
Procedure
• Specify eRSL*
• Calculate SS
• Calculate mean (m) and standard deviation (s) of LTD
• Based on m, s, q (ROQ) and SS: calculate % backorders/cuts
(A)
- Iterate with conventional formula: SS=ks, k=f[L(z)=Aq/s]
- Or calculate explicitly per Zipkin: A=(s/q)[F1(zr)-F1(zr+q)]
• Calculate m* and s* of demand over X*
• Based on m*, s*, q and eRSL*: calculate % backorders/cuts*
• Determine estimated percentage of expired inventory based on
symmetric relationship with percentage backorders/cuts
(see calculation mechanics on next slide)
* “Shelf Life” in this context is viewed from internal perspective and has subtracted the “remaining shelf life” guaranteed to external customers upon receipt. It is otherwise known as “Warehouse Life”.
#APICS2016
Calculation MechanicsStart at midpoint (MP), where eRSL=0
• Calculate SSMP
• Calculate % backorders/cuts at SSMP given m, s, q
• Calculate % backorders/cuts at eRSLMP given m*, s*, q
Ascending
• Calculate % backorders/cuts at SSMP+2 given m, s, q
• Calculate % backorders/cuts at eRSLMP+2 given m*, s*, q
• Set % expired at = eRSLMP-2 = % backorders/cuts at eRSLMP+2
• Calculate % backorders/cuts at SSMP+4 given m, s, q
• Calculate % backorders/cuts at eRSLMP+4 given m*, s*, q
• Set % expired at = eRSLMP-4 = % backorders/cuts at eRSLMP+4
• Continue for MP+6, MP+6, etc (or increments other than 2)
Descending
• Calculate % backorders/cuts at SSMP-2 given m, s, q
• Calculate % backorders/cuts at eRSLMP-2 given m*, s*, q
• Set % expired at = eRSLMP+2 = % backorders/cuts at eRSLMP-2
• Calculate % backorders/cuts at SSMP-4 given m, s, q
• Calculate % backorders/cuts at eRSLMP-4 given m*, s*, q
• Set % expired at = eRSLMP+4 = % backorders/cuts at eRSLMP-4
• Continue for MP-6, MP-8, etc (or increments other than 2)
eRSL
(wks)
SS
(wks)
q
(wks)
Avg Cycle
Stock
Target
Inv
Avg Wkly
Demand s m m* s*Cuts %
for SS
Cuts %
for SS*
Expired
%
-10 -2 4 2 0 10,000 37,500 200,000 240,000 41,458 49.7% 81.4% 0.0%
-8 0 4 2 2 10,000 37,500 200,000 240,000 41,458 31.1% 67.1% 0.1%
-6 2 4 2 4 10,000 37,500 200,000 240,000 41,458 16.5% 49.8% 0.4%
-4 4 4 2 6 10,000 37,500 200,000 240,000 41,458 7.3% 32.6% 1.3%
-2 6 4 2 8 10,000 37,500 200,000 240,000 41,458 2.7% 18.5% 3.8%
Midpoint 0 8 4 2 10 10,000 37,500 200,000 240,000 41,458 0.8% 9.0% 9.0%
2 10 4 2 12 10,000 37,500 200,000 240,000 41,458 0.2% 3.8% 18.5%
4 12 4 2 14 10,000 37,500 200,000 240,000 41,458 0.0% 1.3% 32.6%
6 14 4 2 16 10,000 37,500 200,000 240,000 41,458 0.0% 0.4% 49.8%
8 16 4 2 18 10,000 37,500 200,000 240,000 41,458 0.0% 0.1% 67.1%
10 18 4 2 20 10,000 37,500 200,000 240,000 41,458 0.0% 0.0% 81.4%
#APICS2016
Model Correlates Well with Historical Data
SL=40d, Low/Med s SL=40d, Med/High s
#APICS2016
Deriving Formulas from Scratch
𝑠 − 𝑥 𝑔(𝑥)𝑑𝑥
∞‒
S
S = the order up to quantity in a periodic review systemp = production lead time (order to produce)w = warehouse life
f(x) = pdf for weekly demand xx ~ N (m, s)
g(x) = pdf for demand x over period p+w
x ~ N (m(p+w), s (p + w))h(x) = pdf for standard normal distribution
For any given S
E (cases expiring) =
Let z = (x-m’)/s’ → x = zs’ + m’ where m’ = m(p+w)
k = (S-m’)/s’ → S = ks’ + m’ where s’ = s (p + w)
∞‒
k
E (cases expiring) = s’(k-z)h(z)dz
∞‒
k
= s (p + w) (k-z)h(z)dz
Where: h(z) = 𝑒−½𝑧
2
2(standard normal dist: m=0, s=1)
Solve with Numerical Integration
#APICS2016
Other References1. Opher Baron, Managing Perishable Inventory, Rotman
School of Management, University of Toronto, Toronto,
Ontario, Canada. Wiley Encyclopedia of Operations
Research and Management Science, edited by James J.
Cochran, Copyright © 2010 John Wiley & Sons, Inc.
2. Karaesmen, I. Z., Scheller-Wolf, A., and Deniz, B.
Managing Perishable and Aging Inventories: Review and
Future Research Directions. In Kempf, K. G., Keskinocak,
P., and Uzsoy, R., editors, Planning Production and
Inventories in the Extended Enterprise, volume 151 of
International Series in Operations Research &
Management Science, chapter 15, pages 393-436.
Springer US, Boston, MA. 2011.
3. Nahmias, S. Perishable Inventory Systems, volume 160
of International Series in Operations Research &
Management Science. Springer US, Boston, MA. 2011.
