Forecasting & Planning

Forecasting
• Actual demand when available
• Forecasts
– When lead time is insufficient
– On push side of push-pull frontier
• Logistics rarely forecasts except input on
constraints
• Term (shorter terms are more reliable)
– Long-term (3+ years)
– Medium-term (1 year or so)
– Short-term (less than 1 year)
 Nature of Forecasting
• Future is uncertain
• “Forecasts are always wrong”
(track accuracy data)
• Bias
• Benefit-cost
• Aggregate more reliable than item forecast
• Check data sources and simplify models
• Use demand history instead of sales history
• Consumer demand not intermediate customer
 Bullwhip Effect
• 1960’s “beer game”: Lack of communication root
cause
• Distribution center order treated as demand
(safety stock, overstock)

Supplier Factory
Distributor Retailer

Customer
Small demand uncertainty becomes more and
more distorted.
Factors Affecting Demand
Trend line (linear)
Seasonal
peaks
Actual demand line

Demand Base demand

(average, here
over four years)

Random variation

Year 1 Year 2 Year 3 Year 4

Time
 Forecasting Process
Step 1: Step 8: Step 9:
Determine purpose Forecast Perform sales and
operations planning (S&OP)

Step 2:
Step 7:
Determine aggregation and Step 10:
Test against historical data
units Periodically review and
improve models for accuracy

Step 3: Step 6:
Determine time horizon Prepare data

Step 4: Step 5:
Visualize data Choose method or model
 Forecasting Methods
Qualitative Quantitative

• Beware of bias • Time-series:

• Judgmental/expert – Naïve
judgment – Simple moving average
• Delphi method – Weighted moving
average
– Expert opinions collated
anonymously – Exponential smoothing
– Successive rounds for
consensus
– Groupthink and “stake in
the ground”
Deseasonalizing

Data for
next two
slides
Deseasonalized values
closer to year average
Simple and Weighted Moving Average
Smooths out random spikes or dips:

(M1 + M2 + M3 )
3-Month Moving Average =
3

(15 + 12.96 + 14 )
January 2016 forecast = = 13.99 units
3

If recent periods are better predictors:

(1  M1) + ( 2  M2 ) + ( 3  M3 )
3-Month Weighted Moving Average =
6

(1  15 ) + ( 2  12.96 ) + ( 3  14 )
January 2016 forecast = = 13.82 units
6
Sum of weights
Exponential Smoothing and Re-seasonalizing
• Smoothing constant (alpha, α): 0–1
(percentage)

New Forecast = (  Last Period's Demand ) + (1 −  )  Last Period's Forecast 

January 2016 forecast = ( 0.3  14 ) + (1 − 0.3 )  16.2 = 15.54 units

30% weight 70% weight

• Reseasonalize
– 15.54 units × 2.214 (January 2015 seasonal index)
= 34.4 units, the forecast for January 2016
Quantitative: Associative (Causal)
Simple Regression
• Independent (predictor) and dependent variable (predicted)
y =  + x

Roofing Sales =  + (   Prior Month's Housing Starts )

Roofing Sales
USD 1,600,000
USD 1,400,000
USD 1,200,000
USD 1,000,000
USD 800,000
β (slope, here USD per housing start)
USD 600,000
USD 400,000
α (intercept)
USD 200,000
USD 0
0 20 40 60 80 100 120 140 160
Housing Starts
Coefficient of Correlation (r)
• A statistical measure of the degree to which
changes to the value of one variable predict
change to the value of another
• Range of values between -1.0 and +1.0
r = 1.0 is perfect positive correlation

r = 0.0 is not correlated at all

r = -1.0 is perfect negative correlation

 How Logistics Uses Forecasts
3-5 Year Annual Forecast
Forecast
▪ Size and number of ▪ Staffing levels and
warehouses equipment
▪ Average shipments ▪ Capacity to book with
per shipping line carriers in request for
proposal (RFP) or
invitation to tender
(ITT)
Operational
Forecast
▪ Worker, warehouse
capacity constraints
▪ Number of vehicle
payloads given minimum
orders, weight
restrictions, etc.
 Accuracy (Error Rates)
• Absolute deviation:
Forecast Error = Actual – Forecast
– stated as an absolute value
MAD = A − F
• Mean absolute deviation (MAD): n
Average of absolute deviations
• Mean squared error (MSE): Sum of (Errors for Each Period)
2

