ls industry deep dive … · ls industry deep dive eyeon pid november 13th 2019 1. 2 agenda 16h00...
TRANSCRIPT
LS industry deep dive
EyeOn PID
November 13th 2019
1
2
Agenda
16h00 Kick off
16h15 No-touch S&OP – the human factor (Freek Aertsen)
17h00 Getting your inventory health under control (Maarten Van Liempd / Vasco Werners)
17h45 Integrate industry specific constraints & requirements into your planning (Bart Paridaen)
18h30 Dinner
20h00 Wrap-up & Drinks
3
Introducing your hosts for today
Vasco Werners Maarten Van Liempd Bart Paridaen
Loek Lemmens Freek AertsenLuc Van Wouwe
70 Specialists in realizing forecasting and planning improvements
EyeOn Company profile
• Focus on design and implementation
• Line management experience (f.e. Sales, SCM & Finance) and university educated
• Cross functional & hands-on mentality
• Proven track records in interim management
• Data science team with experts in planning and forecasting modelling
• System independent
• Industry focus; industrial companies, its suppliers and customers
• Network facilitator
• Deliver concrete projects with a short throughput time
(max. 100 days) and clear deliverables
4
5
EyeOn - Full service offering inforecasting and planning
Project Interim Academy
Execute & improve
planning processes with
temporary resources
Advance forecasting &
planning specialists in their
career
Forecast Services
Deliver best possible
forecast. Optimize
inventory parameters
Data Science & Solutions
Deliver technology and
models solving key
customer challenges in
turnkey solutions.
Develop & implement
tailored planning &
forecasting solutions
6
EyeOnKnowledge partner
Industry networks
Expert Sessions / Idea Labs
Benchmarks
Teaching: EyeOn Academy
Research & Publications
7
Agenda
16h00 Kick off
16h15 No-touch S&OP – the human factor (Freek Aertsen)
17h00 Getting your inventory health under control (Maarten Van Liempd / Vasco Werners)
17h45 Integrate industry specific constraints & requirements into your planning (Bart Paridaen)
18h30 Dinner
20h00 Wrap-up & Drinks
No-touch S&OPThe human factor!
Freek Aertsen
November 2019
8
9
10
No-touch planning is feasible in my organization in....
0%
5%
10%
15%
20%
25%
30%
35%
2 years 5 years 10 years 20 years never Don't want it
85
Simplify the world
Let’s put some sense in it!
11
σ 𝑃 − 𝐴
σ𝑃∗ 100 = 𝑥%
σ 𝐴 − 𝐹
σ𝐴∗ 100 = 𝑥%
Forecast Accuracy Manufacturing Schedule Adherence
How to improve (1) Forecasts and plans and (2) actuals
Improving!
Better forecasts and plans
Forecasting – using external data and analytics
• Causal forecasting using macro economic indicators
• NPI forecasting
• Promotion forecasting
• Improved time series forecasting
Supply planning – up to date parameters
• Data engineering and IOT to have up to date
planning parameters
• MEIO inventory optimization
Better actuals - reduce data latency
Market – what is happing with our product?
• Consumer insights, reviews
• Sell-through
Supply – faster response
• IOT predicting and preventing machine breakdown
• Integrate S&OP and S&OE
• Compliance improvement
• QC release by priority
12
Man and machine
13
No-Touch planning requires different building blocks
EyeOn vision
Integral process for immediate and fact based decision making- Along value chain
- Different functional areas
- Linking strategy to execution
Demand
Making the change
Inventory
Supply
E2E, S&OP, IBP
15
No-Touch planning requires different building blocks
EyeOn vision
Integral process for immediate and fact based decision making- Along value chain
- Different functional areas
- Linking strategy to execution
Demand
Making the change
Inventory
E2E, S&OP, IBP
Supply
Excellent data- Data engineering
- Collection (internal –
external)
- Data storage
- ……
EyeOn Analytics in Demand Planning benchmark shows limited use of
additional data sources!
