Centralizovano planiranje, optimizacija zaliha i automatizacija poručivanja u maloprodajnom lancu je bila tema predavanja koje je ispred kompanije Tools Group održao Richard Wallis. U okviru prezentacije je prikazan način optimizacije procesa nabavke u trgovini koja zavisi od promocije proizvoda.

1. OPTIMIZACIJA LANCA SNABDEVANJA
Godišnja Konferencija Srpske Logističke Asocijacije
November 15, 2016
Richard D Wallis
Technical Director & CSA Global Operations
ToolsGroup
Supply Chain Planning & Optimisation with Automated
Replenishment in Promotion Driven Markets

2. Company Confidential All Rights Reserved2
Who are ToolsGroup
EUROPE
Amsterdam
Milan
Munich
Barcelona
London
Stockholm
Belgrade
AMERICAS
Boston
Buenos Aires
Mexico D.F.
Ontario
Quito
Sao Paulo
ASIA/Pacific
Manila
Bombay
Tokyo
Penang
Sydney
MIDDLE EAST
Tel Aviv
AFRICA
Cape Town
Our Vision
Accelerating Business Performance
through Market Driven Demand Analytics,
Demand Planning and Supply Chain Optimization
Supply Chain Specialist
ToolsGroup is a Supply Chain specialist with over
25 years business engineering & optimisation
experience in global and regional operations with
almost 400 customer references

3. Company Confidential All Rights Reserved3
Single Bolt-On Model Solution for Integrated Supply Chain Planning
DEMAND
MODELING
&
ANALYTICS
ERP SCP
SO99+ Single Model
Forecast &
Variability
Stocking
LevelsMULTI-
ECHELON
INVENTORY
OPTIMIZATION
Recommended
Orders
DASHBOARDS & PERFORMANCE MANAGEMENT
INTEGRATED BUSINESS PLANNING
Demand
Collaboration
Hub
Supply
Collaboration
Hub
REPLENISHMENT
and PRODUCTION
PLANNING
Demand
Scenarios

4. Company Confidential All Rights Reserved4
Recommended by Gartner

5. Company Confidential All Rights Reserved5
Retailer 1 Network
Retailer M Network
Retailer N Network
Vendor Network
At each network tier/location:
• Orders are fulfilled from stock
Typical Single Stage Planning & Replenishment Model
At each network tier/location:
• Projected demand and inventory are compared
• .. replenishment orders are calculated
• .. and passed upstream (as demand)
Demand=Tražnja
Supply=Snabdevanje

6. Company Confidential All Rights Reserved6
• The right hand does not know what the left hand is doing
• The system is (blindly) reactive
• The Bullwhip (Forrester) disease is present
• Inventory at any network tier is typically oversized
• The inefficiency adds excessive cost to the total supply chain
The Single Stage Planning & Replenishment Model
• “Beer Game” anyone ?

7. Company Confidential All Rights Reserved7
Plants /
Suppliers
Central
Warehouse
Regional
Warehouses
How much of this is YOUR network ?
Retailer DC’s
CONSUMERS
Stores
RETAILER NETWORKVENDOR NETWORK

8. Company Confidential All Rights Reserved8
The Fully Automated Supply Chain: Dynamic Multi-Echelon Replenishment
Processing a ‘time-phased’
Stochastic Demand Signal
across the entire network

9. Company Confidential All Rights Reserved9
New Generation Planning & Replenishment Models
Three key processes must be tamed and synchronised in order to
enable a fully automated supply chain:
• Forecast
• Inventory Plan
• Replenishment (supply and deployment)
But how can you automate the forecast in a business with highly promotion
driven demand ?

10. Company Confidential All Rights Reserved10

11. Company Confidential All Rights Reserved11
Mastering the Long Tail
• Ever larger number of SKU/Ls with highly
heterogeneous behaviour:
• Fast
• Slow
• Erratic
• …
• The challenge is to understand the
behaviour of each SKU-Location across
multiple dimensions:
• Quantity
• Order-line frequency
• Variability
• Forecast error
Self-adaptive forecasting techniques allow
business to address this growing issue and
provide a reliable base line forecast at the most
detailed market levels

12. Company Confidential All Rights Reserved12
New Generation Promotional Planning Models
Machine Learning (AI) techniques facilitate the automation of promotion planning
in the forecast.
The first step is to establish a set of promo types, characterised by:
• characteristics of the promo
• duration
• location
Then the promos can be clustered into behavioural types: uplift characteristics.
This allows:
• the modelling of the historical performance
• the application of the model to future events