MSE =
Number of Forecast Periods
Average of errors squared then summed
• Mean absolute percentage error (MAPE):
 A −F 
Average of summed forecast error   A  %
MAPE =  
percentages n
 Accuracy (Error Rates)
• Tracking signal:
Used to indicate the existence of any
positive or negative bias in a forecast
– Algebraic sum of forecast errors, divided by
the mean absolute deviation value

Algebraic Sum of Forecast Errors

Tracking Signal =
MAD

• Exceptions:
Outliers, could be errors or not
Manufacturing/Distribution Balancing
$
External Balancing Internal Balancing
• Price • Production flexibility
• Lead time • Inventory holding
• Both promote substitutes – Safety stock helps to prevent
(yours or others) stockouts, but makes
oversupply issues worse
• Marketing methods
– Sales incentives
– Trade discounts
– Consumer promotions
 Logistics Balancing

$
External Balancing Internal Balancing
• Switch from manual to EDI • Reconfigure transportation
ordering routes or modes
• Offer substitutes located • Adjust stock levels at
closer distribution centers
• Offer free shipping • Drop ship from alternative
• Leaving trailers at distribution sites or channels
customers to minimize back • Tracking devices for more
haul detailed and accurate
tracking
Distribution Channels
Production

Retail

Business-to-
Business-to- Consumer (B2C)
Business (B2B)

Ecommerce

Supplier to
Customer
 Operations Planning and Control
Strategic Planning

Forecasting Business Planning

S&OP: Consensus Demand Resource Planning

Demand Management
Plan and Production Plan (RP)

Distribution Master Scheduling (MS): Rough-Cut Capacity

Requirements Planning Master Production Planning (RCCP)
(DRP) Schedule(MPS)

Material Requirements Capacity Requirements

Planning (MRP) Planning (CRP)

Production Activity Final Assembly

Purchasing Capacity Control
Control (PAC) Scheduling (FAS)
 Supply Planning
• Production planning

Production Rate =
(Ending Inventory - Beginning Inventory ) + Forecast
Number of Periods

=
(12,000-10,000 ) + 100,000
= 8,500 units per month
12
• Resource planning
• Inventory planning
• Distribution requirements planning (DRP)
• Performance metrics and targets
 Resource Planning
Resource Profile Units Capacity (Monthly)
Condenser Shop Hours 35,000
Final Assembly m3 8,000

Bill of Resources Units Family A Family B Family C

Condenser Shop Hours 3 5 5
Final Assembly m3 0.6 1.2 1.4

Resource Units Family Family Family Total Load Capacity Load vs.
Plan A B C Capacity
Jan. Plan Units 5,000 2,000 1,500 8,500 n/a n/a
Condenser
Hours 15,000 10,000 7,500 32,500 35,000 92.9%
Shop
Final
m3 3,000 2,400 2,100 7,500 8,000 93.8%
Assembly
Source: Adapted from David F. Ross, Distribution Planning and Control—Managing in the Era of Supply Chain Management, Third Edition.
Fulfillment Center Capacity Bar Chart
Fulfillment Center Capacity
10000 Exceeds Capacity
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Released
ReleasedLoad
Load Planned
Planned Load
Load
Demand Management Process
Demand Plan Dashboard----Units
Demand for Product Family XYZ in Units
15,000

10,000

5,000

0
Units
–3 –2 –1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Historical Demand Demand Plan for 18 Months in Future