Data collection
0 1 2 3 4 5 6 7 8 9 10
Order data
Point of sales - sell out
Point of sales - stock in trade
Consumer review data from the web
Price development of our and competitor products over time
Promotional activities of customers
Weather forecasts
Historical promotion information
Facebook data
Google search behaviour data
Webscraping
Other
Macro economic indicators
Usage of data sources (1 - 10 scale, N= 25)
Excellent data - process mining to determine missing master data to enable higher
planning reliability in Pharma company
Data engineering
>7M records of stock movements
18
No-Touch planning requires different building blocks
EyeOn vision
Integral process for immediate and fact based decision making- Along value chain
- Different functional areas
- Linking strategy to execution
Demand
Making the change
Inventory
E2E, S&OP, IBP
Supply
Excellent data- Data engineering
- Collection (internal –
external)
- Data storage
- ……
Application enabled- Advanced Planning
Systems
- Cloud based tools
- Visualization
From machine to human centred systems
Systems evolve
HCI (Human Computer Interaction)
HMI (Human Machine Interaction)
Human Machine Engineering / Human Machine Design
From machine to …….
The planning systems evolution
HMI (Human Machine Interaction)
HCI (Human Computer Interaction)
Human Machine Engineering / Human Machine Design
?
From machine to…
Changing APS landscape
Strong focus on:• Platform – connecting appilcations• Open architecture – connecting external data • ML / AI capabilities• User interface• Scenario’s, simulations
• Recommendations instead of exceptions (this will change the role of the planner!)
22
No-Touch planning requires different building blocks
EyeOn vision
Integral process for immediate and fact based decision making- Along value chain
- Different functional areas
- Linking strategy to execution
Demand
Making the change
Inventory
E2E, S&OP, IBP
Supply
Excellent data- Data engineering
- Collection (internal –
external)
- Data storage
- ……
Application enabled- Advanced Planning
Systems
- Cloud based tools
- Visualization
High quality analytics- Applying algorithms in
the supply chain.
- Machine learning - AI
- Role of data scientists
- ….
What is all the fuzz about?
AI Taxonomy
Is part of a broader family of machine learning methods based on artificial neural networks. It has been applied to fields including image and speech recognition, natural language processing, audio recognition, machine translation, bioinformatics, drug design, medical image analysis, material inspection and board game programs.
The scientific study of algorithms and statistical models that computer systems use to perform a specific task without using explicit instructions, relying on patterns and inference instead. It is seen as a subset of artificial intelligence
Deep learningMachine learning (ML) Artificial Intelligence (AI)
The study of "intelligent agents": any device that perceives its environment and takes actions that maximize its chance of successfully achieving its goals. Mimic human intelligence.
More and clever machine learning techniques and algorithms
New techniques become available
• Gradient boosting• Generalized linear models• Support vector machines• Nearest neighbors• Decision trees• Neural networks• And so on…
25
No-Touch planning requires different building blocks
EyeOn vision
Integral process for immediate and fact based decision making- Along value chain
- Different functional areas
- Linking strategy to execution
Demand
Making the change
Inventory
E2E, S&OP, IBP
Supply
Excellent data- Data engineering
- Collection (internal –
external)
- Data storage
- ……
Application enabled- Advanced Planning
Systems
- Cloud based tools
- Visualization
High quality analytics- Applying algorithms in
the supply chain.
- Machine learning - AI
- Role of data scientists
- ….
Organizational
readiness- How:
• Is analytics organized?
• Training / coaching
• Changing R&R
- …
26
27
28
29
What will remain?
Trainers • Advanced algorithms need to be trained how to perform• Train algorithms statistical forecast based on historical demand and other available data sources like
Facebook, twitter, google or Point-of-Sales information• Requires deep analytical knowledge• Often centrally organized (cross- functional, size matters)
Explainers • AI delivers conclusions (forecasts, plans) from a black-box• Requires human experts in the field (Demand Planners) to explain the outcome to non-experts
(Sales)• Often decentral, close to the ‘student’
Sustainer • Unconstrained data gathering can harm privacy and break laws • AI uses data several data sources and interacts with humans • Continuously work to ensure that systems work properly, safely, responsibly, within the existing
laws and ethically • Often organized centrally
31
No-Touch planning requires different building blocks
EyeOn vision
Integral process for immediate and fact based decision making- Along value chain
- Different functional areas
- Linking strategy to execution
Demand
Making the change
Inventory
E2E, S&OP, IBP
Supply
Excellent data- Data engineering
- Collection (internal –
external)
- Data storage
- ……
Application enabled- Advanced Planning
Systems
- Cloud based tools
- Visualization
High quality analytics- Applying algorithms in
the supply chain.