4. Baron O, Berman O, Perry D. Stochastic (τ, S) models
for managing inventory of perishable items. Working paper.
Rotman School of Management, University of Toronto;
2008.
5. Brown GW, Lu JY, Wolfson RJ. Dynamic modeling of
inventories subject to obsolescence. Manag Sci
1964;11(1):51–63.
6. Nahmias S, Pierskalla W. Optimal ordering policies for a
product that perishes in two periods subject to stochastic
demand. Nav Res Logist Q 1973;20:207–229.
7. Broadheim E, Derman C, Prastacos GP. On the
evaluation of a class of inventory policies for perishable
products such as blood. Manag Sci 1975;22:1320–1325.
9. Lian Z, Liu L. A discrete-time model for perishable inventory
systems. Ann Oper Res 1999;87:103–116.
10. Weiss H. Optimal ordering policies for continuous review
perishable inventory models. Oper Res 1980;28:365–374.
11. Graves S. The application of queuing theory to continuous
perishable inventory system. Manag Sci 1982;28:400–406.
12. Nahmias S, Perry D, Stadje W. Actuarial valuation of
perishable inventory systems. Probab Eng Infor Sci 2004;18:219–
232.
13. Nahmias S. Perishable inventory theory: a review. Oper Res
1982;30(4):680–708.
14. Prastacos GP. Allocation of a perishable product inventory.
Oper Res 1981;29:95–107
#APICS2016
CASE STUDIES
#APICS2016
Case Study: Coconut WaterLong lead time, long shelf life products
9 month shelf life* 6 month shelf life***
* “Shelf Life” in this context is viewed from internal perspective and has subtracted the “remaining shelf life” guaranteed to external customers upon receipt. It is otherwise known as “Warehouse Life”.** Target inventory includes Safety Stock + Avg Cycle Stock + In Transit
#APICS2016
Case Study: Coconut WaterLong lead time, long shelf life products
**
* “Shelf Life” in this context is viewed from internal perspective and has subtracted the “remaining shelf life” guaranteed to external customers upon receipt. It is otherwise known as “Warehouse Life”.** Target inventory includes Safety Stock + Avg Cycle Stock + In Transit
Zoomed In View
9 month shelf life* 6 month shelf life*
#APICS2016
Case Study: Naked Juice SmoothiesShort lead time, short shelf life products
* “Shelf Life” in this context is viewed from internal perspective and has subtracted the “remaining shelf life” guaranteed to external customers upon receipt. It is otherwise known as “Warehouse Life”.** Target inventory includes Safety Stock + Avg Cycle Stock + In Transit
40 day shelf life* 20 day shelf life*
**
#APICS2016
Rules of ThumbFor Example: Max Stock Policy = ½ Shelf Life*
For the 4 examples studied, this policy will result in less than 10% expired ( , )
* “Shelf Life” in this context is viewed from internal perspective and has subtracted the “remaining shelf life” guaranteed to external customers upon receipt. It is otherwise known as “Warehouse Life”.** Target inventory includes Safety Stock + Avg Cycle Stock + In Transit
** **
Coconut Water Smoothies
#APICS2016
For More Info on Inventory Optimization
#APICS2016
2011 Inventory Optimization excerpt 1
#APICS2016
2011 Inventory Optimization excerpt 2
#APICS2016
Today’s presenter
Fun facts and APICS highlights
Tim Rowell- Supply Chain Sr. Manager
- 19 years at PepsiCo/Tropicana
- APICS CPIM since 2003
Contact Info:1001 13th Avenue EastBradenton, FL [email protected]/pub/tim-rowell/0/281/275•
1991 Bachelors of Science in Mechanical Engineering, University of Florida
1991 Started at Westinghouse in Nuclear Safety Analysis
1997 MBA (MSIA), Carnegie Mellon Graduate School of Industrial Administration
1997 Started at Tropicana, a division of Seagram's
1999 PepsiCo acquired Tropicana
1999 Began work on Tropicana inventory model with Dr. Sridhar Bashyam (Frito Lay US)
2000 Consulted with Dr. Steven Nahmias on periodic review and continuous review formulas
2000 Established inventory model based on Fill Rate for Tropicana business
2000 PepsiCo acquired Quaker/Gatorade
2001 QTG synergy
2003 APICS: became Certified in Production and Inventory Management (CPIM)
2003 Presented at APICS International Conference: Tropicana Inventory Planning Methodology
2005 Began work on Project OneUp including Multi-Echelon Inventory Optimization for QTG
2006 Applied Dr. Paul Zipkin (Duke/Fuqua) models for Continuous Review and Periodic Review
2007 PepsiCo acquired Naked Juice
2008 Mark Entsminger documented/circulated derivation for quantity-based supply variability
2009 PepsiCo implemented Project OneUp Release 4
2009 PepsiCo reacquired PBG
2010 LSS inventory optimization for ingredients (chilled supply chain)
2011 Presented at APICS International Conference: Essential Inventory Truths
2012 PepsiCo acquired 100% of ONE
2013 APICS International Conference: Michigan State PHD Steven Melnyk presented on SC Resilience
2013 Implemented APICS/MS Supply Chain Resilience strategies for coconut water
2014 PepsiCo NAB APICS Pilot
2015 Performed deep dive optimization and segmentation of Naked Juice business
2015 Developed model to correlate % backorders/cuts and % expired as function of target inventory
2016 Vetted model with academia and industry, including Terra Technology / E2open
2016 PepsiCo Inventory Optimization organization alignment
2016 Presented at APICS International Conference: Practical Application of SC Resilience Strategies
Luke 10:27
#APICS2016
THANK YOU