View
2.136
Download
4
Tags:
Embed Size (px)
DESCRIPTION
By Trevor Miles, VP of thought leadership, Kinaxis
Citation preview
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 1
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 2 Copyright © 2011 Kinaxis Inc. All Rights Reserved.
Inventory Optimization: A lot more than theory Trevor Miles director, thought leadership e: [email protected] | m: +1.647.248.6269 | t: @milesahead
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 3
Agenda
• Intro to Kinaxis • How are we doing? • Demand and Supply Chain Segmentation • Classical IO Methods • Classical Example • Multi-Echelon Inventory Optimization • Conclusion
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 4
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 5
Companies can’t predict the future, …build organizations that will
survive and flourish under …any possible future. Source: McKinsey Quarterly, Dynamic management: Better decisions in uncertain times, December 2009
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 6
Market Dynamics
• Top CPG companies forecast performance Terra Technologies
– MAPE for a one month lag was 31% + 12% • Forecast Error Range: 19% - 43%
– Eight years ago: 36% + 10% MAPE
• High-Tech/Electronics – anecdotal – Struggle to get better than 50% MAPE
• Where will breakthrough performance come from? – Learning to forecast and plan better? – Learning to respond profitably to plan variance?
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 7 Copyright © 2011 Kinaxis Inc. All Rights Reserved.
How are we doing?
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 8
Inventory Performance – Computer Hardware
ê $10B
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 9
Inventory Performance – Household CPG
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 10
Inventory Performance – Automotive OEM
é $10B
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 11
Inventory Performance – Pharmaceutical
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 12 Copyright © 2011 Kinaxis Inc. All Rights Reserved.
Demand Segmentation
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 13
Demand Segmentation
• Each point is a specific SKU • Coefficient of variation = Std.Dev./Mean
80% of Volume
80% of Variability
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 14
Demand Segmentation
• Each point is a specific SKU • Coefficient of variation = Std.Dev./Mean
80% of Volume
80% of Variability
How would this graph look for: • Revenue? • Margin?
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 15
Supply Chain Segmentation
• Each point is a specific SKU • Coefficient of variation = Std.Dev./Mean
Mak
e-to
-Sto
ck
Configure-to-Order
Pul
l
Make-to-Order Rationalize?
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 16
Supply Chain Segmentation
• Each point is a specific SKU • Coefficient of variation = Std.Dev./Mean
Mak
e-to
-Sto
ck
Configure-to-Order
Pul
l
Make-to-Order Rationalize?
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 17
Supply Chain Segmentation for Product Family
Venu Nagali , Procurement Risk Management (PRM) at Hewlett-Packard Company, Stanford Risk Management Roundtable, November 13, 2006 http://www.gsb.stanford.edu/scforum/login/pdfs/HP%20%20PRM%20Nov%2006%20Venu%20Nagali.ppt
Flexible quantity contract
Demand forecast (units)
Time
Fixed quantity contract
0
100
200
300
400 Uncommitted
Hi scenario
Base scenario
Lo Scenario
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 18
Supply Chain Segmentation for Product Family
Time 0
100
200
300
400
Hi
Base
Lo
Time
Hi
Base Lo
Time
Hi
Base
Lo
• Where will breakthrough performance come from?
– Learning to forecast and plan better? – Learning to respond profitably to plan variance?
High-Tech/Electronics
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 19
Product Life Cycle Segmentation
Toolkit: Frameworks to Design and Enable Supply Chain Segmentation, Matthew Davis, Gartner, 19 May 2011
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 20 Copyright © 2011 Kinaxis Inc. All Rights Reserved.
Classic IO Methods
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 21
Common Safety Stock Calculation
• Demand rate: – the amount of items consumed by customers, on average, per unit
time. • Lead time:
– the delay between the time the reorder point (inventory level which initiates an order) is reached and renewed availability.
• Service level: – the desired probability that a chosen level of safety stock will not
lead to a stock out. Naturally, when the desired service level is increased, the required safety stock increases as well.
• Forecast error: – an estimate of how far actual demand may be from forecasted
demand. Expressed as the standard deviation of demand.
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 22
Common Safety Stock Calculations
• Service Factor * Forecast Error * √𝐿𝑒𝑎𝑑 𝑇𝑖𝑚𝑒 – Where’s the representation of supply variability? – How is demand variability associated with forecast error?
• (Demand Variability Factor) * (Service Factor) * (Lead-Time Factor) * (Order Cycle Factor) * (Forecast-to-Mean-Demand Factor) – Where’s the representation of supply variability?
• Service Factor * √𝜇↓𝐿𝑇 ↓↑2 ∗ 𝛿↓𝐷 ↓↑2 + 𝜇↓𝐷 ↓↑2 ∗ 𝛿↓𝐿𝑇 ↓↑2 – Includes supply variability
• What about Multi-Echelon Inventory Optimization? – Have you looked at the math?