- Machine learning - AI
- Role of data scientists
- ….
Organizational
readiness- How:
• Is analytics organized?
• Training / coaching
• Changing R&R
- …
Decision focused
culture• Fact based
collaboration
• Effective meetings
• Reducing cognitive
bias
• ….
The EyeOn Analytics benchmark shows decision making uses human knowledge and simple data and facts, limited use of advanced data analysis
Data driven culture: rely on data!
0 1 2 3 4 5 6 7 8 9 10
Intuition
Personal Experience
Consultation with others
The Hippo
Simple data and facts
Advanced data analysis
Decision making (1 - 10 scale, N = 25)
Understand what people do with it!
Relying on data!
Individual human behavior - bias • Segmentation - differentiation• Close system for small changes (<20%)• Close system for non-value adding updates (FVA)• Use planners for the good (thumbs – up)• Recommendations instead of exceptions
Group decision making• Understand dynamics• Dashboards for instant resolving• Roles in meeting – devil’s advocate
34
No-Touch planning requires different building blocks
Summary - EyeOn vision
• Although it is extremely hyped, analytics really changes planning
• It is often old technology getting accessible / applicable
• Algorithms are demanding and need to be fueled by (a lot of) high quality data
• Though it is coined ‘no-touch’ it makes sense to prepare your organization
• Many small steps make a big difference
Let’s build the future together!
www.eyeon.nl
Eindhoven
Amsterdam
Antwerp
Geneva
DublinFreek [email protected]
T +31 6 29 07 23 87
36
Agenda
16h00 Kick off
16h15 No-touch S&OP – the human factor (Freek Aertsen)
17h00 Getting your inventory health under control (Maarten Van Liempd / Vasco Werners)
17h45 Integrate industry specific constraints & requirements into your planning (Bart Paridaen)
18h30 Dinner
20h00 Wrap-up & Drinks
Inventory Health
Life Science Pre Event
November 13th, 2019
38
Content
• Building blocks inventory management• Stock decomposition• Lead time analysis• Lot size analysis• Inventory Projection• Dashboarding and insights (demo)
Building blocks for inventory management
39
• regular check of actualperformance vs data setting
• clear process to ‘engineer’required parameters
• stage gate process forproduct pipeline
MASTER DATA
40
Transparent process to set theparameters for (groups of) SKUbased on agreed characteristicsand threshold. Appointed processowner.
PARAMETER DECISION
• regular check of actualperformance vs data setting
• clear process to ‘engineer’required parameters
• stage gate process forproduct pipeline
MASTER DATA
Building blocks for inventory management
41
A process with clear ownershipand accountability to review andset targets and identify andimplement optimizationopportunities.Clear who are the ‘suppliers’ and‘customers’ of the process.
INVENTORY MANAGEMENT
Transparent process to set theparameters for (groups of) SKUbased on agreed characteristicsand threshold. Appointed processowner.
PARAMETER DECISION
• regular check of actualperformance vs data setting
• clear process to ‘engineer’required parameters
• stage gate process forproduct pipeline
MASTER DATA
Building blocks for inventory management
42
A process with clear ownershipand accountability to review andset targets and identify andimplement optimizationopportunities.Clear who are the ‘suppliers’ and‘customers’ of the process.
INVENTORY MANAGEMENT
• plan towards targets• proactive management of
projected deviations• clear ‘escalation’ in case of
unexpected events• fact based decisions to
deviate
INVENTORY CONTROL
Transparent process to set theparameters for (groups of) SKUbased on agreed characteristicsand threshold. Appointed processowner.