13. Company Confidential All Rights Reserved13
-3 -2 -1 0 1 2 3 4
1.01.52.02.5
#week
norm.sale
Automating the Forecasting of Promotions & the Cleaning of the History
Profitable promotion planning entails accurately splitting baseline demand from promotional uplift
PAST | Learning FUTURE | Forecasting
PROMO 01 PROMO 02 PROMO 03 PROMO 04 PROMO 05 PROMO 06
FUTURE
PROMO 01
FUTURE
PROMO 02
Baseline Demand Actual Demand Baseline Forecast Promotional Forecast
Uplift models are applied to the future baseline to generate an accurate and reliable forecast

14. Company Confidential All Rights Reserved14
Demand Phenomena
Historical
Demand
Historical
Demand
Machine Learning Engine in SO99+
New Products
Web
Media
Promotions
(auxiliary engines)
OTHER
Specific dataset
extraction
Demand
Modeling
SPECIAL
ACTIONS
Promo
Variables
EXTERNAL
VARIABLES
Media
Variables
Sentiment
Data LAUNCH
PROFILES
Introduction
Variables
SEASONALITY
Statistical
Variables

15. Company Confidential All Rights Reserved15
Automating the Forecast enables an automated Replenishment Plan

16. Company Confidential All Rights Reserved16
ToolsGroup serves customers with complex demand
environments, intermittent and irregular demand, promotions,
frequent NPIs and heterogeneous product categories
Core Verticals include
CPG Electronics Food &
Beverage
Fashion Healthcare
& Pharma
Aftermarket
Parts
Industrial &
Durables
Retail Specialty
Chemicals
Wholesale
Distribution
Our 98% customer
retention rate is
among the highest in
the industry
Frequently these customers
have large and complex
distribution networks
including vertically
integrated retailers,
aftermarket parts
manufacturers,
multi-enterprise networks,
and multi-channel supply
chains
Innovator in machine learning
and advanced analytics for
supply chain planning
25 years of demand analytics,
supply chain planning and
optimization experience
15-30%
Average decrease
In inventory
& costRATED #1 In
Inventory
Optimization
By Nucleus
Research
2X to 5X
Improvement In
Planner Productivity
2-15%
Average
Improvement in
Service Levels
390 +
Customers with
Offices around the
World
Vaulted into the
LEADERS
quadrant
SOR - 2016
By GARTNER

17. Company Confidential All Rights Reserved17
Some Businesses using the SO99+ Technology

18. Company Confidential All Rights Reserved18
SME running SO99+ Technology
Retail SMEs in competitive growing markets
Company Sector Locations Item/Locations * Country
Fybeca PC & Pharma 150 300,000 Ecuador
Sana Sana PC + HC 500 250,000 Ecuador
Kronans Apotek PC & Pharma 304 350,000 Sweden
KIKO Cosmetics 700 200,000 Italy
Costa Express Coffee >5,000 1,000,000 UK
Nikora Food 130 60,000 Georgia
Generika Pharma >650 1,500,000 Philippines
Oki Doki Snack & Beverage >20 4,000 Ecuador
Novo Mundo White Goods 130 150,000 Mexico
Lilly Personal Care 200 2,500,000 Serbia/Bulgaria
Andrew Page Automotive 135 15,000,000 UK
* Approximate figures

19. Company Confidential All Rights Reserved19

20. Case Studies

21. Company Confidential All Rights Reserved21
Danone Case Study
DEMAND SHAPING
KPIs (PROMO)
2010 2011 2012*
Net UPLIFT %
“TRADITIONAL”
SUPPLY CHAIN KPIs
…BEYOND
NUMBERS…
TOTAL
FORECAST
ERROR
-20%
 ROBUST FOUNDATION
OF A VALUE DRIVEN
S&OP
 CONTINUOUS
OPTIMIZATION OF
INVESTMENT EFFICIENCY
 CONSISTENT BUSINESS
STEERING CYCLE
 FLEXIBLITY AND
RESPONSIVENESS TO
BUSINESS OBJS SHIFTS
-30%
-30%
LOST SALES
(SELL IN)
FG
OBSOLESCENCE
+36% +55%
2010
+6% +8%
2011 2012*
Net ROI
Recent Update
• Forecast accuracy
improved to 92%
• 37 consecutive months
have exceeded 98.7%
service level goal
• Number of demand
planners reduced to 2
• APO used only as data
transmission tool

22. Company Confidential All Rights Reserved22
POS Demand Sensing & Replenishment with 15 Minute Latency