Annual business plan

Product family XYZ:
Key: Prior demand plan Lead time = 2 weeks
Actual Demand Current demand plan Inventory turnover = 2 weeks
Positively Influencing Demand
• Develop products that customers are demanding
• Determine profitable product mix (Range Optimization)
• Set strategic pricing
• Place products at distribution points to establish
presence and customer convenience
• Promote products
• Support customer expectations and needs
• Support organization’s business objectives
 Managing and Prioritizing Demand
• Policies promote optimum profit and service
• Prioritize by customer value
• Rationing, queues, substitute incentives
• Retain scarce inventory at central supply longer
• Time fences reduce oversupply
• Supply and demand evaluate custom orders
• Management prioritizes, not salespersons
• Fulfill all demand if feasible and adds marginal profit
Supply Planning Phase Meeting

Supply/Demand Supply/Demand
Match Mismatch

▪ Production plan matches demand ▪ Supply develops alternative plans:

plan. − Produce above demand to meet
later spikes.
− Increase capacity by hiring, adding
shifts, overtime, equipment, or
outsourcing (or opposite).
− Reduce demand plan (last resort).
 Monthly S&OP Process
Step 1:
Data Gathering Statistical forecast updated.

Step 2: Statistical forecast reviewed by product/brand,

Demand Planning marketing, sales.

Step 3: Supply management team may alter operations

Supply Planning plan if necessary.

Step 4: Key players review data, set executive meeting

Pre-Meeting agenda.

Step 5: VPs meet monthly to review decisions and

Executive Meeting strategy.
Controlling Priorities: Master
Scheduling
Master Scheduling (MS) Grid
Frozen zone Slushy zone Liquid zone

Period 1 2 3 4 5 6 7 8 9 10
Forecast 20 22 21 25 24 23 21 21 25 25
Customer orders 19 17 15 11 9 5 2 1 0 0
PAB 50 31 14 49 24 0 27 6 35 10 35
ATP 14 15 43 49
MPS 50 50 50 50

Demand Time Planning Time

Fence (DTF) Fence (PTF)
Materials Requirements Planning
Master What to produce?
production How many?
schedule When needed?

Planning
factors Material Requirements Planning Inventory
status
• Gross-to-net calculations
• Planned order recommendations On hand?
Bills of • Exception action messages On order?
Lead time?
material
Safety stock?

Parts to make? Planned Tells us:

Quantities? production/ Ordered?
purchase Orders to expedite?
Orders to cancel?
orders Orders to de-expedite?
 Inventory Planning
Family (in units) Family A Mini-Refrigerator
Period 0 1 2 3 4 5 6 7
Forecast 5,200 5,400 4,900 4,700 4,800 5,100 5,000
Production Plan 5,000 5,000 5,000 4,867 4,867 4,867 7,000
Ending Inventory Plan 1,500 1,300 900 1,000 1,167 1,233 1,000 3,000
Qtr. Inventory Target 1,000 1,000
Max Inventory (OK?) 2,000 OK OK OK OK OK OK FIX
Min Inventory (OK?) 1,000 OK FIX OK OK OK OK OK

Source: Adapted from David F. Ross, Distribution Planning and Control – Managing in the Era of Supply Chain Management, 3rd edition.

Production Rate =
(Ending Inventory - Beginning Inventory ) + Forecast
Number of Periods

=
(1,000 - 1,500 ) + (5,200 + 5,400 + 4,900 ) = 5,000 units per month in Q1
3
 Logistics Capacity Check
Capacity
Tactical-level check
• Warehouse capacity – aggregate by storage type
• Labor and equipment capacity and load – set standard hours
by product family
– Two categories: receiving and put-away, picking and shipping

• Transportation volume – information from distribution

requirements planning (DRP)
Distribution Requirements Planning
Forecasting Planned order releases

Where selling?

