Results of outside-in planning testing. The use of market signals decreases demand latency and improves synchronization market-to-market. Implicit in the testing and definition is challenging five paradigms: accepting that the order is a poor proxy for demand, that demand moves from time-phased data to the management of flow through the building of demand visibility, and the improvement in outcomes by shifting from functional metrics to balanced scorecards, the construction of bi-directional orchestration capabilities, and the need for a supply chain planning master data layer.
9. •About LKQ
LKQ Corporation is an eCommerce
provider of alternative and specialty parts
to repair and accessorize automobiles and
other vehicles.
The Company is global, with operations in
North America, Europe, and Taiwan. LKQ
offers its customers a broad range of
Original Equipment (OE) recycled and
aftermarket parts, replacement systems,
components, equipment, and services to
repair and accessorize automobiles, trucks,
and recreational and performance vehicles.
Currently, the Company has 44,000
Employees and 1,600 Locations in 31
Countries.
10. The Problem
Orchestration
Channel Sensing
Sensing Supply Sensing
Variability: Cycles, conversions, and grades
Growth
Customer
Service
Margin
Inventory
Turns
Safety
Return on
Invested
Capital
Balanced Scorecard
Listening
Post
Pattern
Recognition
of Channel
Data
Interest
Social
Graph Mining
Contract
Manufacturing
Alternate
Sourcing
Alternate Bill
of Materials
Unified Data Model
Planning Master Data
Baseline Demand Plan Feasibility
Events
Risk Sensing
Demand
Shaping
Portfolio
Shifts
Quality
Sensing
Logistics
Sensing
Leadtime
Variability
Commodity
Price Shifts
Platform
Changes
Social
Sentiment
Weather
Conversion
Rates
Yield
Maintenance
Schedules
Reverse Bill
of Materials
Alternate
Routing
Factory
Reliability/Capabilitie
s
Alternate
Distribution
Alternate
Channel
Images Market
Shifts
Weather
Pattern
Recognition
Images
Postponement
Calendar(s)
Replenishment
Policies
Allocation
The goal was to build an “inventory buy
plan” in each S&OP cycle with a playbook
based on inflation, supply-side variation
and risk, and changes in demand.
11. Pilot Scope
SKU Number SKU Description
1343305 BREMSSCHEIBE HA FUER AUDI
1343349 BREMSSCHEIBE HA FUER
1343696 BREMSSCHEIBE HA FUER
2215644 BREMSSCHEIBE 2-TEILIG HA
2256825 BREMSSCHEIBE HA FUER
3124261 BREMSSCHEIBE HA FUER ALFA
3281997 BREMSSCHEIBE HA FUER AUDI
3285253 BREMSSCHEIBE HA FUER AUDI
3521850 BREMSSCHEIBE HA FUER
3573804 BREMSSCHEIBE HA FUER
3681698 BREMSSCHEIBE HA FUER
4607865 BREMSSCHEIBE HA FUER
4625360 BREMSSCHEIBE HA FUER AUDI
4628860 BREMSSCHEIBE HA FUER AUDI
4630096 BREMSSCHEIBE HA FUER AUDI
5271371 BREMSSCHEIBE HA FUER
5271372 BREMSSCHEIBE HA FUER
5415080 BREMSSCHEIBE 2-TEILIG VA
7327083 BREMSSCHEIBE HA FUER
1. No of SKUs- 19
2. Distribution Network - 1 CDC, 150+ Branches (Germany
and Austria)
3. Location are grouped under the following:
• Garage, E-Commerce, Wholesale
• Garage, E-Commerce
• Garage, Wholesale, E-Commerce, Key Account
• Garage
• Garage, Wholesale
4. Sales Region – Austria, Germany, Slovenia
12. FVA Analysis
Forecast Value Add (FVA) analysis by comparing the
accuracy against the shipment (at CDC level for all items)
data. The analysis includes the forecast generated by LKQ,
o9 Stat model and o9 machine learning model by
comparing against the naive method (moving average of 5
months) as base forecast. The period of validation is
between M01-2022 to M05-2022
13. Insights
▪ LKQ branch and CDC replenishment were out of sync. Neither was right. The branch
demand plan is structurally too low, whereas the CDC forecast is too high.
▪ The COV at a branch level was too high to forecast.
▪ Before the pilot, the organization was unaware of the issues.
▪ Thirteen market factors tested. Four were selected—COVID levels, GDP, car registrations,
and inflation.
16. Scenario overview
SCOPE: Modeling of 2022 Data | Jan thru’ Jun | Weekly Buckets
Data inputs 1.
Enterprise Sync
LKQ Forecast
2.
Enterprise Sync
o9 Statistic Model
3.
Market Signal
Sensing
Market Driven
4
Market Signal
Sensing
& Supply lead time
o9 Machine Learning
5
Bi-directional:
Market Driven
Demand & Supply
lead time & price
Forecast LKQ Stat Fcst
sales history
o9 Stat Fcst
sales history
o9
Order + Market
Signals
o9 ML Fcst
(Order Invoices +
Market Signals)
o9 ML Fcst
(Order Invoices +
Market Signals)
Supply Lead Time Lead Time
Purchase Price
Safety Stock
calculation
Using Actual lead time from 2021 and variability Assumptions on
improved supplier
reliability
Assumptions on
improved supplier
reliability
Purchase
price
Increase by 30% from
Feb
19. Corning Demand Insights
Corning o9 Statistical
Enterprise
o9 Market Driven
Inputs Customer
Orders
Customer
Orders
Customer orders
+ Market Signals
Market Driver
Lags
Unknown Unknown ~ 3-4 Months
Forecast Value
Added (FVA)
-4.16% 0.55% 5.43%
Forecast Bias Negative Negative /
Balanced
Balanced
Pilot Insights:
● Outside-In forecasts consistently outperform inside-out statistical forecasting approaches
● Correlation does not equal causality – but can be helpful in the model (Temperature)
● COV analysis confirmed forecastability at the product family level
20. Corning - Summary Observations
• The division has a negative FVA (~4.2%). The FVA technique is new for the group.
• ML Drivers that showed strong correlation and ability to enhance FVA by ~+10%
• Selling Price
• Temperature
• GDP
• Covid
• The organization looks at price management and demand in disconnected processes.
• There is a need to enhance insights on pricing communication lead times, and effective
dates
• Inflation Rate
• Housing Starts
• 5G Subscriptions
• 5G Sites