IE 8580 Module 2: Transportation in the Supply Chain - PowerPoint Presentation

Slide1

IE 8580 Module 2:Transportation in the Supply Chain

Lecture

2.1:

Transportation Networks

Slide2

The Bis Corporation

Paint manufacturer

Established in 1964

8 manufacturing plants17 warehouses2,000 retail stores4,000 SKUsOwned by 12 shareholders and run by a new CEO

IE 8580, mason@clemson.edu

2

Slide3

Problem

Although a high profit margin business (gross margin of 20%), idea is that the distribution system could be improved

Designed 15 years ago

Produce and store at manufacturing plantsPick, load, and ship to warehouseUnload and store at the warehousePick, load, and deliver to storesNeed to

re-engineer their distribution and supply chain network

IE 8580, mason@clemson.edu

3

Slide4

Design Requirements

What is the best network configuration?

1-tier network not efficient

Given the network, where should the inventory be held?4,000 SKUs – not obvious how to position inventoryWhich plant should produce which product?IE 8580, mason@clemson.edu

4

Slide5

Data

Aggregate demand

Grouped customers into 550 zones and products into 5 families

Collect dataDemand by SKU per product family for each zoneAnnual production capacity by SKU’s at each facility Max capacity by SKU at each warehouseTransportation costs per product family per mile between all production facilities and warehousesSetup cost to establish a new warehousePotential locations for new warehouses

IE 8580, mason@clemson.edu

5

Slide6

Other Relevant Information

CEO requires delivery time from warehouse to customer < 24 hours

Implies distance between warehouse and customer zone <450 miles

Growth industry with 7%, 3%, 6%, 5%, and 6% annual growth rate projected for families 1 through 5, respectivelyQuestionsShould Bis keep current strategy or switch to 2-tier?Is the model good enough to trust the results?What is the optimal inventory positioning?

Should production of SKUs be consolidated?

IE 8580, mason@clemson.edu

6

Slide7

What is Network Planning?

Way

a firm structures and manages the physical supply chain – suppliers, plants, warehouses, distribution centers,

cross-docking facilitiesTypically, there are three stepsNetwork design: number, location, size of plants, warehouses, DC’s

Inventory positioning: where to keep stock, how muchResource allocation: capacity allocation, outsourcing, offshoring

IE 8580, mason@clemson.edu

7

Slide8

General Network Design

Mathematical models are used to identify

Supply points: suppliers and plants

Distribution: warehouses, DC’s, cross-dockingConnecting pathwaysAssumptions are used to make the problem more tractableBecoming less critical with better computing and approximation algorithms

IE 8580, mason@clemson.edu

8

Slide9

Focus on Warehouses

Admittedly this is a very simple example but it lets us

work

all the way through a scenarioAdapting this “strategy” to your environment will require some thinking and creativity … and maybe working with a university!

Assume supplier and customer locations are fixedDetermine

Number of warehouses

Location of warehouses

Size of warehouses

Space allocation for warehouses

Assignment of products to warehouses to customers

IE 8580, mason@clemson.edu

9

Slide10

Increasing the Number of Warehouses …

Improves service level

Increases inventory cost

Increases overhead and start-up costsReduces outbound transportation costsIncreases inbound transportation costsWhat is the right thing to do?IE 8580, mason@clemson.edu

10

Slide11

We Need Data—Some is Easy to Get/Accurate, Some Is Not

Locations

customers, warehouses, DC’s, production facilities and suppliers

Product information Volumes and special transport considerationsDemandsExisting and forecastedTransportation costsWarehouse costsLabor, inventory holding, material handling, etc.

Delivery restrictionsShipment sizes and frequency of deliveriesCustomer service requirements

IE 8580, mason@clemson.edu

11

Slide12

If your problem is big, effective data aggregation is critical

Customers can often be aggregated according to location

According

to 3 or 5 digit ZIP codeProducts can be aggregate according to similaritiesDistribution patterns or product typesThis sword has two edgesAdvantages:

smoothes variability, produces a tractable problemDisadvantage: easy to over-do

and get meaningless answers

IE 8580, mason@clemson.edu

12

Slide13

Ground transportation rates in the US are available

Internal fleet – (cost/mile) x miles

External fleet

TLeach state is a zone (except FL, CA, TX)cost per mile established from each zone to all other zones