Distribution Requirements Planning Inventory

Distribution status
network • DC resupply requirements
• Central supply/manufacturing
plans production On hand?
How many?
• Gross to net On order?
• Planned order recommendations Lead time?
• Exception messages Safety stock?
Customer • Inputs to MPS
orders

Payloads to order? Ordered?

Drivers, handlers? Logistics Resupply Orders to expedite?
DC space? capacity action plan cancel?
Shipping lanes? requirements de-expedite?
Distribution Requirements Planning
Push Hybrid systems Pull
systems (e.g., DRP) systems

Forecasts and schedules Push to given Partners determine

centrally coordinated. echelon, pull from own orders.
there, use retail
demand data. Drawbacks
Drawbacks Bullwhip effect if partners
Customers don’t don’t collaborate.
Coordination
determine own orders. and control. Doesn’t account for needs
Doesn’t account for of other SC partners.
local conditions. Responsive to
Ignores supplier’s ability.
local demand.
DRP Grid, Prior to Planned Orders
DRP Grid Safety Stock: 0 units
Min. order quantity: 50 units
Lead time: 2 weeks
Week 1 2 3 4 5 6 Lot size: 50 units
Gross Requirements 110 110 110 110
Scheduled Receipts 100
PAB 170 60 50 -60 -170
Net Requirements 0 0 60 170
Planned Order Receipts 0 0
Planned Order Releases 0 0

PAB = Beginning Inventory or Prior Period PAB + Scheduled Receipts

+ Planned Order Receipts – Gross Requirements
DRP Grid, with a Planned Order
Safety Stock: 0 units
DRP Grid Min. order quantity: 50 units
Lead time: 2 weeks
Lot size: 50 units
Week 1 2 3 4 5 6
Gross Requirements 110 110 110 110
Scheduled Receipts 100
PAB 170 60 50 40 -70
Net Requirements 0 0 60 70
Planned Order Receipts 0 0 100
Planned Order Releases 100 0

Net Requirements = Beginning Inventory or Prior Period PAB

+ Scheduled Receipts – Gross Requirements
 DRP Grid, Completed Safety Stock: 0 units
Min. order quantity: 50 units
DRP Grid Lead time: 2 weeks
Lot size: 50 units
Week 1 2 3 4 5 6
Gross Requirements 110 110 110 110 110 110
Scheduled Receipts 100
PAB 170 60 50 40 30 20 10
Net Requirements 0 0 60 70 80 90
Planned Order Receipts 0 0 100 100 100 100
Planned Order Releases 100 100 100 100 100 150

Given safety stock, rather than if PAB would go negative,

planned orders are triggered when PAB would go below
safety stock level.
 DRP Logic
(Lead time = 1 week) (Lead time = 2 weeks)
DC A: Week ~ 6 7 DC B: Week ~ 6 7 8
Gross Reqs. 300 Gross Reqs. 500
PAB 170 170 270 PAB 200 200 200 200
Net Requirements 130 Net Requirements 300
Planned Order Receipts 400 Planned Order Receipts 500
Planned Order Releases 400 Planned Order Releases 500

Central Supply: Wk. ~ 3 ~ 5 6 7 MS Grid: Week ~ 2 3 4

Gross Reqs. 900 Gross Reqs. 600
PAB 500 500 500 200 200 PAB 0 200 200
Net Requirements 400 MPS 800
Planned Order Rec. 600
Lot sizes: DC A: 400
Planned Order Rel. 600
DC B: 500 Central: 600
(Lead time = 3 weeks)
Source: APICS CPIM Basics of Supply Chain Management
Exceptions and Ordering Policies

Exception Action Messages

DRP and
Order Policy
• Releases • Lot-for-lot
• Lead-time violations • Lot-for-lot above minimum
• Cancel notices quantity
• Expedite scheduled receipts • Lot size quantities
• De-expedite scheduled • Fixed period requirements
receipts • Min-max
• Economic order quantity