PARAMETER DECISION
• regular check of actualperformance vs data setting
• clear process to ‘engineer’required parameters
• stage gate process forproduct pipeline
MASTER DATA
Building blocks for inventory management
43
A process with clear ownershipand accountability to review andset targets and identify andimplement optimizationopportunities.Clear who are the ‘suppliers’ and‘customers’ of the process.
INVENTORY MANAGEMENT
• plan towards targets• proactive management of
projected deviations• clear ‘escalation’ in case of
unexpected events• fact based decisions to
deviate
INVENTORY CONTROL
Transparent process to set theparameters for (groups of) SKUbased on agreed characteristicsand threshold. Appointed processowner.
PARAMETER DECISION
• regular check of actualperformance vs data setting
• clear process to ‘engineer’required parameters
• stage gate process forproduct pipeline
MASTER DATA
Building blocks for inventory management
44
Inventory control tools
• Stock decomposition• Lead time analysis• Lot size analysis• Inventory Projection• Dashboarding and insights
Stock Decomposition
45
46
Inventory decomposition
Safety stocks
Cycle
Stock
Regulatory stock
Anticipation stock
Stock type Description Determinants
Cycle stock Stock produced in lots or batches
that fulfills demand over replenishment cycle (time between two production starts)
Lead time, batch size Calculated
Safety stock Required stock to buffer for
uncertainty in demand (order quantity) or supply (lead-time)
Demand variability (Q) and
supply variability (@lead time)
Calculated
Anticipation stock Required stock to anticipate on
future known events
Tenders, quotes, anticipated
factory shut downs for maintenance, etc
Agreed
Regulatory stock Stock required for contractual or
legal reasons
Regulatory (mandatory) and
contractually agreed quantitiesC, SS and A are part of R
Agreed
47
Inventory decomposition shows stock health
Time
Inve
nto
ry
Overstock
Understock
Healthy Stock
Safety stock
Cycle stock
48
Example deliverable: making your inventory transparent
Safety Stock
Cycle Stock
Blocked Stock
Overstock• Excess• Obsolete• Strategic• …
• Returns• Quality Inspection• …
Safety45%
Cycle13%
Blocked23%
Total overstock
19%
Total Stock
49
Drivers for overstock and understock
WIP
Stock in
transit
On Hand Stock
Strategic Stock
Blocked Stock
Active stock
Safety stock
Cycle stock
Excess stock
Active stock
Safety stock Cycle stockExcess stock
• Imbalance safety stock• Lead time uncertainty
• Duplication of safety stock
• Lot size effects• Actual and planned lead
time different• Incorrect ROP
EOLEarlyBias
Pre-buildEtc.
Lead time analysis
50
Introduction lead time
• Analysis of the actual replenishment behavior of both receiving and supplying plants
Relevance for inventory management
• Safety stock: lead-time is an input to the “APICS” formula; purpose to cover demand and
supply uncertainty during the replenishment lead time
• Cycle stock: structural arrival of replenishment orders too early raises the physical cycle
stock.
• Service: structural late arrival of replenishment orders causes stock out and hence service
loss.
51
Impact of differences between actual and master data lead time
Typically in industry is the focus on late deliveries and not on early deliveries.Late deliveries directly impact the service provided.Early arrival of replenishments guarantee availability and have an often neglected working capital impact.
>
52
Early deliveries Late deliveries
52
Inbound lead time
53
Production time Goods Receipt
Actual lead timeArrived
early
Total replenishment lead timeInbound view
• In case replenishment orders from different suppliers structurally arrive earlier than planned at the receiving plant,
there is a high probability that the Goods Receipt component is too high.
• Hypothesis: local inventory targets for receiving plants provide a pervert incentives to plan with longer lead times
and wait with Goods Receipt processing till needed.
Observed behavior inbound view
Expected behavior Unexpected behavior
Loc 1:
Loc 2:
Loc 3:
Loc 4:
OK Late → Early
OK Late → Early OK Late → Early
OK Late → Early
54
How to improve inventory health?