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 23
What about Supply Uncertainty?
• Types of supply uncertainty: – Lead-time uncertainty – Yield uncertainty – Inspection failure – Disruptions
• Strategies for dealing with supply uncertainty – Safety stock inventory – Dual sourcing – Improved forecasts
• Has big effect on inventory levels
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 24
Dimensionless Analysis
• Safety Stock = Service Factor * √𝜇↓𝐿𝑇 ↓↑2 ∗ 𝛿↓𝐷 ↓↑2 + 𝜇↓𝐷 ↓↑2 ∗ 𝛿↓𝐿𝑇 ↓↑2
• Safety Stock = Service Factor * 𝜇ͿԤԥԦԧԨԩԪԫԬԭԮԯՠֈ֍֎֏ࢪࢩࢨࢧࢦࢥࢤࢣࢢࢡࢠࡪࡩࡨࡧࡦࡥࡤࡣࡢࡡࡠ࡞࡛࡚࡙ࡘࡗࡖࡕࡔࡓࡒࡑࡐࡏࡎࡍࡌࡋࡊࡉࡈࡇࡆࡅࡄࡃࡂࡁࡀ࠾࠽࠼࠻࠺࠹࠸࠷࠶࠵࠴࠳࠲࠱࠰࠭ࠬࠫࠪࠩࠨࠧࠦࠥࠤࠣࠢࠡࠠࠟࠞࠝࠜࠛࠚ࠙࠘ࠗࠖࠕࠔࠓࠒࠑࠐࠏࠎࠍࠌࠋࠊࠉࠈࠇࠆࠅࠄࠃࠂࠁࠀ߿߾߽ٟؠׯ𝐿𝑇 * 𝜇↓𝐷 * √𝐶𝑜𝑉↓𝐷 ↓↑2 + 𝐶𝑜𝑉↓𝐿𝑇 ↓↑2 – CoV = 𝛿⁄𝜇
• Safety Stock / (𝜇ͿԤԥԦԧԨԩԪԫԬԭԮԯՠֈ֍֎֏ࢪࢩࢨࢧࢦࢥࢤࢣࢢࢡࢠࡪࡩࡨࡧࡦࡥࡤࡣࡢࡡࡠ࡞࡛࡚࡙ࡘࡗࡖࡕࡔࡓࡒࡑࡐࡏࡎࡍࡌࡋࡊࡉࡈࡇࡆࡅࡄࡃࡂࡁࡀ࠾࠽࠼࠻࠺࠹࠸࠷࠶࠵࠴࠳࠲࠱࠰࠭ࠬࠫࠪࠩࠨࠧࠦࠥࠤࠣࠢࠡࠠࠟࠞࠝࠜࠛࠚ࠙࠘ࠗࠖࠕࠔࠓࠒࠑࠐࠏࠎࠍࠌࠋࠊࠉࠈࠇࠆࠅࠄࠃࠂࠁࠀ߿߾߽ٟؠׯ𝐿𝑇 * 𝜇↓𝐷 ) = Service Factor * √𝐶𝑜𝑉↓𝐷 ↓↑2 + 𝐶𝑜𝑉↓𝐿𝑇 ↓↑2 – Safety Stock = Units
– 𝜇ͿԤԥԦԧԨԩԪԫԬԭԮԯՠֈ֍֎֏ࢪࢩࢨࢧࢦࢥࢤࢣࢢࢡࢠࡪࡩࡨࡧࡦࡥࡤࡣࡢࡡࡠ࡞࡛࡚࡙ࡘࡗࡖࡕࡔࡓࡒࡑࡐࡏࡎࡍࡌࡋࡊࡉࡈࡇࡆࡅࡄࡃࡂࡁࡀ࠾࠽࠼࠻࠺࠹࠸࠷࠶࠵࠴࠳࠲࠱࠰࠭ࠬࠫࠪࠩࠨࠧࠦࠥࠤࠣࠢࠡࠠࠟࠞࠝࠜࠛࠚ࠙࠘ࠗࠖࠕࠔࠓࠒࠑࠐࠏࠎࠍࠌࠋࠊࠉࠈࠇࠆࠅࠄࠃࠂࠁࠀ߿߾߽ٟؠׯ𝐿𝑇 = Periods
– 𝜇↓𝐷 = Units/Period
– Safety Stock / (𝜇ͿԤԥԦԧԨԩԪԫԬԭԮԯՠֈ֍֎֏ࢪࢩࢨࢧࢦࢥࢤࢣࢢࢡࢠࡪࡩࡨࡧࡦࡥࡤࡣࡢࡡࡠ࡞࡛࡚࡙ࡘࡗࡖࡕࡔࡓࡒࡑࡐࡏࡎࡍࡌࡋࡊࡉࡈࡇࡆࡅࡄࡃࡂࡁࡀ࠾࠽࠼࠻࠺࠹࠸࠷࠶࠵࠴࠳࠲࠱࠰࠭ࠬࠫࠪࠩࠨࠧࠦࠥࠤࠣࠢࠡࠠࠟࠞࠝࠜࠛࠚ࠙࠘ࠗࠖࠕࠔࠓࠒࠑࠐࠏࠎࠍࠌࠋࠊࠉࠈࠇࠆࠅࠄࠃࠂࠁࠀ߿߾߽ٟؠׯ𝐿𝑇 * 𝜇↓𝐷 ) = 𝑈𝑛𝑖𝑡𝑠 /𝑃𝑒𝑟𝑖𝑜𝑑𝑠 ∗ 𝑈𝑛𝑖𝑡𝑠⁄𝑃𝑒𝑟𝑖𝑜𝑑 = 𝑈𝑛𝑖𝑡𝑠 ∗ 𝑃𝑒𝑟𝑖𝑜𝑑/𝑃𝑒𝑟𝑖𝑜𝑑 ∗𝑈𝑛𝑖𝑡 = 1
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 25
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 26 Copyright © 2011 Kinaxis Inc. All Rights Reserved.