LTLrates based on class of cargo

ZIP code

to

ZIP code

There are ways to estimate

mileage

Straight line for short distances; long distance correction for earth’s curvature

“Fudge factor” because roads are not straight

Google Earth/Maps API calls

IE 8580, mason@clemson.edu

13

Slide14

Warehouse costs are higher than most folks think

Handling costs

Labor

and utilityTypically proportional to annual flow through

warehouse

Fixed costs

Equipment

and rent, for example

Not

related to volume

Storage costs

Inventory

holding costs typically computed as proportional to the average inventory level

Finding

the average inventory level for a particular customer is sometimes difficult – one solution is the inventory turnover ratio

IE 8580, mason@clemson.edu

14

Slide15

Other warehouse related items

Capacity

Inventory

turnover ratio can help to estimatePotential locationsGeographical and infrastructure locationsNatural resources and laborTaxes and incentives

Service level requirementsMinimum distance or maximum time

Future demand considerations

IE 8580, mason@clemson.edu

15

Slide16

Model building and validation

Use most appropriate tool

OR model, simulation

Solve for plant conditions and verify output is reasonable approximation of realityBe careful of extrapolationUse this as a starting point for decision making

IE 8580, mason@clemson.edu

16

Slide17

Example (Page 92, Simchi Levi et al.)

Single product, two plants, two warehouses, 3 customers

Plant capacities: 140,000 and 60,000

Customer demands: 50,000, 100,000, and 50,000Unit distribution costs ($)

P1

P2

C1

C2

C3

W1

0

4

3

4

5

W2

5

2

2

1

2

IE 8580, mason@clemson.edu

17

Slide18

The Basic Problem

Plant 1

W/H 1

Customer 1

0

5

4

2

3

4

5

2

1

2

Plant 2

W/H 2

Customer 2

Customer 3

140

60

50

100

50

IE 8580, mason@clemson.edu

18

Slide19

Surely you can do this one in your head, right?Who needs math, anyways?!?!?

IE 8580, mason@clemson.edu

19

Slide20

Approach #1

Since shipping from W/H 2 is cheaper than W/H 1

for both customers, use

only W/H 2.

Plant 1

W/H

1

Customer 1

0

5

4

2

3

4

5

2

1

2

Plant 2

W/H 2

Customer 2

Customer 3

140

60

50

100

50

140

60

50

100

50

IE 8580, mason@clemson.edu

20

Slide21

Approach #2

Select the path with the minimum cost starting with Customer 1 and proceed sequentially.

Plant 1

W/H

1

Customer 1

0

5

4

2

3

4

5

2

1

2

Plant 2

W/H 2

Customer 2

Customer 3

140

60

50

100

50

50

6

0

90

IE 8580, mason@clemson.edu

21

Slide22

Resulting Costs

Heuristic 1

140K units*$5/unit+60K*$2+50K*$2+100K*$1+50K*$2 = $1,120,000

Heuristic 2

50K units*$0/unit+90K*$5+60K*$2+50K*$3+100K*$1+50K*$2 = $920,000

IE 8580, mason@clemson.edu

22

Slide23

Are you going to use and bet your career on:

1. opinions based on experience and gut feel

or

2. hardcore, mathematical analysis

So How Well Do You Trust Yourself?

IE 8580, mason@clemson.edu

23

Slide24

Comparisons

Heuristic 1: $1,120,000

Heuristic 2: $ 920,000

Optimal: $740,000

You just overpaid by:

51%

if you used Heuristic 1

25%

if you used Heuristic 2

IE 8580, mason@clemson.edu

24

Slide25

Simulation and OptimizationOptimization

Models are typically inferior because assumptions have to be made for tractability

Solutions are determined; that is, you don’t have to guess them – you just develop the model

Solutions can be great starting points for system designSimulationModels can be quite accurate with many of the stochastic elements capturedIt only shows system performance based on the controls that you guessCan be of great value when you have a really good starting point and want to “tune” things or play “what if”

IE 8580, mason@clemson.edu

25

Slide26

Scope of Network ConfigurationIn practice, there are both greenfield and brownfield requirements

Greenfield

Complete new design because you are just starting or because the existing structure has been modified so much that a fresh approach is required

BrownfieldMust modify within the existing structure or include new features like adding warehouse-to-warehouse flow or customer specific service level requirementsYour authors call the ability to accommodate this range “flexibility” and it is critical to network configuration design

IE 8580, mason@clemson.edu

26