• Insights are key in understanding your inventory! Analyze and visualize your inventory periodically
• Focus is often on optimizing safety stock levels, while cycle stock also makes up big part of total inventory
• Are your master data lead times similar to the actual lead times?
• Are there any strange lot sizes, such as an MOQ of one year of demand
55
Total Stock
56
Drivers for overstock and understock
WIP
Stock in
transit
On Hand Stock
Strategic Stock
Blocked Stock
Active stock
Safety stock
Cycle stock
Excess stock
Active stock
Safety stock Cycle stockExcess stock
• Imbalance safety stock• Lead time uncertainty not
included• Duplication of safety stock
• Lot size effects• Actual and planned lead
time different• Incorrect ROP
EOLEarlyBias
Pre-buildEtc.
Lot sizes
57
High MOQ Effect - stock height driven by MOQ value rather than demand
58
Inve
nto
ry
Replenishment lead-time
ROP
Safety stock
Equal to the demand over the replenishment lead-time
Inve
nto
ry
Replenishment lead-time
ROP
Safety stock
Equal to MOQ
Situation 2: MOQ >> Demand during LT
Situation 1: No MOQ
1. In situation 1, the replenishment quantity is equal to the demand over the lead-time. Assuming that orders are coming in as expected, the next replenishment will be planned after one cycle of replenishment lead-time
2. In situation 2, the replenishment quantity is driven by the MOQ. In case the MOQ is very high, it will take a long time before the inventory has moved to the minimum stock level.
>
In case of high MOQs the safety stock can be relatively low
59
Inve
nto
ry
Situation 2: High MOQ
1. Risk of a stock-out increases at the end of the replenishment cycle2. In case the MOQ is very high, the number of replenishment cycles is limited. Therefore the number of
occasions when the stock is developing towards the minimum stock level is also limited. This effect results in lower safety stock values.
>
Situation 1: No MOQ
Inventory policy for “rare demand of large orders”
60
Production
Product A
Replenishment LT: 2 weeksMOQ: 11,4 kgAvg. demand: 0,84 kg / weekStd. demand: 1,03 kg / weekPolicy: ROP (k = 2,05 → 98%)
Product B
Replenishment LT: 9 weeksMOQ: 75 kgAvg. demand: 0,84 kg / weekStd. demand: 1,03 kg / weekPolicy: non-stocking item
The current policy ignores the effect of the large demand orders. If an order occurs the available stock quantity is not enough and replenishment takes two weeks. Demand is rare, only for 2 out of the 89 weeks there will be shortage →97,7 % service.
Demand for B is a rare event; happens approximately once per 89 weeks (= 75 / 0,84)
If there is demand of B it triggers a demand of 75 kg of A.
Starting point
Implication
• Recommendation: Keep no stock of A (97,7% is also realized with ROP = 0). The current ROP is useless.• Alternative: set the reorder point equal to the demand order quantity (service will be 100%)
Recommendation
Inventory Projection
62
Latest inventory snapshot
Supply Plans
Net Requirement Plans
63
Inventory projection
Root cause analysis on stock development
using inventory projection
• Root cause analysis helps in
understanding main drivers of stock
increase
• Example:
- Detect mismatch between Supply
and Net Requirements per month
or per location to prevent
obsolete stock
64
Stock coverage projection to detect
excess and obsolete cases
• Based on latest Net Requirement
Plans projected inventory can be
categorized in stock:
- Sold within 1 year
- Sold between 1 and 2 years
- Sold after 2 years
- No requirements at all
• Early detection can reduce the future
excess and/or obsolete stock
65
Scenario Analysis
• By changing the original supply or net
requirement plan, the inventory
projection is updated
• Benefits:
- Quantitative arguments in
discussions
- Which adjustment to make to close
the budget gap
66
Inventory Control and
visualization
67
Online demo
68
Go to our website:
https://www.eyeon.nl/insights/powerbi/
for an online demo
68
Let’s build the future together!