Classical Example
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 27
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 28
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 29
Effect of Lead Time Variability on SS
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 30 Copyright © 2011 Kinaxis Inc. All Rights Reserved.
Multi-Echelon Inventory Optimization
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 31
An Approximate Method
• Assume that each stage carries sufficient inventory to deliver product within S periods “most of the time” – Definition of “most” depends on service level – S is called the committed service time (CST)
• We simply ignore the times that the stage does not meet its CST – For the purposes of the optimization – Allows us to pretend LT is deterministic
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 32
Net Lead Time
• Each stage has a processing time T and a CST S
• Net lead time at stage i = Si+1 + Ti – Si
3 2 1
T3 T2 T1
S3 S2 S1
“bad” LT “good” LT
Prof. Larry Snyder, Multi-Echelon Inventory, Lehigh University, June 15, 2006
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 33
Net Lead Time vs. Inventory
• Suppose Si = Si+1 + Ti – e.g., inbound CST = 4, proc time = 2, outbound CST = 6 – Don’t need to hold any inventory – Operate entirely as pull (make-to-order, JIT) system
• Suppose Si = 0 – Promise immediate order fulfillment – Make-to-stock system
Prof. Larry Snyder, Multi-Echelon Inventory, Lehigh University, June 15, 2006
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 34
Net Lead Time vs. Inventory
• Precise relationship between NLT and inventory:
• NLT replaces LT in classical formula • Ignores effect of supply variability • Choose S (committed service level) at each stage • Efficient algorithms exist for finding optimal S values
– Minimize holding cost while meeting customer service – Optimal for only a few stages to hold inventory
• Essentially decomposes multi-echelon problem into multiple stages
NLTzNLTy σµ α+×=*
Prof. Larry Snyder, Multi-Echelon Inventory, Lehigh University, June 15, 2006
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 35
A Hybrid Push-Pull System
• Part of system operated produce-to-stock, part produce-to-order
• Moderate lead time to customer • Influenced by postponement strategy
PART 1 DALLAS ($260)
15 7
8
PART 2 CHARLESTON ($7)
14
PART 4 BALTIMORE ($220)
5
PART 3 AUSTIN ($2)
14
6
8
5
PART 5 CHICAGO ($155)
45
PART 7 CHARLESTON ($30)
14
PART 6 CHARLESTON ($2)
32
8
30
7
8
9 45
14
32
push/pull boundary
Prof. Larry Snyder, Multi-Echelon Inventory, Lehigh University, June 15, 2006
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 36
Practical & Pragmatic, but …
• Ignores – Multi-sourcing – Alternate parts – Alternate routing – ECO/ECN – Product life-cycle stage
• Especially new product introduction
• …and what about – Changes in product mix? – NPI effect on component requirements? – Transportation costs? – Labor costs?
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 37 Copyright © 2011 Kinaxis Inc. All Rights Reserved.
Conclusion
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 38
Be Practical & Pragmatic
• Some observations on Inventory Optimization – More about SC design than Operations – All ‘theories’ based upon rule-of-thumb – Effort to maintain should be less than effort to deploy – Demand changes more quickly than inventory policy
• Some practical suggestions – Start with segmentation
• Customers • Products
– Use theories to set ball park inventory levels – Convert these to periods of supply (DOS, WOS, …)
• Adjusts automatically for seasonality and life-cycle stage
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 39 Copyright © 2011 Kinaxis Inc. All Rights Reserved.
Thank you! Questions? Trevor Miles | e: [email protected] | m: +1.647.248.6269 | t: @milesahead
Copyright © 2011 Kinaxis Inc. All Rights Reserved. 40