www.eyeon.nl
Eindhoven
Amsterdam
Antwerp
Geneva
DublinMaarten van LiempdT: +31 (0)88 883 00 25
replenishment lead time
requested customer lead time
Replenishment lead-time = required time to have new stock available again
Make-to-Order: We have sufficient time available to produce the material and deliver the customer within the requested lead-time
replenishment time
requested customer lead time
Make-to-Stock/Forecast: We don’t have sufficient time available to produce the material and deliver the customer within the requested lead-time. And as a result stock is required to meet the requested lead-time
Lead time vs. stock
72
73
Agenda
16h00 Kick off
16h15 No-touch S&OP – the human factor (Freek Aertsen)
17h00 Getting your inventory health under control (Maarten Van Liempd / Vasco Werners)
17h45 Integrate industry specific constraints & requirements into your planning (Bart Paridaen)
18h30 Dinner
20h00 Wrap-up & Drinks
74
Integrate industry specific constraints & requirements into your planning
Key elements for planning:• Master data strategy• Planning signal• Planning specifics
Focus Topic 2: Flavour management
Focus Topic 1: New product introductions
75
Focus Topic 1: New product introductionsIntroduction
• Disparate systems and data sources
• NPI requires a cross functional approach – involvement of planning
• Forecast consensus – 1 aligned input for planning
• Planning signal visibility – ability to propagate your planning signal upstream
• Assessing the impact of NPIs on existing portfolio (phase in-phase out, cannibalization)
• Ability to simulate E2E impact of NPI scenarios
• Conversion between value and volume
76
Focus Topic 1: New product introductionsMaster data strategy: NPI codes & “early” MD
• Setup of “early” master data to capture forecast signal and propagate upstream in the supply chain• Trade off between effort and speed• Enable local simulation modelling in your planning system
A customer example
78
Focus Topic 1: New product introductionsPlanning signal: consolidate different demand inputs
• Consolidation of different demand inputs (e.g. LRFP, commercial forecast, clinical forecast)
• Agree on demand signal to be used for operational supply planning
• Anticipate upside and downside scenarios to assess RCCP impact
• Estimating and systemizing brand cannibalization impact (predecessor – succesor relations)
79
Focus Topic 1: New product introductionsA customer example – Forecast Integration Session
80
Focus Topic 1: New product introductionsA customer example – Forecast Integration Session
81
Focus Topic 1: New product introductionsA customer example – Forecast Integration Session
82
Focus Topic 1: New product introductionsA customer example – Forecast Integration Session
83
Focus Topic 1: New product introductionsA customer example – Forecast Integration Session
84
Focus Topic 1: New product introductionsA customer example – Forecast Integration Session
85
Focus Topic 1: New product introductionsPlanning specifics – E2E visibility
• Create E2E plan visibility
• Embed regulatory dependencies (approval timings)
• Clinical vs commercial supply
• Mid term RCCP vs operational planning (launch plans)
• Formalize risk based decisions in S&OP
86
Focus Topic 2: Flavour managementIntroduction
87
Focus Topic 2: Flavour managementMaster data strategy: SKUs vs recipes
Key considerations
• Stability of the process
• Stability of the plan
• Reporting at aggregated level
• Data maintenance effort and cost
88
Focus Topic 2: Flavour managementPlanning signal: upstream flavour propagation
• Balance between regulatory constraints and planning decisions
• Capture flavour preferences as part of the demand signal
• Upstream flavour propagation
89
Focus Topic 2: Flavour managementPlanning specifics: balancing flavour demand and supply
Key considerations
• At what level do youtake your decisions?
• A dynamic planning challenge
• Flavour maintenance
• Assess bridging impact of regulatory changes
Let’s build the future together!
www.eyeon.nl
Eindhoven
Amsterdam
Antwerp
Geneva
Dublin
Bart ParidaenT: +31 (0)88 883 00 12
Many different nationalities, one team!
EyeOn as a partner
91
500+Years of combined experience
12Nationalities Consultants with different backgrounds: SCM, finance, sales & marketing, data science
89Companies where projects were delivered in 2018
10Countries where we operated in 2018